LG Electronics, Inc.v.ATI Technologies ULC

Patent Trial and Appeal Board

Apr 14, 2016

11746453 (P.T.A.B. Apr. 14, 2016)

Trials@uspto.gov Paper 62
571-272-7822 Entered: April 14, 2016

UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

LG ELECTRONICS, INC.,
Petitioner,

v.

ATI TECHNOLOGIES ULC,
Patent Owner.
____________

Case IPR2015-00325
Patent 7,742,053 B2
____________

Before JONI Y. CHANG, BRIAN J. McNAMARA, and
RAMA G. ELLURU, Administrative Patent Judges.

CHANG, Administrative Patent Judge.

FINAL WRITTEN DECISION
35 U.S.C. § 318(a) and 37 C.F.R. § 42.73

IPR2015-00325
Patent 7,742,053 B2

2
I. INTRODUCTION
LG Electronics, Inc. (“LG”) filed a Petition requesting an inter partes
review of claims 1, 2, and 5–7 (“the challenged claims”) of U.S. Patent No.
7,742,053 B2 (Ex. 1001, “the ’053 patent”). Paper 2 (“Pet.”). Patent
Owner, ATI Technologies ULC (“ATI”), filed a Preliminary Response.
Paper 12 (“Prelim. Resp.”). Upon consideration of the Petition and
Preliminary Response, we instituted this trial as to claims 1, 2, and 5–7 of
the ’053 patent on June 15, 2015. Paper 13 (“Dec.”).
Subsequent to institution, ATI filed a Patent Owner Response
(Papers 21, 22, “PO Resp.”); LG filed a Reply to the Patent Owner Response
(Papers 33, 34, “Reply”); and ATI filed a sur-reply to LG’s Reply with
respect to the antedating issue (Papers 39, 40).1 An oral hearing was held on
February 10, 2016.2
We have jurisdiction under 35 U.S.C. § 6(c). This Final Written
Decision is issued pursuant to 35 U.S.C. § 318(a). For the reasons discussed
herein, and in view of the record in this trial, we determine that LG has
shown by a preponderance of the evidence that claims 1, 2, and 5–7 of the
’053 patent are unpatentable.

1 The parties filed a confidential version and a redacted version of their
papers. The Decisions denying the parties’ Motions to Seal these documents
and supporting evidence are entered concurrently with this Final Written
Decision. Papers 63, 64. The citations to these papers are to the unredacted
versions.
2 A transcript of the oral hearing is entered in the record as Paper 61 (“Tr.”).
IPR2015-00325
Patent 7,742,053 B2

3
A. Related Matter
The ’053 patent is asserted in Advanced Micro Devices, Inc. v. LG
Electronics, Inc., No. 3:14-cv-01012-SI (N.D. Cal.). Pet. 1.
B. The ’053 Patent
The ’053 patent discloses a computer system for multithreaded
graphics processing. Ex. 1001, 2:36–41. The system includes a memory
device for storing command threads and an arbiter for providing a command
thread to a command processing engine, based on a priority scheme. Id. at
2:48–52, 3:29–35; see Paper 13, 2–3.
C. Illustrative Claim
Of the challenged claims, claims 1 and 5 are independent. Claim 2
depends from claim 1, and claims 6 and 7 depend directly from claim 5.
Claim 5, reproduced below, is illustrative of the challenged claims.
5. A graphics processing system comprising:
at least one memory device comprising a first portion operative
to store a plurality of pixel command threads and a second
portion operative to store a plurality of vertex command threads;
an arbiter, coupled to the at least one memory device, operable
to select a command thread from either of the plurality of pixel
command threads and the plurality of vertex command threads;
and
a plurality of command processing engines, coupled to the
arbiter, each operable to receive and process the command
thread.
Ex. 1001, 8:4–15 (emphases added).
IPR2015-00325
Patent 7,742,053 B2

4
D. Prior Art Relied Upon
LG relies upon the following prior art references:
Lindholm US 7,015,913 B1 Mar. 21, 2006 (Ex. 1004)
Stuttard US 7,363,472 B2 Apr. 22, 2008 (Ex. 1005)
Moreton US 7,233,335 B2 June 19, 2007 (Ex. 1006)
Whittaker US 5,968,167 Oct. 19, 1999 (Ex. 1007)
Kimura US 6,105,127 Aug. 15, 2000 (Ex. 1008)
Admitted Prior Art – Figure 1, and the Background of the Invention
Section of the ’053 patent. Ex. 1001, 1:22–2:6, Fig. 1.
E. Instituted Grounds of Unpatentability
We instituted this trial based on the following grounds (Dec. 36–37):
Claims Basis References
5–7 § 102(e) Moreton
1 and 2 § 103(a) Moreton and Whittaker
1, 2, and 5–7 § 103(a) Lindholm in view of the Admitted Prior Art
1, 2, and 5–7 § 103(a) Stuttard in view of the Admitted Prior Art

II. ANALYSIS
A. Claim Construction
In an inter partes review, claim terms in an unexpired patent are given
their broadest reasonable construction in light of the specification of the
patent in which they appear. 37 C.F.R. § 42.100(b). Under the broadest
reasonable interpretation standard, claim terms are given their ordinary and
customary meaning as would be understood by one of ordinary skill in the
IPR2015-00325
Patent 7,742,053 B2

5
art in the context of the entire disclosure. In re Translogic Tech., Inc., 504
F.3d 1249, 1257 (Fed. Cir. 2007).
“command thread”
Each of independent claims 1 and 5 recites “at least one memory
device comprising a first portion operative to store a plurality of pixel
command threads and a second portion operative to store a plurality of
vertex command threads.” Ex. 1001, 7:11–15, 8:5–8 (emphases added).
Before institution, ATI urged us to construe “command thread” as “a
sequence of commands.” Prelim. Resp. 12–13. ATI also argued that a
command thread does not encompass an instruction. Id. at 12.
In the Decision on Institution (Dec. 6–7), we noted that the word
“command” is used in the Specification of the ’053 patent consistent with its
plain and ordinary meaning, as including an instruction. See, e.g., Ex. 1001,
4:21–27; MICROSOFT COMPUTER DICTIONARY 111 (5th ed. 2002) (Ex. 3001)
(defining “command” as an “instruction to a computer program that, when
issued by the user, causes an action to be carried out”). Notably, the
Specification discloses that “[u]pon the execution of the associated
command of the command thread, the thread is thereupon returned to the
station 302 or 304 at the same storage location with its status updated, once
all possible sequential instructions have been executed.” Ex. 1001, 4:21–27
(emphasis added). Dr. Nader Bagherzadeh testifies that, in the context of
computer multithreading, a stream of instructions is called a thread.
Ex. 1003 ¶¶ 23–24. This is consistent with the usage of the word “thread” in
IPR2015-00325
Patent 7,742,053 B2

6
the prior art of record. See, e.g., Ex. 1005, 5:19–30. We further note that
the plain meaning of “thread,” in the context of computer programming,
means a process that is part of a larger process or program. MICROSOFT
COMPUTER DICTIONARY 518 (5th ed. 2002) (Ex. 3001). We, therefore,
disagreed with ATI, in our Decision on Institution, that a command thread
does not encompass an instruction, as it would be inconsistent with the
term’s plain and ordinary meaning. Rather—for purposes of the Decision on
Institution—in light of the Specification, we construed the claim term
“command thread” to encompass a stream of instructions or a process that is
part of a larger process or program. Dec. 6–7.
After institution, ATI does not challenge our claim construction. PO
Resp. 30. In fact, ATI’s expert, Dr. Andrew Wolfe, testifies that one of
ordinary skill in the art would have understood that the term “command
thread” requires instructions. Ex. 2151 ¶¶ 57–58, 118. As such, we discern
no reason to change our claim construction of “command thread” for this
Final Written Decision.
“arbiter”
Each of independent claims 1 and 5 recites “an arbiter, coupled to the
at least one memory device, operable to select a command thread from
either of the plurality of pixel command threads and the plurality of vertex
command threads.” Ex. 1001, 7:16–19, 8:9–12.
In its Petition, LG proposes to construe the claim term “arbiter” as
“any implementation of hardware and/or software that receives and provides
IPR2015-00325
Patent 7,742,053 B2

7
a thread.” Pet. 9. As support, LG cites to the Specification, which explains
that an “arbiter may be any implementation of hardware, software or
combination thereof such that the arbiter receives the command thread and
thereupon provides the command thread to a command processing engine.”
Id. (citing Ex. 1001, 2:48–52).
Before institution, ATI argued that LG’s proposed construction
“ignores arbitration,” and proposed that the claim term “arbiter” should be
construed as “a component for picking out a command thread among
available pixel and vertex command threads.” Prelim. Resp. 10–11. ATI’s
proposed construction, however, improperly would import other claim
language—“picking out a command thread among available pixel and vertex
command threads”—into the construction of the claim term “arbiter.” Such
a construction also would render other claim language superfluous—e.g.,
“select a command thread from either of the plurality of pixel command
threads and the plurality of vertex command threads,” recited in claim 5.
Moreover, the Specification explains that “arbiter 204 retrieves a command
thread via connection 214 and provides the retrieved command thread to the
command processing engine.” Ex. 1001, 3:8–10 (emphasis added). As
such, we declined to adopt ATI’s proposed construction. Rather—for
purposes of the Decision on Institution—in light of the Specification, we
construed the claim term “arbiter” as any computer hardware, software, or
combination thereof that receives and provides a command thread. Dec. 7–
8. After institution, neither party proffers a different construction for this
term. See PO Resp.; Reply. Upon review of this record, we discern no
IPR2015-00325
Patent 7,742,053 B2

8
reason to change our claim construction of “arbiter” for this Final Written
Decision.
“command processing engine”
Claim 5 recites “a plurality of command processing engines, coupled
to the arbiter, each operable to receive and process the command thread.”
Ex. 1001, 8:13–15 (emphasis added). Claims 6 and 7 directly depend from
claim 5, and further recite “wherein the plurality of command processing
engines comprises at least one arithmetic logic unit” and “at least one
texture processing engine,” respectively. Id. at 8:16–21 (emphases added).
LG proposes to construe “command processing engine” as “any
implementation of hardware and/or software that processes commands.”
Pet. 9. In its Patent Owner Response, ATI asserts that an ordinarily skilled
artisan would have understood that the “command processing engine”
limitation recited in claim 5 requires each command processing engine to be
able to process both pixel and vertex command threads. PO Resp. 31;
Ex. 2151 ¶ 125.
LG disagrees, arguing that the disputed limitation recited in claim 5
merely requires that the command processing engines are capable of
processing either a pixel command thread or a vertex command thread
because claim 5 recites “an arbiter . . . operable to select a command thread”
and “a plurality of command processing engines . . . each operable to receive
and process the command thread.” Reply 14 (emphases added by LG).
IPR2015-00325
Patent 7,742,053 B2

9
Upon review of the claim language and Specification, we agree with
LG. Nothing in the claim language requires a single command processing
engine to be able to process both pixel and vertex command threads.
We note that ATI’s proposed claim construction essentially requires
each of the plurality of command processing engines to have the capability
to process all of the command threads selected by the arbiter—excluding
type-specific processing engines. The Specification of the ’053 patent,
however, does not support such a narrow construction. Notably, the
Specification discloses that the “command processing engine may be any
suitable engine as recognized by one having ordinary skill in the art for
processing commands, such as a texture engine, an arithmetic logic unit, or
any other suitable processing engine.” Ex. 1001, 2:59–62 (emphasis added).
Additionally, claims 5–7 are directed to the preferred embodiment illustrated
in Figure 4 of the ’053 patent, which is reproduced below.

IPR2015-00325
Patent 7,742,053 B2

10
As shown in Figure 4, processing system 300 includes first
reservation station 302, which contains pixel command threads; second
reservation station 304, which contains vertex command threads; arbiter
306; arithmetic logic unit 308; and texture engine 310. Id. at 3:62–4:33.
The Specification explains that arbiter 306 retrieves pixel command thread
324 and vertex command thread 326 and then provides thread 328, which
may be either pixel command thread 324 or vertex command thread 326, to
texture engine 310. Id. at 4:28–33. Arbiter 306 further provides the other
thread 330 to arithmetic logic unit (“ALU”) 308. Id. Upon execution of the
command, ALU 308 and text engine 310 return command threads 332 and
334 to the appropriate reservation station 302 or 304. Id. at 4:34–41. As the
Specification explains, multiple command operations may be performed by
ALU 308 or texture engine 310, but, in order to switch a thread from ALU
308 to texture engine 310, that thread must be returned back to the
appropriate reservation station 302 or 304 and re-retrieved by arbiter 306
and thereupon provided to the other unit 308 or engine 310, respectively. Id.
at 4:42–48.
Significantly, the Specification indicates that a thread is provided to a
specific type of command processing engine (e.g., an ALU or texture engine)
based on the type of operations (e.g., an ALU or texture operation). Id. at
3:62–4:48. ATI’s own expert, Dr. Wolfe, testifies that “vertex command can
also involve texture operations.” Ex. 2151 ¶¶ 159, 165 (noting that “at least
10–20% of vertex command threads involving texture operations should be
processed by the texture fetch processor”). Dr. Wolfe also explains that
IPR2015-00325
Patent 7,742,053 B2

11
pixel command threads can involve ALU operations. Id. ¶¶ 161–162 (noting
“pixel command threads involving ALU operations, which contribute to 20–
30% of the ALU operations”). This means that about 10–20% of vertex
command threads involving texture operations cannot be processed in the
ALU, and 20–30% of pixel command threads involving ALU operations
cannot be processed in the texture engine. Put simply, ALU and texture
engines are type-specific processing engines, and each unit or engine cannot
process all of the threads selected by an arbiter.
Construing the “command processing engines” limitation recited in
claim 5 to exclude type-specific processing engines, as proposed by ATI,
would be inconsistent with the Specification. Id. at 2:59–62, 3:62–4:48,
Fig. 4. Additionally, ATI’s proposed claim construction would be
inconsistent with claims 6 and 7, which recite “wherein the plurality of
command processing engines comprises at least one arithmetic logic unit”
and “at least one texture processing engine,” respectively. Id. at 8:16–21.
More importantly, ATI’s proposed claim construction would import
improperly an extraneous negative limitation into the claims—excluding
type-specific processing engines. Hoganas AB v. Dresser Indus., Inc., 9
F.3d 948, 950 (Fed. Cir. 1993) (“It is improper for a court to add
‘extraneous’ limitations to a claim, that is, limitations added ‘wholly apart
from any need to interpret what the patentee meant by particular words or
phrases in the claim.’”). Such a claim construction would not be reasonable
as it would exclude the very embodiment that provides written description
support for the claims at issue. See Kaneka Corp. v. Xiamen Kingdomway
IPR2015-00325
Patent 7,742,053 B2

12
Group Co., 790 F.3d 1298, 1304 (Fed. Cir. 2015) (“A claim construction
that excludes a preferred embodiment is rarely, if ever, correct.”).
For all of the reasons discussed above, we decline to adopt ATI’s
proposed claim construction that excludes type-specific processing engines
and that requires each command processing engine to be able to process all
of the command threads selected by an arbiter. Rather, consistent with the
plain meaning of the claim language, we construe the “command processing
engine” limitation recited in claim 5 as requiring each command processing
engine to be coupled to an arbiter and operable to receive and process a
command thread selected by the arbiter.
B. Antedating Lindholm, Moreton, and Stuttard
LG asserts that each of the following U.S. patents qualifies as prior art
under 35 U.S.C. § 102(e)3 against the challenged claims of the ’053 patent,
which has an effective filing date of September 29, 2003:
Lindholm filed June 27, 2003 (Ex. 1004, at [22]);
Moreton filed April 21, 2003 (Ex. 1006, at [22]); and
Stuttard filed October 9, 2001 (Ex. 1005, at [22]).
Pet. 10. LG relies upon the U.S. filing dates of these references as the prior
art dates under § 102(e). Id. Indeed, neither Lindholm nor Moreton claims
the benefit of an earlier-filed U.S. application. Ex. 1004, 1; Ex. 1006, 1.

3 Because the ’053 patent was filed before the enactment of the Leahy-Smith
America Invents Act, Pub. L. No. 112-29, 125 Stat. 284 (2011) (“AIA”), the
pre-AIA version of 35 U.S.C. § 102(e) applies in this trial.
IPR2015-00325
Patent 7,742,053 B2

13
Stuttard claims under §§ 120 and 365(c), as a continuation, the benefit of
International Application No. PCT/GB00/01332 (“the ’332 PCT
application”), which was filed on April 7, 2000, and published by the
International Bureau in the English language as International Patent
Publication No. WO 00/62182 (“the ’182 international publication”) on
October 19, 2000. Ex. 1005, at [63], 1:5–9. The international filing date,
April 7, 2000, however, is not a U.S. filing date for prior art purposes under
§ 102(e) because the ’332 PCT application was filed prior to November 29,
2000, the effective date of § 102(e).4 Therefore, the effective date of
Stuttard as prior art is its U.S. filing date, October 9, 2001. See Sun Studs,
Inc. v. ATA Equip. Leasing, Inc., 872 F.2d 978, 983 (Fed. Cir. 1989) (“When
patents are not in interference, the effective date of a reference United States
Patent as prior art is its filing date in the United States, as stated in § 102(e),
not the date of conception or actual reduction to practice of the invention
claimed or the subject matter disclosed in the reference patent.”).
As an initial matter, we note that the ’182 international publication
also is a prior art reference under § 102(b) because it was published on
October 19, 2000, which is more than one year prior to the effective filing
date of the ’053 patent (September 29, 2003). Ex. 1005, at [63], 1:5–9;

4 See Sections 4505, 4508 of the American Inventors Protection Act of 1999,
Pub. L. No. 106-113, 113 Stat. 1501A-552, 565–67 (1999), as amended by
the Intellectual Property and High Technology Technical Amendment Act of
2002, Public Law 107-273, 116 Stat. 1158, 1902–03 (2002).

IPR2015-00325
Patent 7,742,053 B2

14
Ex. 3003; Ex. 1001, at [63]. Stuttard and the ’182 international publication
have the same disclosure given Stuttard was issued as a patent from a
continuation application of the ’332 PCT application that was published as
the ’182 international publication. See Transco Prods. Inc. v. Performance
Contracting, Inc., 38 F.3d 551, 555 (Fed. Cir. 1994) (“‘Continuation and
‘divisional’ applications are alike in that they are both continuing
applications based on the same disclosure as an earlier application.”).
LG, however, did not assert the ’182 international publication in any of its
grounds of unpatentability.
Here, ATI seeks to disqualify Lindholm, Moreton, and Stuttard as
prior art under § 102(e) by establishing a date of invention prior to the U.S.
filing dates of these references. PO Resp. 15–29. Section 102(e)(2) requires
a prior art patent to have a U.S. filing date “before the invention by the
applicant for patent.” See Loral Fairchild Corp. v. Matsushita Elec. Indus.
Co., Ltd., 266 F.3d 1358, 1362 (Fed. Cir. 2001). In particular, ATI alleges
that the named inventors of the ’053 patent conceived the claimed subject
matter of the ’053 patent no later than August 24, 2001, before the filing
dates of Lindholm, Moreton, and Stuttard. Id. at 22–24, 27. ATI also
contends that the named “inventors were reasonably and continuously
diligent to reduce the claimed subject matter to practice.” Id. at 2, 24–26,
28–29. Thus, ATI argues diligence from the date of conception until the
constructive reduction to practice of the claimed invention on the filing date
of the application that issued as the ’053 patent. ATI also argues that the
IPR2015-00325
Patent 7,742,053 B2

15
claimed invention was actually reduced to practice before the filing dates of
Lindholm and Moreton. Id. at 15–22.
ATI provides the following timeline:

Ex. 2155, 3. An inventor “may date his patentable invention back to the
time of its conception, if he connects the conception with its reduction to
practice by reasonable diligence on his part, so that they are substantially
one continuous act.” Mahurkar v. C.R. Bard, Inc., 79 F.3d 1572, 1577 (Fed.
Cir. 1996).
Constructive Reduction to Practice
Constructive reduction to practice occurs when a patent application on
the claimed invention is filed. Weil v. Fritz, 572 F.2d 856, 865 n.16 (CCPA
1978). On its face, the ’053 patent claims under § 120, as a continuation, the
benefit of the filing date of U.S. Application 10/673,761, filed on
September 29, 2003. Ex. 1001, at [63]. LG does not challenge that the
effective filing date of the ’053 patent is September 29, 2003. See generally
IPR2015-00325
Patent 7,742,053 B2

16
Reply. On this record, we, therefore, determine that ATI’s date of
constructive reduction practice is September 29, 2003.
Conception
“Conception must be proved by corroborating evidence which shows
that the inventor disclosed to others his completed thought expressed in such
clear terms as to enable those skilled in the art to make the invention.”
Coleman v. Dines, 754 F.2d 353, 359 (Fed. Cir. 1985).
Here, ATI alleges that the named inventors conceived the claimed
subject matter no later than August 24, 2001, while designing a graphics
processing unit known as the R400 Graphics Processing System (“the
R400”). PO Resp. 1–3, 7–11, 23–24. ATI relies on Version 0.4 of the R400
Sequencer Specification (Ex. 2010) as evidence of conception. PO Resp. 11.
As support, ATI also proffers a Declaration of Mr. Laurent Lefebvre
(Ex. 2006), one of the named inventors of the ’053 patent, and a Declaration
of Dr. Wolfe (Ex. 2106 ¶ 240) to explain how the R400 Sequencer
Specification discloses every element of each challenged claim.
LG does not dispute the sufficiency of the R400 Sequencer
Specification for showing conception. See generally Reply. Based on the
evidence before us, we are satisfied with the sufficiency of the documents to
show conception, and they also serve as effective corroboration to
Mr. Lefebvre’s testimony as to establishing conception. Exs. 2006, 2106,
2010, 2028, 2040, 2041. On this record, we, therefore, determine that ATI
has demonstrated by a preponderance of the evidence that the named
inventors of the ’053 patent conceived the claimed system no later than
IPR2015-00325
Patent 7,742,053 B2

17
August 24, 2001, prior to the U.S. filing dates of Stuttard, Moreton, and
Lindholm.
Actual Reduction to Practice
In order to establish an actual reduction to practice, the inventor must
establish that: (1) the inventor constructed an embodiment or performed a
process that met all of the claim limitations; and (2) the invention would
work for its intended purpose. Cooper v. Goldfarb, 154 F.3d 1321, 1327
(Fed. Cir. 1998). The inventor’s testimony must be corroborated by
independent evidence. Id. at 1330. A rule of reason applies to determine
whether the inventor’s testimony has been corroborated. Price v. Symsek,
988 F.2d 1187, 1194 (Fed. Cir. 1993). “The rule of reason, however, does
not dispense with the requirement for some evidence of independent
corroboration.” Coleman, 754 F.2d at 360. The requirement of
“independent” corroboration requires evidence other than the inventor’s
testimony. In re NTP, Inc. 654 F.3d 1279, 1291 (Fed. Cir. 2011).
Here, ATI asserts that the register-transfer level (“RTL”) code
(Exs. 2072–87) in hardware-description language for the R400 is a
constructed embodiment of the claimed invention. PO Resp. 3, 16–22. As
support, ATI proffers a Declaration of Dr. Wolfe (Ex. 2106) to demonstrate
how the RTL code maps to each claim at issue, and a Declaration of
Mr. Lefebvre (Ex. 2006) to explain why the first-triangle test shows that the
RTL code worked for its intended purpose. PO Resp. 3.
LG counters that ATI fails to provide sufficient evidence to establish
that the named inventors constructed a physical embodiment because
IPR2015-00325
Patent 7,742,053 B2

18
computer simulations that were never reduced to a physical embodiment
cannot serve as an actual reduction to practice. Reply 2–3. LG also alleges
that ATI has not shown that its code actually worked for its intended purpose
because ATI’s evidence shows that the code had a number of problems. Id.
at 5–7. LG further contends that the evidence does not show the RTL code
passed the first-triangle test on July 1, 2002, as the RTL code has a revision
date of August 5, 2002, which is after the alleged test. Id. at 7; Ex. 2066, 6.
In its Sur-reply, ATI responds that a physical embodiment is not
required to establish an actual reduction to practice. Sur-reply 2–4 (citing
Cooper, 154 F.3d at 1328; Yorkey v. Diab, 601 F.3d 1279, 1290–91 (Fed.
Cir. 2010)). ATI also asserts that the RTL code is an embodiment that
discloses every claim element and maps to hardware. Id. at 3–4. According
to ATI, its evidence shows that the first-triangle test was conducted
successfully. Id. at 4. ATI cites to Mr. Lefebvre’s testimony (Ex. 2006
¶ 47), and the Program Review Slides (Ex. 2062, 4; Ex. 2066, 2, 6) for
support.
Requirement of a physical embodiment
It is well settled that “[t]here cannot be a reduction to practice of the
invention . . . without a physical embodiment which includes all limitations
of the claim.” UMC Elecs. Co. v. United States, 816 F.2d 647, 652 (Fed.
Cir. 1987) (emphasis added). “It is equally well established that every
limitation of the [claim] must exist in the embodiment and be shown to have
performed as intended.” Newkirk v. Lulejian, 825 F.2d 1581, 1582 (Fed.
Cir. 1987) (emphasis added).
IPR2015-00325
Patent 7,742,053 B2

19
Here, based on the evidence in this trial, we are not persuaded that
ATI has demonstrated that the claimed system was actually reduced to
practice. In its Patent Owner Response and Sur-reply, ATI does not contend
that it constructed a hardware device in accordance with the challenged
claims. Nor has ATI established that an implementation based on the RTL
code constitutes the subject matter of the challenged claims in this record.
At best, ATI has shown that the RTL code constitutes software files in a
hardware description language that describes the logical design and behavior
of the system. ATI, however, has not demonstrated that the design inferred
by the RTL code was synthesized into a set of interconnected hardware
devices. It is well-established that an equivalent of a physical embodiment
does not satisfy the first requirement of actual reduction to practice. Eaton
v. Evans, 204 F.3d 1094, 1097–98 (Fed. Cir. 2000) (holding that a party
cannot obviate the first requirement of constructing a physical embodiment
through evidence of testing an equivalent).
At the outset, we note that the claims at issue are apparatus claims
directed to a graphics processing system, and not a circuit design, as ATI
alleges (Sur-reply 4). See Ex. 1001, 7:10–8:28. As Dr. Wolfe testifies, both
circuit design and fabrication are necessary components of constructing a
physical integrated circuit, commonly referred to as “a chip.” Ex. 2106
¶¶ 32–33. It is undisputed that ATI submits no evidence to show that a
sample or prototype for the R400 was fabricated before the filing dates of
the asserted references. PO Resp. 7–9; Sur-reply 3–4; Tr. 80:1–3. A design,
by itself, is not a physical hardware element, let alone a graphics processing
IPR2015-00325
Patent 7,742,053 B2

20
system. Therefore, such a design is insufficient to show an actual reduction
to practice of the claimed graphics processing system. See UMC, 816 F.2d
at 652 (“It is not sufficient for a reduction to practice that [the inventor] built
and tested only a part of the later-claimed model UMC-B accelerometer.”);
Fitzgerald v. Arbib, 268 F.2d 763, 765–66 (CCPA 1959) (noting that the
reduction to practice of a three-dimensional design invention required the
production of an article embodying that three-dimensional design and not a
mere drawing); In re McIntosh, 230 F.2d 615, 619 (CCPA 1956) (holding
that drawings and models could establish only conception, but not actual
reduction to practice, because “there is nothing of record to suggest that
appellant ever made or tested a full-sized airplane constructed in accordance
with any of the appealed claims”).
We are not persuaded by ATI’s argument that a physical embodiment
is not required to establish an actual reduction to practice of the recited
system. PO Resp. 19–21; Sur-reply 2. ATI’s reliance on Cooper to support
its argument is misplaced. Sur-reply 2 (citing Cooper, 154 F.3d at 1328). In
fact, Cooper specifically quotes UMC, which states that “[t]here cannot be a
reduction to practice of the invention . . . without a physical embodiment
which includes all limitations of the claim,” and further explains that “the
physical embodiment relied upon as an actual reduction to practice must
include every limitation.” Cooper, 154 F.3d at 1327 (quoting UMC, 816
F.2d at 652; citing Correge v. Murphy, 705 F.2d 1326, 1329 (Fed. Cir.
1983)) (emphases added). Moreover, it was undisputed that a physical
embodiment was constructed in Cooper. The inventor in Cooper conducted
IPR2015-00325
Patent 7,742,053 B2

21
successfully a series of experiments with physical embodiments of the
claimed artificial vascular graft. Cooper, 154 F.3d at 1324 (“The parties
agree that a single successful graft in a dog would constitute an actual
reduction to practice.”).
ATI’s reliance on Yorkey also is unavailing. PO Resp. 17; Sur-reply 2
(citing Yorkey, 601 F.3d at 1290–91). Yorkey involved method claims
directed to a method of measuring saturation of a blood constituent, whereas
the instant trial involves apparatus claims directed to a graphics processing
system comprising an arbiter coupled to a memory device and a plurality of
command processing engines. Yorkey, 601 F.3d at 1282; Ex. 1001, 7:10–
8:28. To show an actual reduction to practice of a claimed apparatus, as
here, the inventor must construct a physical embodiment that includes all the
claimed apparatus elements. UMC, 816 F.2d at 652; Fitzgerald, 268 F.2d at
765–66; Correge, 705 F.2d at 1329 (“The physical embodiment relied upon
as an actual reduction to practice of the invention . . . must include every
essential limitation.” (emphasis added)).
Furthermore, in Yorkey, it was uncontested that the method was
performed successfully in measuring blood oxygen saturation, and the
software program implementing the central equation of the eta methodology
was tested using data collected from patients in hospitals and in-house
clinical studies. Yorkey, 601 F.3d at 1287–90. Such circumstances are not
present here. In fact, ATI does not contend that the computer system
executing the RTL code encompasses the claimed graphics processing
system, but rather the RTL code, by itself, which purportedly “defines the
IPR2015-00325
Patent 7,742,053 B2

22
actual hardware of the chip” and serves as “an accurate representation of a
hardware chip design,” is sufficient as an embodiment to show actual
reduction to practice. PO Resp. 8–9, 16–21; Sur-reply 3–4; Ex. 2106 ¶¶ 31,
32. Simply mapping a software code to hardware features is insufficient to
show that the hardware features actually existed. See Newkirk, 825 F.2d at
1582–83 (noting that proof of actual reduction to practice requires showing
that the claimed apparatus features actually existed).
Calling the RTL software code an embodiment or implementation of
the claimed system does not change the fact that the claimed hardware
elements do not exist. At most, the RTL code represents a logical behavior
design from which a hardware implementation can be inferred. However,
“there can be no actual reduction to practice if the constructed embodiment
. . . lacks an element recited in the [claim] or uses an equivalent of that
element.” Eaton, 204 F.3d at 1097 (emphasis added). Hence, we do not
discern that either Cooper or Yorkey supports ATI’s proposition that no
physical embodiment is required, or that a “representation of a hardware
chip design” constitutes an actual reduction to practice of a physical
integrated circuit.
ATI also argues that a physical embodiment is not required simply
because a commercially acceptable embodiment is not required for
establishing actual reduction to practice. PO Resp. 19–21; Sur-reply 2–3.
ATI, however, conflates the two distinct and separate requirements of actual
reduction to practice: (1) constructing a physical embodiment, and
(2) testing the embodiment. “An invention is actually reduced to practice
IPR2015-00325
Patent 7,742,053 B2

23
when [1] it is put into physical form and [2] shown to be operative in
environment of its practical contemplated use.” See Technical Dev. Corp. v.
United States, 597 F.2d 733, 746–47 (Ct. Cl. 1979). For the testing
requirement, “[l]aboratory tests, rather than tests under actual use or service
conditions, may be sufficient to constitute actual reduction to practice if the
conditions of the test adequately simulate the conditions of practical use.”
Id. at 747. However, permitting laboratory test simulation to satisfy the
testing requirement does not eliminate the physical embodiment
requirement. See Eaton, 204 F.3d at 1097; Wetmore v. Quick, 536 F.2d 937,
942 (CCPA 1976) (finding that, despite successful testing, there was no
actual reduction to practice where the embodiment used an equivalent of an
element).
ATI’s reliance on Scott v. Finney also is misplaced, as the Court in
Scott was addressing the testing requirement. 34 F.3d 1058, 1061 (Fed. Cir.
1994) (noting that actual reduction to practice “does not require that the
invention, when tested, be in a commercially satisfactory stage of
development” (emphasis added)). In Scott, there was no dispute that the
physical embodiment actually existed. Id. at 1059. A videotape showed an
operation where the surgeon inserted Dr. Scott’s prototype device into an
anesthetized patient. Id. at 1060–1063. Nothing in Scott suggests that a
physical embodiment is not required, as alleged by ATI. Id. More
importantly, in Eaton, the Federal Circuit expressly rejected the argument
that satisfying the testing requirement eliminates, or acts as a surrogate for,
the requirement of constructing a physical embodiment, because such an
IPR2015-00325
Patent 7,742,053 B2

24
argument “misapprehends this Court’s precedent and conflates the two
requirements.” Eaton, 204 F.3d at 1098 (holding that a party cannot obviate
the first requirement for a physical embodiment through evidence of testing
an equivalent, regardless of the quality of such evidence). “Put simply, these
are two distinct requirements and a party must satisfy each one to establish
an actual reduction to practice.” Id.
We also are not persuaded by ATI’s argument that the inventive
process for the claimed invention ended after chip design, “so it is logical
that RTL is a valid reduction to practice.” IPR2015-00330, Paper 36,
Sur-reply 2. This argument squarely contradicts our reviewing Court’s
precedent that requires a physical embodiment, and that there is no actual
reduction to practice if the embodiment is merely an equivalent. Eaton, 204
F.3d at 1097–98; Wetmore, 536 F.2d at 942; Martin v. Snyder, 214 F.2d 177,
180 (CCPA 1954) (holding that doctrine of equivalents does not apply when
determining whether a constructed embodiment contained every element of
the invention). Moreover, ATI’s argument conflates actual reduction to
practice with conception, which is established when the inventive process
ends. See Hybritech Inc. v. Monoclonal Antibodies Inc., 802 F.2d 1367,
1376 (Fed. Cir. 1986) (Conception is the “formation in the mind of the
inventor of a definite and permanent idea of the complete and operative
invention as it is thereafter to be applied in practice.”); Townsend v. Smith,
36 F.2d 292, 295 (CCPA 1930) (defining conception as “the complete
performance of the mental part of the inventive act”). As discussed above,
actual reduction to practice is established only when the invention “is put
IPR2015-00325
Patent 7,742,053 B2

25
into physical form and shown to be operative in environment of its practical
contemplated use.” Technical Dev., 597 F.2d at 746–47 (emphasis added).
In light of the foregoing, and consistent with our reviewing Court’s
precedent, we determine that a physical embodiment of the claimed graphics
processing system is required here, and that a circuit design, by itself, is
insufficient to establish actual reduction to practice.
ATI does not contend that the computer system executing the RTL
code encompasses the claimed graphics processing system, but rather the
RTL code, alone, is an embodiment of the claimed system. PO Resp. 7–9,
16–21; Sur-reply 3–4. ATI argues that the entire chip-design industry
recognizes that once a chip has been defined in a hardware-description
language, as here, “an embodiment of the chip has been constructed.” PO
Resp. 18–19. ATI alleges that the RTL code “defines the actual hardware of
the chip,” and, as a result, the RTL code is “an implementation” of the
graphic processing unit (“GPU”). Id. at 8–9.
We are not persuaded by ATI’s argument that the RTL code, alone, is
sufficient to meet the first requirement of actual reduction to practice—
constructing a physical embodiment. PO Resp. 7–9, 16–20; Sur-reply 3–4.
At best, the RTL code is a chip design or a software program that represents
the logical design behavior of the claimed graphics processing system. As
ATI’s own expert, Dr. Wolfe, explains, RTL code is “generally used to
model, define, and instantiate a hardware design.” Ex. 2106 ¶ 23.
According to Dr. Wolfe, “[w]hile the design representation at this stage may
resemble software, its primary purpose is to be an accurate representation of
IPR2015-00325
Patent 7,742,053 B2

26
a hardware chip design.” Id. ¶ 32. Dr. Wolfe also testifies that the “R400
Emulator Code . . . emulates the behavior of the graphics-processing system
recited in claims 1, 2, 5, 6, and 7 using software that executes on a
computer.” Ex. 2106 ¶ 37. Based on Dr. Wolfe’s testimony, we observe
that the RTL code is merely a representation of a chip design or a software
program that emulates the behavior of a physical system that includes
hardware elements, but it is not a physical embodiment of that system.
ATI further asserts that it is an integrated circuit design company, so
that the actual reduction to practice was in the RTL code. PO Resp. 3. ATI
argues that the fabrication phase begins only after the RTL code has been
thoroughly tested, and the designer’s final product is the RTL code, which is
converted into a “GDSII or tape-out file and sent to a circuit-fabrication
facility.” Id. at 9 (emphasis in the original). According to ATI, the
fabrication process “is very expensive, so a graphics-processing company,
such as ATI, will typically tape-out a GPU only if the design is ready for
commercialization—meaning that the GPU design has passed hundreds, if
not thousands, of tests.” Id.
ATI’s evidence, however, shows that it was scheduled to obtain a
sample or prototype of the R400 in May 2002. Ex. 2040, 8. ATI confirms
that there was a plan, at the outset, to make a prototype of the chip.
Tr. 81:10–16. ATI also acknowledges that “there was actually never a tape
out, a sample, or a production for the R400,” but ATI did not submit any
evidence as to why ATI could not have obtained a prototype of the R400,
other than commercially related reasons. Tr. 80:2–11, 82:5–10.
IPR2015-00325
Patent 7,742,053 B2

27
In any event, either allegation that ATI is a chip-design company or
that the fabrication phase of making a chip is a very expensive process, even
if they were true, does not discharge ATI from satisfying the requirement of
constructing a physical embodiment. Eaton, 204 F.3d at 1098; Fitzgerald,
268 F.2d at 765–66. In addition, ATI’s arguments are premised improperly
on the notion that constructing a physical embodiment would require the
inventors to fabricate, personally themselves, a chip with a commercially
acceptable design. It is well-settled that a reduction to practice can be done
by another on behalf of the inventor. Solvay S.A., v. Honeywell Int’l Inc.,
742 F.3d 998, 1000 (Fed. Cir. 2014); In re De Baun, 687 F.2d 459, 463
(CCPA 1982) (“[T]here is no requirement that the inventor be the one to
reduce the invention to practice so long as the reduction to practice was done
on his behalf.”). “Commercially acceptable structure or operation is not
necessary for a reduction to practice.” Goodrich v. Harmsen, 442 F.2d 377,
383 (CCPA 1971). Furthermore, the inventors had the opportunity to
establish constructive reduction to practice by filing a patent application
upon conception, avoiding the cost and delay of constructing a physical
embodiment of their invention. See Weil, 572 F.2d at 865 n.16; In re
Mulder, 716 F.2d 1542, 1545 (Fed. Cir. 1983) (holding that appellants are
entitled to rely on their filing date for a constructive reduction to practice);
see also Grabowsky v. Gallaher, 39 App. D.C. 548, 551–52 (D.C. Cir. 1913)
(“Appellant seeks to excuse this delay by the plea that he was a poor man
and did not have the means to actually reduce the invention to practice,
which, undoubtedly, would have involved considerable expense. Granting
IPR2015-00325
Patent 7,742,053 B2

28
this to be true, the record fails to show that he was unable to pay the cost of
preparing and prosecuting an application, which it was his duty to do.”).
Here, ATI chose to wait more than two years from conception to file a patent
application, while others—Stuttard, Moreton, and Lindholm—filed their
applications earlier. Naber v. Cricchi, 567 F.2d 382, 385 n.5 (CCPA 1977)
(“Public policy favors the early disclosure of invention.”); Gould v.
Schawlow, 363 F.2d 908, 921 (CCPA 1966).
Based on the evidence before us, we determine that ATI has not
demonstrated by a preponderance of the evidence that the claimed invention
was actually reduced to practice. For the reasons discussed below, even if
we were to accept the RTL code as a physical embodiment of the claimed
system, we find that ATI fails to provide sufficient evidence in this record to
show that the RTL code was tested successfully and adequately to prove that
the claimed system would have worked for its intended purpose.
Testing requirement
To show actual reduction to practice, the inventor also must
demonstrate that the claimed invention would work for its intended purpose.
Cooper, 154 F.3d at 1327. Here, ATI relies on Mr. Lefebvre’s testimony
(Ex. 2006 ¶¶ 44–52) to show that the RTL code (Exs. 2072–87), submitted
as evidence of actual reduction to practice, worked for its intended purpose,
and several PowerPoint slides (e.g., Ex. 2066, 2, 6) to corroborate
Mr. Lefebvre’s testimony. PO Resp. 13–15, 21–22; Sur-reply 1, 4. LG
counters that the high-level slides are insufficient corroborating evidence,
IPR2015-00325
Patent 7,742,053 B2

29
and that ATI does not proffer any witness to explain how the slides relate to
the claimed elements. Reply 2. On this record, we agree with LG.
We recognize that “some inventions are so simple and their purpose
and efficacy so obvious that their complete construction is sufficient to
demonstrate workability.” Mahurkar, 79 F.3d at 1572. “Complex
inventions and problems in some cases require laboratory tests that
accurately duplicate actual working conditions in practical use,” however.
Scott, 34 F.3d at 1062. “[T]he testing requirement depends on the particular
facts of each case, with the court guided by a common sense approach in
weighing the sufficiency of the testing.” Id. at 1061. Commercial perfection
or absolute replication of the circumstances of the claimed invention’s
ultimate use is not required. Id. at 1063. “Laboratory tests, rather than tests
under actual use or service conditions, may be sufficient to constitute actual
reduction to practice if the conditions of the test adequately simulate the
conditions of practical use.” Technical Dev., 597 F.2d at 747; Elmore v.
Schmitt, 278 F.2d 510, 513 (CCPA 1960).
Here, ATI does not assert that the claimed graphics processing system
is so simple that it needs no testing to show actual reduction to practice. PO
Resp. 13–15. Rather, ATI asserts that, “[t]o validate the R400 design and
before a tape-out could occur, the design had to pass hundreds, if not
thousands, of tests.” Id.; Ex. 2006 ¶ 47. In the context of the ’053 patent,
we agree with ATI that testing is required to show that the claimed system
would work for its intended purpose, but, on this record, we do not find that
IPR2015-00325
Patent 7,742,053 B2

30
there is sufficient evidence to show that the claimed invention would work
for its intended purpose.
Mr. Lefebvre testifies that the R400 team developed the C++
emulation code and RTL code for the various blocks of the R400, and the
code was tested extensively on individual blocks and the entire graphics core
during the development process. Ex. 2006 ¶¶ 44–46. Yet, Mr. Lefebvre and
ATI direct our attention to only one reported result of a “first triangle” test,
which was allegedly conducted on July 1, 2002. Id. ¶¶ 44–49; PO Resp. 13–
15, 21–22. That report is provided in the form of two high-level PowerPoint
slides, reproduced below with a green marking added (Ex. 2066, 2, 6;
Ex. 2155, 27; Papers 63, 64).

IPR2015-00325
Patent 7,742,053 B2

31

Slide 2 shows two triangles with a heading “R400 First HW
Triangle(s)” and Slide 6 is a high-level program schedule, which states, in
part, “Simulate 1 Triangle / Emulate ready for SW” for the July 1, 2002
entry under the “Actual” column. Id. The slides are silent as to what was
tested, how the triangles were generated, and what software program or
computer system generated the triangles. Significantly, the slides do not
indicate whether the key claimed elements were tested—e.g., what priority
scheme, if any, was tested; how many command processing engines, if any,
were tested; or whether each command process engine was tested to
demonstrate that it can execute a pixel or vertex command thread.
The slides are said to be a part of Mr. Peter Pellerite’s presentation.
Sur-reply 4. Yet, ATI did not submit a declaration from Mr. Pellerite, or
IPR2015-00325
Patent 7,742,053 B2

32
anyone else who could explain meaningfully what these slides purportedly
show. Nor does its expert, Dr. Wolfe, testify that he executed the RTL code
and that the code actually worked. Ex. 1020, 305:11–22. In sum, the slides
provide insufficient detail as to whether the RTL code actually passed the
“first triangle” test. See Newkirk, 825 F.2d at 1583 (“Proof of actual
reduction to practice requires more than theoretical capability.”).
Mr. Lefebvre testifies that “I recall the first triangle test being
successful approximately mid-way through the R400 project. Consistent
with my memory, the PowerPoint slide shows that the first triangle was
reported during a meeting on August 30, 2002.” Ex. 2006 ¶ 49.
Mr. Lefebvre also testifies that “[a]ccording to this presentation, the first
triangle was completed on July 1, 2002.” Id. However, Mr. Lefebvre’s
testimony is vague and conclusory, and is not corroborated with independent
evidence. ATI’s declarant, Mr. Watson, testifies that the files submitted as
Exhibits 2072–2104 and 2108, which contain the RTL code and the
emulator code, have a revision date of August 5, 2002. Ex. 2105 ¶ 132.
According to ATI, the RTL code for R400 was not completed until
October 31, 2002. PO Resp. 14–15; Ex. 2068, 1; Ex. 2069, 6; Ex. 2071, 7.
It is unclear whether the particular RTL code with a revision date of
August 5, 2002 (Exs. 2072–87), submitted as evidence of actual reduction to
practice, is the same code that allegedly passed the “first triangle” test on
July 1, 2002. ATI proffers insufficient explanation or credible evidence to
show that they are the same code. As discussed above, the high-level
PowerPoint slides (Ex. 2066, 2, 6) that both ATI and Mr. Lefebvre rely upon
IPR2015-00325
Patent 7,742,053 B2

33
do not identify which software program or computer system generated the
triangles. ATI fails to provide sufficient independent evidence to
corroborate Mr. Lefebvre’s testimony that the RTL code passed the “first
triangle” test on July 1, 2002, and, therefore, Mr. Lefebvre’s testimony is
entitled to little, if any, weight. See NTP, 654 F.3d at 1291; Mahurkar, 79
F.3d at 1577 (noting that the requirement for corroboration of inventor’s
testimony arose out of a concern that inventors testifying at trial would be
tempted to remember facts favorable to their case by the lure of protecting
their patent or defeating another’s patent).
More importantly, neither ATI nor Mr. Lefebvre proffers any
meaningful test results, test input or output files, simulation run logs, or
input test parameters, which a person with ordinary skill in the art would
have expected to see for testing. Ex. 2153, 112:2–14. At a minimum, ATI’s
evidence must show “the conditions of the test adequately simulate the
conditions of practical use.” Technical Dev., 597 F.2d at 747. As discussed
above, the picture of two triangles and high-level program schedule do not
support that the RTL code was tested successfully to establish that the
claimed graphics processing system would work for its intended purpose.
See McDonnell Douglas Corp. v. United States, 670 F.2d 156, 161 (Ct. Cl.
1982) (Proof of actual reduction to practice must show “that the invention
will perform its intended function beyond a probability of failure.”).
In his declaration, Mr. Lefebvre also mentions other tests. Ex. 2006
¶¶ 44–47 (“We ran many tests on the R400 during its development. . . . This
code was tested extensively . . . . Tests could be run on both the emulation
IPR2015-00325
Patent 7,742,053 B2

34
code and the RTL code, and these tests could be run on individual blocks or
the entire graphics core.”). Again, Mr. Lefebvre’s testimony is vague and is
not corroborated sufficiently with independent evidence. There is no
evidence in this record that those other tests were conducted successfully.
As such, Mr. Lefebvre’s testimony is entitled to little, if any, weight. See
NTP, 654 F.3d at 1291; Mahurkar, 79 F.3d at 1577.
The PowerPoint presentation slides cited by ATI are too general,
showing high-level schedules and general descriptions of the “first triangle”
test. PO Resp. 13–15, 21–22; Ex. 2057, 6, 9; Ex. 2061, 2–5; Ex. 2062, 4;
Ex. 2025, 5; Ex. 2071, 7.
For example, one of the PowerPoint presentation slides (Ex. 2058, 1;
Papers 63, 64) is reproduced below:

This high-level slide does not show any test results, test input or
output files, simulation run logs, or input test parameters. ATI’s evidence
IPR2015-00325
Patent 7,742,053 B2

35
does not provide sufficient detail or explanation to demonstrate that the
testing would simulate actually the characteristics and environment of an
operative graphics processing system in its practical, contemplated setting.
See Technical Dev., 597 F.2d at 747–48; Elmore, 278 F.2d at 513;
McDonnell Douglas, 670 F.2d at 162 (“Tests which fail to simulate the
varying and multiple conditions of the invention’s intended environment will
not serve to prove the operability, stability and reliability of the invention for
practical use.”). ATI’s evidence as a whole is insufficient to establish that
the RTL code was tested successfully or adequately to show that the claimed
system would work for its intended purpose. See Gordon v. Hubbard, 347
F.2d 1001, 1007 (CCPA 1965) (finding the evidence as a whole insufficient
to establish actual reduction to practice when it “simply relates that tests
were performed, and gives no definite indication or suggestion what results
were attained or the results were satisfactory”).
In his Declaration, Mr. Lefebvre also provides an example as to how
the RTL code arbitrates between vertices and pixels. Ex. 2006 ¶¶ 50–51.
However, the code alone does not necessarily show that it would accurately
reproduce the operating conditions in which the claimed graphics processing
system would be encountered in practical use. As discussed above, ATI
acknowledges that, “[t]o validate the R400 design and before a tape-out
could occur, the design had to pass hundreds, if not thousands, of tests.” PO
Resp. 13; Ex. 2006 ¶ 47. Without a sufficient showing of a successful test,
the RTL code, by itself, does not demonstrate adequately that the claimed
system would work for its intended purpose.
IPR2015-00325
Patent 7,742,053 B2

36
In addition, as LG points out, ATI’s evidence indicates that there were
significant problems in the code. Reply 5–7. For example, the entry on
August 26, 2002, in an ATI document log, indicates that the code had a
clamping problem that produced a degenerate triangle. Ex. 2048, 50.
Mr. Lefebvre explains that a “degenerate triangle is a triangle that has all
three vertices along the same axis, so basically transforms the area into a
zero area triangle and makes the triangle disappear.” Ex. 1035, 151:19–
152:5. ATI’s logs reveal this clamping problem was still an issue in
February 2003. Ex. 2048, 31–32 (“Fixing clamping problem in the
emulator.”). As additional examples, the document logs for the sequencer’s
source code indicate the following problems: (1) the entry on July 15, 2002,
stating “[f]ixed the event interface in the SQ [sequencer] . . . which was then
crashing the numerical library” (Ex. 2048, 54); (2) the entry on August 5,
2002, stating “[f]ixed a parameter generation bug in SQ [sequencer] which
was causing the emulator to crash” (id. at 52); (3) the entry on August 26,
2002, stating “[f]ixed 3 bugs in the HW accurate interpolators,” which was
causing two triangle tests to fail (id. at 50); (4) the entry on January 7, 2003,
stating “[f]ound a major bug in the SQ [sequencer]” (id. at 36); (5) the entry
on June 19, 2003, stating “this was causing a failure on a WQL test” and
“[t]his was causing r400sx_wrapper _01.ccp to fail (this is a test that I wrote
to duplicate the WQL test that was failing in order to run it on HW)” (id. at
21); and (6) the entry on November 11, 2003, stating “[t]hese shaders were
all broken because the address register was not refreshed prior to use”
(Ex. 2052, 46). In June 2003, one of ATI’s engineers made a change to the
IPR2015-00325
Patent 7,742,053 B2

37
connection between the SX block and the SQ and explained that he was “not
sure how this was working at all with the live SX” beforehand. Ex. 2049,
57. On October 8, 2003, an engineer “[a]dded needed include files” and
remarked that it was “[s]trange how these compiled before this.” Id. at 35.
Mr. Vargas also testifies that the version of the code provided by ATI for his
review may not have been able to pass a triangle test, noting that the
comments in the RTL code indicate that there were known problems.
Ex. 1013 ¶¶ 43–47 (citing Ex. 2072, 38:7, 42:11; Ex. 2078, 2:1–4; Ex. 2080,
20:17–18; Ex. 2081, 2:1–4). Although Dr. Wolfe testifies that the RTL code
maps to each claimed element, Dr. Wolfe did not testify that he executed the
RTL code and that the code actually worked. Ex. 1020, 305:11–22.
In its Sur-reply, ATI does not provide sufficient explanation or
credible evidence as to why a code that forms a degenerate triangle would be
considered to have passed the “first triangle” test. Rather, ATI responds that
the comments identified by LG are related to the emulator code—the code
that ATI does not rely on for its actual reduction to practice. Sur-reply 4.
Dr. Wolfe, however, testifies that chip “designers would ordinarily use the
emulator code such as this to design the integrated circuits as part of their
design process.” Ex. 2106 ¶¶ 38, 142. ATI confirms that the emulator code
“can be tested in a simulated environment to determine whether the GPU
design functions properly and is, therefore, an important first step in the
GPU design process.” PO Resp. 8. ATI further relies on the development
and testing of the emulator code, as activities that were directed toward
actual reduction to practice of the claimed system, to establish reasonable
IPR2015-00325
Patent 7,742,053 B2

38
diligence. Id. at 24–26. In light of the evidence before us, we are not
persuaded that the RTL code, without testing, establishes that the claimed
system would work for its intended purpose.
We have considered ATI’s evidence in this record. We, however, are
not persuaded that the evidence as a whole supports that the RTL code
submitted by ATI, as evidence of actual reduction to practice (Exs. 2072–
87), was tested successfully under conditions that adequately simulate the
conditions of practical use. McDonnell Douglas, 670 F.2d at 163 (indicating
that no reduction to practice until physical tests showed capability of
actually working). For the foregoing reasons, even if we were to accept
ATI’s proposition that the RTL code is a physical embodiment of the
claimed system, we determine that ATI fails to demonstrate by a
preponderance of the evidence that the RTL code was tested successfully to
show that the claimed system would work for its intended purpose.
Reasonable Diligence
“The reasonable diligence standard balances the interest in rewarding
and encouraging invention with the public’s interest in the earliest possible
disclosure of innovation.” Griffith v. Kanamuru, 816 F.2d 624, 626 (Fed.
Cir. 1987). During the period in which reasonable diligence must be shown,
there must be continuous exercise of reasonable diligence. McIntosh, 230
F.2d at 619; see also Burns v. Curtis, 172 F.2d 588, 591 (CCPA 1949)
(referring to “reasonably continuous activity”). A party alleging diligence
must account for the entire critical period. Griffith, 816 F.2d at 626. Even a
short period of unexplained inactivity is sufficient to defeat a claim of
IPR2015-00325
Patent 7,742,053 B2

39
diligence. Morway v. Bondi, 203 F.2d 742, 749 (CCPA 1953); Ireland v.
Smith, 97 F.2d 95, 99–100 (CCPA 1938).
To satisfy the reasonable diligence requirement, “the work relied on
must ordinarily be directly related to reduction to practice of the invention.”
Naber v. Cricchi, 567 F.2d 382, 385–86 (CCPA 1977) (citing Anderson v.
Scinta, 372 F.2d 523 (CCPA 1967); Gunn v. Bosch, 181 USPQ 758 (BPAI
1973); Moore v. Harris v. Hale, 92 USPQ 187 (BPAI 1951)) (holding that
the work done directed at improving oxide and nitride layer deposition
techniques generally applicable to all MNOS devices, not merely the drain-
source protected device of the claimed invention, did not satisfy the
requirement of reasonable diligence). The work done directed to the generic
invention cannot be relied upon as evidence of diligence for the specific
claimed invention. In re Nelson, 420 F.2d 1079, 1081 (CCPA 1970) (“[W]e
must consider the evidence only as it relates to the specific invention
claimed.”). A party alleging diligence must provide corroboration with
evidence that is specific both as to facts and dates. Gould, 363 F.2d at 920;
Kendall v. Searles, 173 F.2d 986, 993 (CCPA 1949). The rule of reason
does not dispense with the need for corroboration of diligence that is specific
as to dates and facts. Gould, 363 F.2d at 920; Kendall, 173 F.2d at 993;
Coleman, 754 F.2d at 360.
Here, ATI bears the burden of production in antedating the asserted
references. See Dynamic Drinkware, LLC v. Nat’l Graphics, Inc., 800 F.3d
1375, 1378–80 (Fed. Cir. 2015). ATI acknowledges that it must establish
reasonable diligence from just before each reference’s filing date until its
IPR2015-00325
Patent 7,742,053 B2

40
own constructive reduction to practice, September 29, 2003, because we
determine that ATI did not meet its burden of establishing actual reduction
to practice. PO Resp. 27. As discussed above, ATI’s conception date is
August 24, 2001, which predates Stuttard’s filing date (October 9, 2001),
Moreton’s filing date (April 21, 2003), and Lindholm’s filing date (June 27,
2003). Id. at 24–29. Accordingly, ATI must establish reasonable diligence
for the following three critical periods:
Critical Period 1 – from just before Stuttard’s filing date,
October 9, 2001, to September 29, 2003;
Critical Period 2 – from just before Moreton’s filing date,
April 21, 2003, to September 29, 2003; and
Critical Period 3 – from just before Lindholm’s filing date,
June 27, 2003, to September 29, 2003.
To establish reasonable diligence during all three critical periods, ATI
relies upon a Declaration of Mr. Lefebvre (Ex. 2006), and the metadata in
document logs and folder histories (Exs. 2048–52, 2107) to show that over
100 project managers and designers worked every non-holiday business day
to reduce the claimed system to practice. PO Resp. 11–13, 24–26, 28–29.
As support, Mr. Lefebvre provides a calendar (Ex. 2006, Part V, 31–56) to
show that at least one person on the R400 project team worked on the R400
design every non-holiday business day from August 24, 2001 until
September 29, 2003. Ex. 2006 ¶ 42, Part V; Exs. 2048–52, 2107. The
calendar references the following document logs and folder histories: R400
Sequencer Emulator Folder History (Ex. 2048, “Em.”); R400 Sequencer
IPR2015-00325
Patent 7,742,053 B2

41
Parts Folder History (Ex. 2049, “Parts”); R400 Document Library Folder
History (Ex. 2050, “Lib.”); R400 Architecture Folder History (Ex. 2051,
“Arch.”); R400 GFX Testing Folder History (Ex. 2052, “Test”); and R400
Shader Pipe Parts Folder History (Ex. 2107, “SP Parts”). Ex. 2006, 31.
In its Reply, LG counters that ATI’s calendar (Ex. 2006, Part V, 31–
56) is directed to the entire R400 project, and not simply the claimed system
of the ’053 patent. Reply 8–9. LG notes, for example, the GFX Testing
Folder History (Ex. 2052) includes all of the tests on the R400 project, and
ATI failed to parse out which tests were related to the blocks that allegedly
embody the claims and which tests related only to other blocks. Id. at 11.
According to LG, “ATI’s work for an entire graphics chip containing
numerous subparts wholly unrelated to these features cannot establish
diligence.” Reply 9.
In response, ATI disagrees, asserting that the project team could not
test the sequencer in isolation, which was the “heart of the chip” and “there
was no way to apply the correct stimulus on just the sequencer.” Sur-reply
5. ATI also submits that at least one document was checked in every
business day during the critical periods, and the metadata associated with
each check-in date describes the work that occurred. Id.
Upon consideration of the evidence in this entire record, we agree
with LG. ATI’s calendar, metadata, document logs, and folder histories
(nearly 1,300 pages) are not self-explanatory and do not explain
meaningfully as to which tasks are reasonably necessary for reducing the
claimed elements to practice, and which tasks are directed toward
IPR2015-00325
Patent 7,742,053 B2

42
developing and testing other chip designs and optional features. Ex. 2006,
Part V, 33–56; Exs. 2048–52, 2107. Moreover, Mr. Lefebvre’s testimony,
the only testimony that attempts to explain the work ATI relies upon in
support of establishing reasonable diligence, is vague and not sufficiently
corroborated by independent evidence. Ex. 2006 ¶¶ 41–43 (“This metadata
also shows work on the design and development of the R400 generally.”).
“Mere work does not necessarily constitute diligence.” Gunn, 181 USPQ at
761. To satisfy the reasonable diligence requirement, the work relied on
must be directed toward, or reasonably necessary for, the reduction to
practice of the claimed system. See Naber, 567 F.2d at 385.
We are not persuaded by ATI’s contention that the work allegedly
performed for the entire R400 project satisfies the reasonable diligence
requirement. As discussed below, the work done for the entire R400 project
includes developing and testing other chip designs and optional features to
improve graphic processing systems generally, and not merely for the
claimed elements. ATI’s general allegation that the sequencer, shader, and
texture blocks could not be worked and tested in isolation, or the sequencer
was “the heart of the chip,” does not explain sufficiently why the work
performed on other chip designs and optional features was reasonably
necessary for the reduction to practice of the claimed system. PO Resp. 11–
12; Sur-reply 5. That allegation also squarely contradicts ATI’s own
evidence, showing that block level testing can be conducted—e.g.,
Mr. Lefebvre testifies that “[t]ests could be run on both the emulation code
and the RTL code, and these tests could be run on individual blocks or the
IPR2015-00325
Patent 7,742,053 B2

43
entire graphics core.” Ex. 2006 ¶¶ 44–47; see also Ex. 2056, 14 (“Tests
from block level tests”); Ex. 2069, 8 (“Block Level Test Status”); Ex. 2049,
57 (“2003/07/01 by llefebvr . . . Now works on the SQSP testbench.”).
Notably, ATI’s evidence shows that the R400 project included
developing and testing other chip designs: R200 (Ex. 2057, 7–10, 12–14),
RV250 (id. at 11, 15), R300 (id. at 11, 13, 15), and RV450 (id. at 3, 16, 17).
See also Ex. 2039, 5 (“Adding R500 modifications.”); Ex. 2041, 6 (“The
R450, aimed at a volume high end market.”); Ex. 2048, 26 (“Neede[d] to
validate load on the R500”); Ex. 2050, 359 (“Updated Appendix for delta
between R300 and R400”), 391 (“R300 figures showing the wrapping
policies”); Ex. 2059, 6 (“R300 Bring up”); Ex. 2061, 9 (“Entire SC Team
consumed for 2 weeks for R300 Hangs (not yet resolved)”). Yet, ATI
includes all of the activities for the entire R400 project, as evidence of
reasonable diligence, without specifically explaining which activities are
directed toward reducing the claimed elements to practice, and which
activities are directed toward developing and testing other chip designs.
Nor does ATI explain why activities that are directed toward the other chip
designs are reasonably necessary for the reduction to practice of the claimed
elements, which allegedly are embodied only in the R400 design (PO
Resp. 11). In light of the foregoing, we determine that ATI fails to provide
evidence that is specific both as to facts and dates for each of the three
critical periods, during which diligence is required. Gould, 363 F.2d at 920;
Kendall, 173 F.2d at 993.
IPR2015-00325
Patent 7,742,053 B2

44
Additionally, ATI relies on the first design of R400, as described in
Version 0.4 of the R400 Sequencer Specification (Ex. 2010), to establish
conception of the claimed invention as of August 24, 2001. PO Resp. 27;
Tr. 48:20–49:12. According to ATI, this first design describes all of the
claimed elements. PO Resp. 11; Tr. 49:2–4 (This first design “shows that
the inventors conceived of the claimed invention, everything that is required
for the claimed invention.”). Notwithstanding that, Mr. Lefebvre redesigned
the R400, after the conception date, to include an optional feature that is not
recited in the claims at issue—a sequencer that is capable of processing an
unlimited number of clauses. PO Resp. 12–13; Ex. 2006 ¶ 15. Mr. Lefebvre
explains that “Microsoft wanted the sequencer to be able to run shaders with
an unlimited number of clauses/instructions.” Ex. 2006 ¶ 30. According to
ATI, the reason for the redesign was Microsoft asked “for more general
purpose – longer-living model.” PO Resp. 12; Ex. 2060, 5. ATI’s activities
appear more in the nature of commercial development, which is not accepted
as an excuse for delay. See Griffin, 816 F.2d at 627; Fitzgerald, 268 F.2d at
766 (“It is well settled that efforts to exploit an invention commercially do
not constitute diligence in reducing it to practice.”); Naber, 567 F.2d at 385–
86 (rejecting the proposition that the reasonable diligence requirement was
satisfied, notwithstanding delay due to general work on layer deposition
techniques needed to produce a commercially-acceptable device).
ATI’s evidence indicates that during the time period between
August 24, 2001, and April 19, 2002, Mr. Lefebvre was working on the
second chip design to include the optional feature. Ex. 2006 ¶¶ 14, 15;
IPR2015-00325
Patent 7,742,053 B2

45
Exs. 2010, 2028; Ex. 2039, 4 (“Changed the interfaces to reflect the changes
in the SP,” and “Changed the spec to reflect the new R400 architecture.”).
ATI also acknowledges that “the evidence goes to both designs and more
than both designs.” Tr. 51:18–19. Therefore, merely stating that the
metadata shows “work on the design and development of the R400
generally” and that “over 100 ATI project managers and designers worked
on the R400 project every non-holiday business day” does not satisfy ATI’s
burden of production. Ex. 2006 ¶¶ 41–42; PO Resp. 28–29. The work done
that is directed to the second design for a commercially-acceptable device
does not constitute reasonable diligence in reducing the claimed elements to
practice. See Naber, 567 F.2d at 385–86; Nelson, 420 F.2d at 1081;
Fitzgerald, 268 F.2d at 766. This time period overlaps with Critical Period 1
by at least six months—from just prior to Stuttard’s filing date, October 9,
2001, to April 19, 2002. Without a meaningful explanation as to which
activities listed in the metadata are directed to the second chip design and
which activities are directed to reducing the claimed elements to practice,
ATI fails to provide a sufficient account or adequate facts to support a
showing of the continuity of activities required for reasonable diligence
during the entire Critical Period 1. See Rieser v. Williams, 255 F.2d 419,
424 (CCPA 1958) (finding that there was no showing of diligence where no
activity was shown during the first thirteen days of the critical period).
According to the R400 Architecture Proposal (dated November 13,
2000), the R400 was scheduled to “tape-out” by April 2, 2002, for
fabricating of the chip, “samples” in May 2002, and “production” in
IPR2015-00325
Patent 7,742,053 B2

46
November 2002. Ex. 2040, 8. The R400 Top Level Specification (dated
March 11, 2001) also states the “tape-out” for the R400 was scheduled to
occur in July 2002. Ex. 2041, 6. Significantly, ATI’s evidence indicates at
least a delay of nearly eight months due to the redesigning for the optional
feature—comparing Version 0.4 of the R400 Sequencer Specification (dated
August 24, 2001) describing the first design, with Version 2.0 of the R400
Sequencer Specification (dated April 19, 2002) describing the second
design. In fact, Mr. Lefebvre testifies during cross-examination that “the
R400 project itself never got taped out.” Ex. 1035, 70:20–21. Yet, ATI
does not explain sufficiently why implementing the optional feature is
reasonably necessary for reducing the claimed elements to practice. See
Griffith, 816 F.2d at 626 (“The correct inquiry is . . . whether it is reasonable
for [the inventor] to require the public to wait for the innovation, given the
well settled policy in favor of early disclosure.”).
Additionally, the R400 Architecture Proposal (Ex. 2040) and the
R400 Top Level Specification (Ex. 2041) indicate that the R400 includes
other optional features that are not recited in the challenged claims of the
’053 patent—e.g., “nearly transparent dual chip,” features for supporting
“dual monitor,” and a single programmable pipeline that is used for 2D
video, 3D vertex, and 3D pixel operations. Exs. 2040, 2041. These
documents also reveal that the R400 project includes work for improving
other optional features of the prior design, the R300. See Ex. 2040, 7 (“We
want to further improve the anti-aliasing used in the R300 by reducing the
needed memory, and possibly increasing the number of samples per pixel.
IPR2015-00325
Patent 7,742,053 B2

47
The goal is more than fifty percent of the performance and less than three
times the memory of anti-aliased rendering. We should also look into
improved methods.”), 8 (“I would still like to aim for 2x the internal
processing capability of the R300”). As ATI’s documents describe, the
optional features increased costs and delays. See, e.g., Ex. 2040, 6 (“To be
able to address the very high end desktop/enthusiast market we will support
a glueless two chip design instead of a 256 bit bus. . . . There will be costs
added to the base chip to support this. Design time, pins, and area will be
impacted by adding this support.”), 13 (“Since we support dual monitor, this
[total bandwidth requirement] is doubled.”); Ex. 2041, 7.
It is well settled that “[d]elays in reduction to practice caused by an
inventor’s efforts to refine an invention to the most marketable and
profitable form have not been accepted as sufficient excuses for inactivity.”
Griffin, 816 F.2d at 627; Schweyer v. Thomas, 68 F.2d 953 (CCPA 1934)
(explaining that efforts toward commercial exploitation of an invention not
yet reduced to practice do not constitute diligence); see also Naber, 567 F.2d
at 385 (holding that, as “there need not be commercial utility to have a
reduction to practice,” the inventor cannot rely on the work related to layer
deposition techniques, which was required to produce a “useful device,”
rather than a “mere laboratory device”). On this record, ATI does not
explain adequately why developing and testing those optional features were
reasonably necessary for reducing the claimed elements to practice. Nor
does ATI explain why the team of engineers and designers could not have
designed, built, and tested a chip embodying the claimed elements, without
IPR2015-00325
Patent 7,742,053 B2

48
those optional features. See Naber, 567 F.2d at 385 (noting that it is the
inventor’s burden to reconcile the waiting period with the reasonable
diligence requirement).
Mr. Lefebvre testifies that he and his colleagues worked to implement
the design for the R400 by updating the R400 Sequencer Specification, and
by developing, testing, and debugging the emulation code and RTL code,
including other components that supported and interacted with the
sequencer. Ex. 2006 ¶¶ 34–43; Exs. 2007, 2009–2018, 2020–2039.
Mr. Lefebvre also testifies that the “metadata shows work that was necessary
for implementing the R400 design,” and “we could not work on or test the
sequencer block or the shader pipe block in isolation.” Ex. 2006 ¶ 42. Once
again, Mr. Lefebvre’s testimony is too general and conclusory.
As discussed above, ATI’s evidence shows that the R400 project
involves other chip designs and optional features that are not recited in the
claims at issue. Yet, Mr. Lefebvre does not explain meaningfully and
specifically which activities listed in the metadata are directed to the
reduction to practice of the claimed elements and which activities are
directed to developing and testing other chip designs and optional features.
Instead, Mr. Lefebvre relies upon a calendar and metadata (nearly 1,300
pages) for the entire R400 project (Ex. 2006, 31–56; Exs. 2048–52, 2107),
without identifying specific facts and dates as to the particular work that was
reasonably necessary for reducing the claimed features to practice. See
Gould, 363 F.2d at 920; Kendall, 173 F.2d at 993. Such evidence fails to
corroborate Mr. Lefebvre’s testimony for establishing that the R400 team
IPR2015-00325
Patent 7,742,053 B2

49
performed reasonably continuous activities to reduce the claimed system to
practice for each of the three critical periods during which diligence is
required. As a result, Mr. Lefebvre’s testimony (Ex. 2006 ¶¶ 34–43) is
entitled to little, if any, weight. See NTP, 654 F.3d at 1291.
According to ATI, the RTL code for the R400 design was completed
by October 31, 2002. PO Resp. 15; Ex. 2071, 7. ATI submitted thirty-one
versions of the R400 Sequencer Specification (Exs. 2007, 2009–18, 2020–
39), but the last version is dated on May 1, 2003. Ex. 2039, 5; Ex. 2006
¶ 36. ATI also proffered nineteen versions of the R400 program review
slide presentations (Exs. 2053–2071), but the last version is dated on
November 12, 2002, which states that the first samples or prototype was
scheduled for June 25, 2003 (Ex. 2071, 7). As discussed above, ATI
confirms that there was a plan, at the outset, to make a prototype of the
R400. Tr. 81:10–16. ATI also acknowledges that “there was actually never
a tape out, a sample, or a production for the R400,” but ATI did not submit
any evidence as to why ATI could not have obtained a prototype of the
R400, other than commercially related reasons. Tr. 80:2–11, 82:5–10.
However, “efforts to exploit an invention commercially do not constitute
diligence in reducing it to practice.” Fitzgerald, 268 F.2d at 766; Griffin,
816 F.2d at 627; Naber, 567 F.2d at 385–86.
On this record, ATI does not provide sufficient explanation or
credible evidence that is specific both as to facts and dates, regarding what
particular work was done during the time period between June 25, 2003, and
September 29, 2003, that is directed toward, or reasonably necessary for, the
IPR2015-00325
Patent 7,742,053 B2

50
reduction to practice of the claimed elements. See Gould, 363 F.2d at 920;
Kendall, 173 F.2d at 993. This time period overlaps with all three critical
periods by about three months—e.g., from just prior to June 27, 2003,
Lindholm’s filing date, to September 29, 2003. As such, ATI fails to
provide a sufficient account or adequate facts to support a showing of the
continuity of activities required for reasonable diligence during each of the
three critical periods. See Rieser, 255 F.2d at 424; Fitzgerald, 68 F.2d at
766 (affirming a determination of lack of reasonable diligence where there
was less than 1 month of inactivity during critical period).
In addition, ATI’s reliance on Keizer is misplaced. Keizer v. Bradley,
270 F.2d 396, 398 (CCPA 1959). In Keizer, the Court noted that the Board
excused the inventor from the usual requirement that the activity must be
directed toward the claimed invention because the evidence supported that
the work required to develop an operative receiver was necessary for the
reduction to practice of the claimed automatic chroma control. Id. Such
circumstances are not present here. As we discussed above, ATI has not
shown that the other chip designs and optional features were required to
develop and test the claimed elements.
We also are not persuaded by ATI’s argument that the metadata
associated with each check-in date describes the work that occurred.
Sur-reply 5. Rather, we agree with LG that ATI’s metadata, document logs,
and folder histories do not show the actual work being done. Reply 12–13;
Ex. 1035, 124:22–125:12, 126:3–5 (“the tool only collects the check-ins and
not the work that goes in between the check-in”), 127:1–7 (“The perforce
IPR2015-00325
Patent 7,742,053 B2

51
logs we have would not tell you that [work was being done on each of those
days in between].”).
As noted above, ATI’s calendar, metadata, document logs, and folder
histories (nearly 1,300 pages) do not explain meaningfully the nature of the
work that was reasonably necessary for reducing the claimed elements to
practice. Ex. 2006, Part V, 33–56; Exs. 2048–52, 2107. ATI confirms that
“a detailed description of the work that was done is not included in that log.”
Tr. 75:17–20.
Notably, ATI’s metadata merely provide shorthand notations, vague
statements, and generic descriptions of the tasks performed by the designers
and engineers. Exs. 2048–52, 2107. For example, the R400 Shader Pipe
Parts Folder History is reproduced below (Ex. 2107):

Vague and general entries, such as “more changes” and “newly added files,”
are insufficient to indicate what work was done or how it was intended to
IPR2015-00325
Patent 7,742,053 B2

52
further the reduction of the apparatus recited in the challenged claims to
practice. Consequently, such entries are insufficient to show reasonable
diligence.
As another example, many of the entries in the R400 Document
Library Folder History state generically: “deletion” (see, e.g., 2003/09/11
entry), “updates” (see, e.g., 2003/08/29, 2003/08/21, 2003/08/11 entries),
“no change” (see, e.g., 2003/08/18 entry), and “[a]dded info” (see, e.g.,
2003/08/15 entry). Ex. 2050. Given that the descriptions are general and
vague, it is unclear to us what was updated or changed, and how the tasks
are directed toward, or reasonably necessary for, the reduction to practice of
the claimed invention.
ATI and Mr. Lefebvre also do not provide any meaningful
explanation as to how the tasks listed in the metadata specifically relate to
the particular work that was reasonably necessary for the reduction to
practice of the claimed invention. ATI cannot reasonably expect us to
search nearly 1,300 pages for evidence to support ATI’s position, without
some guidance from ATI as to what these document logs and folder histories
show and how they show reasonable diligence in each critical period. See
DeSilva v. DiLeonardi, 181 F.3d 865, 866-67 (Fed. Cir. 1999) (“A brief
must make all arguments accessible to the judges, rather than ask them to
play archeologist with the record.”).
We do not find ATI’s explanation that “since the code kept in the
database is the master copy, checking in updated code would itself constitute
changes to the master copy” meaningful. Sur-reply 5. Even if we reviewed
IPR2015-00325
Patent 7,742,053 B2

53
ATI’s calendar and each of the six document files for each day over a time
period of almost two years, ATI did not provide the updated code for each
check-in date, and, as a result, it would be impossible for us, based on this
record, to determine the changes made to the master copy or the actual work
that occurred for each day. ATI fails to provide a reasonable way for us to
determine whether unexplained lapses have not occurred. See Mulder, 716
F.2d at 1542–46 (affirming a determination of lack of reasonable diligence,
where the evidence of record was lacking for a two-day critical period).
For the foregoing reasons, we determine that ATI fails to demonstrate
by a preponderance of the evidence that there was a continuous exercise of
reasonable diligence to reduce the claimed invention to practice, during each
of the three critical periods. Therefore, Moreton, Lindholm, and Stuttard are
prior art under § 102(e) against the claims of the ’053 patent in this
proceeding.
C. Principles of Law
To establish anticipation, each and every element in a claim, arranged
as recited in the claim, must be found in a single prior art reference. Net
MoneyIN, Inc. v. VeriSign, Inc., 545 F.3d 1359, 1369 (Fed. Cir. 2008). It is
well settled that “the reference need not satisfy an ipsissimis verbis test.” In
re Gleave, 560 F.3d 1331, 1334 (Fed. Cir. 2009); In re Bond, 910 F.2d 831,
832–33 (Fed. Cir. 1990). In an anticipation analysis, “it is proper to take
into account not only specific teachings of the reference but also the
inferences which one skilled in the art would reasonably be expected to draw
IPR2015-00325
Patent 7,742,053 B2

54
therefrom.” In re Preda, 401 F.2d 825, 826 (CCPA 1968); In re Graves, 69
F.3d 1147, 1152 (Fed. Cir. 1995) (“A reference anticipates a claim if it
discloses the claimed invention such that a skilled artisan could take its
teachings in combination with his own knowledge of the particular art and
be in possession of the invention.”). Prior art references must be
“considered together with the knowledge of one of ordinary skill in the
pertinent art.” In re Paulsen, 30 F.3d 1475, 1480 (Fed. Cir. 1994) (quoting
In re Samour, 571 F.2d 559, 562 (CCPA 1978)).
A patent claim is unpatentable under 35 U.S.C. § 103(a) if the
differences between the claimed subject matter and the prior art are such that
the subject matter, as a whole, would have been obvious at the time the
invention was made to a person having ordinary skill in the art to which said
subject matter pertains. KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 406
(2007). The question of obviousness is resolved on the basis of underlying
factual determinations including: (1) the scope and content of the prior art;
(2) any differences between the claimed subject matter and the prior art;
(3) the level of ordinary skill in the art; and (4) objective evidence of
nonobviousness. Graham v. John Deere Co., 383 U.S. 1, 17–18 (1966).
D. The Level of Ordinary Skill in the Art
In determining the level of ordinary skill in the art at the time of the
invention, we note that various factors may be considered, including “type of
problems encountered in the art; prior art solutions to those problems;
rapidity with which innovations are made; sophistication of the technology;
IPR2015-00325
Patent 7,742,053 B2

55
and educational level of active workers in the field.” In re GPAC, Inc.,
57 F.3d 1573, 1579 (Fed. Cir. 1995) (citing Custom Accessories, Inc. v.
Jeffrey-Allan Indus., Inc., 807 F.2d 955, 962 (Fed. Cir. 1986)).
Although the parties’ declarants provide two different definitions of a
person with ordinary skill in the art, the parties do not challenge each other’s
definition; nor do they point out any material differences. Ex. 1003 ¶ 38;
Ex. 2151 ¶ 29; PO Resp.; Reply. Further, it is well-settled that the level of
ordinary skill in the art may be reflected by the prior art of record, as here.
See Okajima v. Bourdeau, 261 F.3d 1350, 1355 (Fed. Cir. 2001);
In re GPAC Inc., 57 F.3d 1573, 1579 (Fed. Cir. 1995); In re Oelrich,
579 F.2d 86, 91 (CCPA 1978). One of ordinary skill is presumed to be
aware of all pertinent prior art. Standard Oil Co. v. American Cyanamid
Co., 774 F.2d 448, 454 (Fed. Cir. 1985). Because there is no material
dispute as to the level of ordinary skill in the art and because the level of
skill is reflected by the prior art of record, we need not explicitly define the
level of ordinary skill.
E. Dr. Bagherzadeh’s Declaration
ATI argues that Dr. Bagherzadeh’s Declaration (Ex. 1003) should be
accorded little or no weight because the Declaration allegedly includes
numerous conclusory statements without corroborating evidence. PO Resp.
59. ATI also contends that Dr. Bagherzadeh lacks meaningful experience in
computer graphics, and did not consider any objective indicia of
nonobviousness. Id. at 59–60. LG counters that Dr. Bagherzadeh’s opinions
IPR2015-00325
Patent 7,742,053 B2

56
on the patentability of the challenged claims are supported adequately with
detailed reasoning, citations to the art, and a view of the technology based on
his experience. Reply 24–25.
We determine that ATI has not articulated a persuasive reason for
giving Dr. Bagherzadeh’s Declaration (Ex. 1003), as a whole, little or no
weight. We have reviewed Dr. Bagherzadeh’s testimony, curriculum vitae,
and cross-examination testimony. Exs. 1003, 1009, 2146. Dr. Bagherzadeh
testifies that he is a Professor in the Department of Electrical Engineering
and Computer Science at the University of California at Irvine. Ex. 1003
¶ 3. Dr. Bagherzadeh also was a member of the technical staff at AT&T
Bell Laboratories, and the Chair of the Electrical/Computer Engineering and
Electrical Engineering/Computer Science Department, as well as an IEEE
Fellow. Id. ¶¶ 5–6. Dr. Bagherzadeh also published at least 94 articles and
five book chapters, many of which relate to computer graphics. Ex. 1009.
There is no material dispute as to the level of ordinary skill in the art.
ATI’s expert, Dr. Wolfe, testifies that one of ordinary skill in the art would
have had at least a bachelor’s degree in electrical or computer engineering or
computer science plus five years of experience in the computer graphics
hardware industry, or a master’s degree in electrical or computer engineering
or computer science plus two years of experience in that industry, or an
equivalent combination of education and experience. Ex. 2151 ¶ 29.
Certainly, Dr. Bagherzadeh’s qualification meets or exceeds that definition.
Given that, we determine that Dr. Bagherzadeh’s qualification and
experience are sufficient to qualify him as an expert in the pertinent field
IPR2015-00325
Patent 7,742,053 B2

57
under Federal Rule of Evidence (“FED. R. EVID.”) 702. See, e.g., Ex. 1003
¶¶ 3–8; Ex. 1009. In addition, there is no requirement of a perfect match
between the expert’s experience and the relevant field. SEB S.A. v.
Montgomery Ward & Co., 594 F.3d 1360, 1373 (Fed. Cir. 2010).
We also are not persuaded by ATI’s argument that Dr. Bagherzadeh
“failed to consider any objective indicia of nonobviousness.” PO Resp. 60.
ATI’s reliance on Rambus Inc. v. Rea, 731 F.3d 1248, 1257 (Fed. Cir. 2013)
is misplaced, because it does not support ATI’s proposition that an expert
declarant must consider evidence of secondary considerations before the
patentee presents such evidence (id.). At the time when LG submitted
Dr. Bagherzadeh’s Declaration in support of its Petition, ATI did not
advance any contention as to secondary considerations of nonobviousness.
ATI cannot reasonably require Dr. Bagherzadeh to consider such evidence
before ATI presents it properly in the proceeding.
For the reasons stated above, we decline to accord Dr. Bagherzadeh’s
Declaration, as a whole, little or no weight, as urged by ATI. Rather, we
exercise our discretion to determine the appropriate weight to be accorded to
the evidence presented, including expert opinion, based on the disclosure of
the underlying facts or data upon which that opinion is based.
F. Anticipation Ground based on Moreton
LG asserts that claims 5–7 are unpatentable under § 102(e) as
anticipated by Moreton. Pet. 48–51. To support its assertion, LG explains
how Moreton describes each claim limitation. Id. LG also relies upon
IPR2015-00325
Patent 7,742,053 B2

58
Dr. Bagherzadeh’s Declaration for support. Ex. 1003. ATI responds that
Moreton does not describe certain limitations, and cites to Dr. Wolfe’s
Declaration for support. PO Resp. 30–34; Ex. 2151. We begin below with a
brief summary of Moreton, and then address the parties’ contentions in turn.
Moreton
Moreton discloses a multithreaded graphics processing system.
Ex. 1006, Abs. Figure 1 of Moreton, reproduced below.

As shown in Figure 1 of Moreton, graphics subsystem 120 includes
graphic processor 125 and local memory 135. Id. at 3:56–66. Graphic
processor 125 contains functional units 140–170, and memory controller
130, which includes thread control buffer 127, address unit 128, and cache
unit 129. Id.
IPR2015-00325
Patent 7,742,053 B2

59
Memory device
Claim 5 recites “at least one memory device comprising a first portion
operative to store a plurality of pixel command threads and a second portion
operative to store a plurality of vertex command threads.” Ex. 1001, 8:5–8.
LG asserts that Moreton describes the claimed memory device. Pet. 42–44,
48–50. According to LG, Moreton’s memory resource includes a first
memory section for storing threads of a first thread type, and a second
memory section for storing threads of a second thread type. Id. (citing
Ex. 1006, Abs., 1:43–58, 6:21–37, 8:60–61, 15:48–67, 16:25–26). ATI does
not dispute LG’s contentions regarding the aforementioned “memory
device” claim limitation. PO Resp. 29–34.
Having considered the Petition and Moreton, we agree with LG.
Indeed, Moreton describes a graphics multithreaded processing system for
processing pixel command threads and vertex command threads. Ex. 1006,
Abs. More importantly, Moreton discloses a memory resource that contains
at least two portions of a memory device, each storing a different thread
type. Id. at 15:48–64.
IPR2015-00325
Patent 7,742,053 B2

60
Figure 5 of Moreton is reproduced below.

As shown in Figure 5 of Moreton, memory resource 200 contains first
memory section 220, and second memory section 520, each having at least
two memory spaces 205, 505 for use by threads executing on graphics
processor 125 (shown in Figure 1, reproduced previously). Id. First
memory section 220 stores a first thread type, and second memory section
520 stores a second thread type. Id. The second thread type is different
from the first thread type. Id.
Based on Moreton’s disclosure, we find that Moreton describes a
“memory device comprising a first portion operative to store a plurality of
pixel command threads and a second portion operative to store a plurality of
vertex command threads,” as recited in claim 5.
IPR2015-00325
Patent 7,742,053 B2

61
Arbiter
Claim 5 recites “an arbiter . . . operable to select a command thread
from either of the plurality of pixel command threads and the plurality of
vertex command threads.” Ex. 1001, 8:9–12 (emphases added). As
discussed above, we construe the claim term “arbiter” as any computer
hardware, software, or combination thereof that receives and provides a
command thread. We also construe “a command thread” to include a stream
of instructions or a process that is part of a larger process or program.
In this regard, LG asserts that Moreton’s memory controller, which
includes a thread control buffer, corresponds to the claimed arbiter. Pet. 48–
49. LG notes that Moreton’s thread control buffer selects and assigns a
thread to the functional units for executing the thread. Id. at 44–46. ATI
disagrees, arguing that Moreton’s thread does not include instructions to
process the pixel or vertex data at the time of thread assignment. PO Resp.
30. ATI also alleges that Moreton’s arbitration is related to memory access
and not to selection of a command thread. Id. at 31.
We are not persuaded by ATI’s arguments and expert testimony (PO
Resp. 30–31; Ex. 2151 ¶¶ 120–22), as they rest primarily on an incorrect
assumption that Moreton’s threads are not command threads, requiring a
command thread to include all of the instructions for processing the pixel or
vertex data at the time of thread assignment. Rather, upon review of
Moreton in its entirety, we agree with LG that Moreton describes the
claimed arbiter.
IPR2015-00325
Patent 7,742,053 B2

62
Similar to the ’053 patent, Moreton discloses a multithreaded graphics
processing system for processing vertex and pixel threads. Ex. 1006, Abs.
Moreton’s system reserves, assesses, and manages memory spaces for use
by threads executing on a graphics processor. Id. at 1:7–10. According to
Moreton, graphics data is processed on a graphics processor through the use
of threads executing on the graphics processor. Id. at 1:14–16. Moreton
also discloses a method for allocation of a memory resource for a plurality of
threads of at least two thread types simultaneously executable in a graphics
processor responsive to a graphics program module, including determining a
first set of threads from the plurality of threads. Id. at 1:48–54.
Significantly, Moreton states that “a thread is a set of processes for
processing a sample according to an instruction set associated with the
sample.” Id. at 6:7–9. A sample refers to “primitive data, surface data, pixel
data, vertex data, fragment data, or the like.” Id. at 5:24–25. Upon
consideration of Moreton as a whole, we determine that Moreton’s threads
are command threads that include instructions to process pixel or vertex
sample data, contrary to ATI’s argument (PO Resp. 30).
ATI’s arguments and expert testimony also conflate Moreton’s
threads with the “instruction sets” associated with the sample. See PO Resp.
30–31; Ex. 2151 ¶¶ 120–22; Ex. 1006, 1:48–58, 6:7–9, 4:34–47. Moreton
discloses that “[a]n instruction set is a specific subset of program
instructions of Graphics Application 111 (executing on Host CPU 114) used
to process an associated sample.” Id. at 12:12–14; see also id. at 5:44–46.
ATI’s arguments also fail to appreciate that there are various instructions in
IPR2015-00325
Patent 7,742,053 B2

63
a graphics processing system for processing pixel and vertex data. Nothing
in the claims requires a command thread to contain all of the instructions
necessary to process the pixel or vertex data at the time of assignment.
We agree with LG and find that Moreton’s memory controller, which
includes a thread control buffer, selects and assigns a thread to a plurality of
functional units for executing the thread. Pet. 44–49; Ex. 1006, 4:4–20,
5:61–66, 7:39–42. Moreton discloses that, when a functional unit receives a
sample, which includes pixel or vertex data, a sample type identifier
associated with the sample, and a pointer to an instruction set associated
with the sample, a thread is assigned to the sample by the thread control
buffer. Id. at 5:61–66. The functional units are programmable units capable
of executing threads in parallel, and performing vertex and pixel operations.
Id. at 5:4–19, 5:65–66.
In light of the foregoing, we determine that LG has demonstrated
sufficiently that Moreton describes an arbiter “operable to select a command
thread from either of the plurality of pixel command threads and the
plurality of vertex command threads,” as recited by claim 5.
A plurality of command processing engines
Claim 5 recites “a plurality of command processing engines, coupled
to the arbiter, each operable to receive and process the command thread.”
Ex. 1006, 8:13–15. In this regard, LG takes the position that Moreton’s
functional units that are connected to a memory controller correspond to the
claimed “command processing engines.” Pet. 49 (citing Ex. 1006, 5:20–23).
ATI responds that Moreton does not disclose a single functional unit that is
IPR2015-00325
Patent 7,742,053 B2

64
capable of processing both pixel and vertex commands. PO Resp. 31–34.
ATI alleges that Moreton’s functional units are type-specific processing
engines, and, therefore, each functional unit cannot process both types of
data. Id. (citing Ex. 1006, 5:10–19, 5:33–40; Ex. 2151 ¶ 127).
ATI’s arguments, however, rest on its proposed claim construction of
the “command processing engine” limitation recited in claim 5, excluding
type-specific processing engines and requiring each command processing
engine to be able to process all of the command threads selected by the
arbiter. PO Resp. 31–34. As discussed above, we decline to adopt ATI’s
proposed claim construction. Rather, we construe the “command processing
engine” limitation as recited in claim 5, in a manner consistent with the plain
meaning of the claim language, as requiring each command processing
engine to be coupled to an arbiter and operable to receive and process a
command thread selected by the arbiter.
As ATI acknowledges, Moreton discloses multiple functional units
that process pixel and vertex data. PO Resp. 33. Indeed, Moreton describes
a multithreaded graphics processing system that includes a plurality of
functional units (command processing engines), each coupled to a memory
controller that includes a thread control buffer (an arbiter) and operable to
receive and process a thread selected by the thread control buffer. Ex. 1006,
5:10–48, 6:15–20. In light of the foregoing, we determine that LG has
established sufficiently that Moreton describes “a plurality of command
processing engines,” as recited in claim 5.
IPR2015-00325
Patent 7,742,053 B2

65
Conclusion on Anticipation
ATI has not raised any additional arguments as to dependent claims 6
and 7 other than those addressed above. Upon review of LG’s analysis and
supporting evidence, we agree with LG’s showing—and adopt it as our
own—that Moreton describes the claimed features recited in claims 6 and 7.
See, e.g., Pet. 37–51; Ex. 1006, 5:10–48, 6:15–20; Ex. 1003 ¶¶ 176–78. For
example, Moreton discloses a functional unit that is “a programmable vertex
processor capable of performing pre-vertex computations (such as lighting
and time-varying spatial offsets), subdivision surface algorithms (as known
in the art), and N-patch algorithms (or ‘normal patch’, as known in the art).”
Ex. 1006, 5:10–15. Moreton also discloses a functional unit that is “a
programmable shader processor capable of performing per-pixel operations,
such as texturing, lighting, bump mapping, or the like.” Id. at 5:15–19. As
Dr. Bagherzadeh testifies, Moreton’s functional units include an arithmetic
logic unit and a texture processing engine. Ex. 1003 ¶¶ 176–78.
For the foregoing reasons, we conclude that LG has shown by a
preponderance of the evidence that claims 5–7 are anticipated by Moreton.
G. Obviousness Ground Based on Moreton and Whittaker
LG asserts that claims 1 and 2 are unpatentable under § 103(a) as
obvious over Moreton and Whittaker. Pet. 37–48. In support of its
assertion, LG provides detailed explanations as to how the combination of
Moreton and Whittaker teaches or suggests each claim limitation and directs
our attention to Dr. Bagherzadeh’s Declaration (Ex. 1003). Id.
IPR2015-00325
Patent 7,742,053 B2

66
ATI counters that Moreton’s thread assignment does not disclose
selecting a command thread as required by claims 1 and 2. PO Resp. 34;
Ex. 2151 ¶ 135. Essentially, ATI relies upon the same arguments presented
in connection with independent claim 5. We have addressed those
arguments in our anticipation analysis above, and conclude that those
arguments are likewise unavailing here.
ATI also argues that Moreton does not disclose an arbiter that
provides a command thread to the command processing engine, as required
by claim 2. PO Resp. 36. ATI further contends that LG fails to provide a
sufficient reason to combine Moreton and Whittaker. Id. at 34–36. As
support, ATI cites to Dr. Wolfe’s Declaration. Ex. 2151.
In our discussion below, we address the parties’ contentions, after
providing a brief summary of Whittaker.5
Whittaker
Whittaker discloses a multithreaded system that includes a media
control core and data processing units for controlling the execution of
instructions. Ex. 1007, 3:36–50. Figure 1 of Whittaker is reproduced below.

5 A summary of Moreton has been provided previously in our anticipation
analysis.
IPR2015-00325
Patent 7,742,053 B2

67

As shown in Figure 1 of Whittaker, each data processing unit 6
includes data processing core 8 that decodes and sequences instructions for
pipeline 10. Id. at 3:46–49. Media control core 2 is a multithreading
processing unit that directs data from inputs to data processing core 8 or to
storage and provides data to outputs. Id. at 3:56–58. Media control core 2
checks “which of the possible operations it could perform have all the
resources available for those tasks to be executed and, of those, which has
the highest priority.” Id. at 3:60–65.
Arbiter
Claim 1 recites “an arbiter . . . operable to select a command
thread . . . based on relative priorities of the plurality of pixel command
threads and the plurality of vertex command threads.” Ex. 1001, 7:16–20.
Claim 2 recites “wherein the arbiter is further operable to provide the
command thread to the command processing engine.” Id. at 7:24–26.
IPR2015-00325
Patent 7,742,053 B2

68
LG asserts that Moreton’s memory controller, which includes a thread
control buffer, corresponds to the claimed arbiter, and that Moreton’s
functional units correspond to the claimed command processing engines.
Pet. 38. As support, LG states that Moreton’s thread control buffer “assigns
threads to data samples and allows instructions and data to be provided to
the plurality of functional units.” Id. at 38–39 (citing Ex. 1006, 5:61–6:2,
12:19–36). LG also notes that Moreton’s memory controller “arbitrates
between hardware components of Graphics Subsystem 120 initiating access
commands to memory resources containing memory spaces used by threads
executing on Graphics Processor 125.” Id. at 39 (citing Ex. 1006, 7:39–42).
Notwithstanding that Moreton does not disclose explicitly selecting a
thread “based on relative priorities” of the threads, LG maintains that the
combination of Moreton and Whittaker would have rendered the claimed
subject matter, as a whole, obvious, as priority selection of threads was
known in the art, as evidenced by Whittaker. Pet. 40–42. LG asserts that
Whittaker discloses a graphics system in which a thread is selected for
processing based on relative priorities of the threads. Id. (citing Ex. 1007,
3:60–67, 8:17–31). LG also submits that it would have been obvious to
modify Moreton’s system to include a priority selection function, in light of
Whittaker, for achieving more efficient data movement and prioritizing
processing of data that requires immediate output, as well as improving
system performance. Id. at 41 (citing Ex. 1007, 3:5–8, 8:29–34).
ATI argues that Moreton does not describe an arbiter that provides a
command thread to a command processing engine, as required by claim 2,
IPR2015-00325
Patent 7,742,053 B2

69
because Moreton’s thread, instead of the memory controller, loads the
instructions to the functional unit. PO Resp. 36. According to ATI, before
the thread control buffer assigns a thread, a functional unit has received
already the pointer to the instruction set. Id.
Once again, ATI’s arguments and expert testimony rest primarily on
an incorrect assumption that Moreton’s thread is not a command thread,
requiring a command thread to include all of the instructions for processing
the pixel or vertex data at the time of thread assignment. PO Resp. 36;
Ex. 2151 ¶¶ 144–45. They also conflate Moreton’s threads with the
“instruction sets” associated with the sample. As discussed above in our
anticipation analysis, we find these arguments unavailing. Rather, upon
review of Moreton as a whole, we find that Moreton’s threads are command
threads that include instructions to process pixel or vertex data. Ex. 1006,
Abs., 1:7–10, 1:14–16, 1:48–54, 5:24–25, 6:7–9. Nothing in the claims
requires a command thread to contain all of the instructions necessary to
process the pixel or vertex data at the time of thread assignment.
IPR2015-00325
Patent 7,742,053 B2

70
Figure 1 of Moreton is reproduced below, with our annotation in
green.

As shown in Figure 1 of Moreton, memory controller 130, which
includes thread control buffer 127 and address unit (“AU”) 128, provides
command threads, data associated with the command threads, and pointers
or addresses to functional units 140–170. Ex. 1006, 5:31–64. More
importantly, Moreton describes that, when functional units receive a sample,
sample type identifier, and pointer to an instruction set associated with the
sample, a thread is assigned to the sample by thread controller buffer 127.
Id. Reading Moreton as a whole, we find that Moreton’s memory controller
provides the threads to the functional units. Id. We agree with LG that
Moreton’s memory controller (arbiter), as modified by Whittaker, is
operable to select and provide a pixel or vertex command thread to a
functional unit (command processing engine), as required by claim 2.
IPR2015-00325
Patent 7,742,053 B2

71
Rationale to combine
ATI asserts that a person of ordinary skill in the art would not have
found it obvious to modify Moreton in view of Whittaker. PO Resp. 34–36.
ATI advances several arguments. Id.
First, ATI alleges that Whittaker’s prioritization scheme does not
involve any pixel or vertex command threads. Id. at 35; Ex. 2151 ¶ 136.
Attacking Whittaker individually, however, does not undermine LG’s
showing of obviousness that is based on the combination of Moreton and
Whittaker. See In re Keller, 642 F.2d 413, 426 (CCPA 1981)
(non-obviousness cannot be established by attacking references individually
where the ground of unpatentability is based upon the teachings of a
combination of references). The test for obviousness is whether the
combination of references, taken as a whole, would have suggested the
patentees’ invention to a person having ordinary skill in the art. In re Merck
& Co., Inc., 800 F.2d 1091, 1097 (Fed. Cir. 1986).
As discussed above, LG relies on Moreton to disclose an arbiter that
selects a command thread from a plurality of pixel and vertex command
threads. Pet. 37–40. LG relies on Whittaker to disclose the priority
selection scheme. Id. at 40–42. Indeed, Whittaker discloses a media control
core that checks “which of the possible operations it could perform have all
the resources available for those tasks to be executed and, of those, which
has the highest priority.” Ex. 1007, 3:60–65 (emphasis added).
IPR2015-00325
Patent 7,742,053 B2

72
Figure 6 of Whittaker is reproduced below.

As shown in Figure 6 of Whittaker, instruction buffers store a
plurality of threads. Id. at 8:18–23. Each thread is sent to resource checker
81, ensuring that the resources required to execute the thread are in place.
Id. at 3:66–67, 8:24–27. If the resources are available, the thread is sent to
priority selection 82 that selects for execution the thread with the highest
priority. Id. at 8:27–31. In view of these disclosures, we are persuaded that
selecting a thread for processing based on relative priorities of the threads
was known in the art at the time of the invention.
We also observe that LG’s articulated rationale for combining
Moreton and Whittaker is supported by Dr. Bagherzadeh’s testimony.
Pet. 41–42 (citing Ex. 1003 ¶¶ 163–166). Specifically, Dr. Bagherzadeh
testifies that it would have been obvious to a person with ordinary skill in
the art to implement the priority selection function in Moreton’s system, in
light of Whittaker, for achieving more efficient data movement, as well as
prioritizing processing of data and improving the performance of the
processing units through load balancing. Ex. 1003 ¶¶ 163–167. On this
IPR2015-00325
Patent 7,742,053 B2

73
record, we credit Dr. Bagherzadeh’s testimony, as it is consistent with the
prior art disclosures. See Ex. 1007, 3:5–8, 8:29–34. Therefore, we are
persuaded by LG’s arguments that one of ordinary skill in the art would have
modified Moreton’s graphics memory controller to select a command thread
between a plurality of pixel command threads and a plurality of vertex
command threads based on relative priorities of the threads, in light of
Whittaker, to improve the efficiency and performance of the graphics
processor. See KSR, 550 U.S. at 417 (“[I]f a technique has been used to
improve one device, and a person of ordinary skill in the art would recognize
that it would improve similar devices in the same way, using the technique is
obvious unless its actual application is beyond his or her skill.”).
Second, ATI argues that “Whittaker was non-analogous art to
Moreton’s technology since Whittaker is not relevant to a graphics
processor.” PO Resp. 35. ATI, however, does not explain why Whittaker’s
thread priority selection disclosure is not reasonably pertinent to the problem
the named inventors of the ’053 patent attempt to solve—namely, selecting a
thread for processing based on relative priorities of the threads. See In re
Clay, 966 F.2d 656, 658–59 (Fed. Cir. 1992) (noting that a reference is
considered to be analogous if it is reasonably pertinent to the particular
problem with which the inventor is concerned, regardless of the field of
endeavor). Rather, ATI merely relies upon an attorney’s argument in
another proceeding. PO Resp. 35 (citing Samsung Electronics Co., Ltd. v.
NVidia Corp., IPR2015-01198, slip op. at 44, 48 (PTAB Sept. 4, 2015)
(Paper 7)). We are not persuaded by ATI’s argument here, as arguments of
IPR2015-00325
Patent 7,742,053 B2

74
counsel cannot take the place of evidence lacking in this record. See Estee
Lauder Inc. v. L’Oreal, S.A., 129 F.3d 588, 595 (Fed. Cir. 1997).
Finally, ATI contends that incorporating Whittaker’s priority selection
scheme into Morton would destroy Moreton’s principle of operation. PO
Resp. 35. ATI’s argument and expert testimony assume that Moreton’s
principle of operation is matching thread type to the same sample type. Id.
at 35–36; Ex. 2151 ¶¶ 137–40. That characterization is overly narrow. As
LG notes, ATI’s argument and expert testimony ignore Moreton’s disclosure
of a plurality of functional units that can process different types of command
threads. Reply 16; Ex. 1006, 5:31–37. Whittaker determines priority by
taking account of the available resources capable of processing the
command thread. Ex. 1007, 3:60–65. For example, when using Whittaker’s
priority scheme, Moreton’s memory controller would determine whether
there is a functional unit available that is capable of processing a vertex
command and, if such a functional unit is not available, the system would
not give that vertex command the highest priority. Id. In light of Whittaker,
it would have been obvious to a person of ordinary skill in the art to
prioritize different types of command threads disclosed by Moreton.
As LG’s expert testifies, implementing Whittaker’s priority scheme in
Moreton’s multi-thread graphics processing system would improve the
efficiency of processing the different types of command threads, and the
performance of the functional units through load balancing. Ex. 1003
¶¶ 163–167. Based on the evidence before us, we are not persuaded that
implementing a priority selection scheme in Morton’s system would destroy
IPR2015-00325
Patent 7,742,053 B2

75
Moreton’s principle of operation for processing different types of command
threads using a plurality of functional units. See In re Umberger, 407 F.2d
425, 430–31 (CCPA 1969) (finding the principle set forth in In re Ratti, 270
F.2d 810, 813 (CCPA 1959), inapplicable where the modified apparatus will
operate “on the same principles as before”).
Having considered the evidence in this record, we determine that LG
has articulated reasoning with rational underpinnings why one with ordinary
skill in the art would have modified Moreton in view of Whittaker, in a
manner to arrive at the subject matter of claims 1 and 2.
Secondary considerations of nonobviousness
Factual inquiries for an obviousness determination include secondary
considerations based on evaluation and crediting of objective evidence of
nonobviousness. Graham, 383 U.S. at 17. Notwithstanding what the
teachings of the prior art would have suggested to one with ordinary skill in
the art at the time of the invention, the totality of the evidence submitted,
including objective evidence of nonobviousness, may lead to a conclusion
that the claimed invention would not have been obvious to one with ordinary
skill in the art. In re Piasecki, 745 F.2d 1468, 1471–1472 (Fed. Cir. 1984).
But, “secondary considerations of nonobviousness . . . simply cannot
overcome a strong prima facie case of obviousness . . . where the inventions
represented no more than ‘predictable use of prior art elements according to
their established functions,’ the secondary considerations are inadequate to
establish nonobviousness as a matter of law.” Wyers v. Master Lock Co.,
616 F.3d 1231, 1246 (Fed. Cir. 2010) (quoting KSR, 550 U.S. at 417).
IPR2015-00325
Patent 7,742,053 B2

76
Here, ATI proffers evidence that purportedly establishes industry
acceptance, arguing that ATI’s unified shader has been adopted widely in
the industry. PO Resp. 60. ATI also contends that companies implementing
a unified shader are able to remain competitive in the marketplace. Id.
(citing Ex. 2142, 14; Ex. 2148, 9, 21; Ex. 2149, 12; Ex. 2150, 5). As support
Dr. Wolfe testifies that Microsoft’s Direct X (DX10) has adopted the unified
shader model. Ex. 2151 ¶¶ 222–223. Dr. Wolfe explains that a “unified
shader” is a single shader that can perform both vertex and pixel
calculations. Id. at ¶ 65.
LG counters that ATI’s evidence rests on the premise that the industry
has embraced the use of a “unified shader” in graphics processors. Reply
25. LG points out that none of the claims in the ’053 patent use the term
“unified shader,” nor does this term appear in the Specification. Id. Upon
review of ATI’s evidence, the claims at issue, and the Specification of the
’053 patent, we agree with LG.
To be of relevance, evidence of nonobviousness must be
commensurate in scope with the claimed subject matter. In re Tiffin, 448
F.2d 791, 792 (CCPA 1971) (evidence of success for cups is not
commensurate in scope with containers). In order to be accorded substantial
weight, there must be a nexus between the merits of the claimed invention
and the evidence of secondary considerations. GPAC, 57 F.3d at 1580.
“Nexus” is a legally and factually sufficient connection between the
objective evidence and the claimed invention, such that the objective
evidence should be considered in determining nonobviousness. Demaco
IPR2015-00325
Patent 7,742,053 B2

77
Corp. v. F. Von Langsdorff Licensing Ltd., 851 F.2d 1387, 1392 (Fed. Cir.
1988). The burden of showing that there is a nexus lies with the patent
owner. Id.
We are not persuaded by ATI’s argument and evidence, as they fail to
establish a nexus between the merits of the claimed invention and the
alleged industry acceptance. As noted by LG, none of the claims at issue
recites a “unified shader.” Ex. 1001, 7:11–9:28. Neither ATI nor Dr. Wolfe
points out which claim limitation requires a “unified shader.” PO Resp. 60;
Ex. 2151 ¶¶ 220–223. In fact, the Specification of the ’053 patent discloses
that the “command processing engine may be any suitable engine as
recognized by one having ordinary skill in the art for processing commands,
such as a texture engine, an arithmetic logic unit, or any other suitable
processing engine.” Ex. 1001, 2:59–62 (emphasis added).
Moreover, although Dr. Wolfe testifies that Microsoft’s DirectX
(DX10) has adopted the unified shader model, Dr. Wolfe does not explain
sufficiently that Microsoft or other companies in the industry would have
accepted the first design of the R400, which allegedly embodies the claimed
system, without other unclaimed features. Ex. 2151 ¶¶ 222–223. In fact,
Mr. Lefebvre, one of the named inventors, testifies that Microsoft rejected
the first design of the R400. Ex. 2006 ¶¶ 15, 30 (“Again, the reason for the
change was to meet the requirement of Microsoft’s API, called DX10. In
particular, Microsoft wanted the sequencer to be able to run shaders with an
unlimited number of clauses/instructions.”); PO Resp. 12; Ex. 2060, 5
(“Microsoft rejecting proposed model – Asking for more general purpose –
IPR2015-00325
Patent 7,742,053 B2

78
longer-living model”). The R400 was redesigned to include features that are
not recited in the claims—e.g., a sequencer that could process an unlimited
number of clauses—in order to meet Microsoft’s acceptance. Ex. 2006
¶¶ 15, 30.
More importantly, the feature—a “unified shader”—that is allegedly
adopted by the industry stems from what was known in the prior art. See,
e.g., Ex. 1005; Ex. 3003. Contrary to ATI’s argument and expert testimony,
the named inventors of the ’053 patent were not the first in the art to disclose
a single graphics processing unit to perform both vertex and pixel
operations. As ATI acknowledges, such a feature was known in the art. PO
Resp. 53; Ex. 2151 ¶ 199 (“Stuttard’s system processes vertex operations
(i.e., geometry) and the so-called pixel operations (i.e., rasterization) . . .
across all processing blocks of the whole system.”). Indeed, both Stuttard
and the ’182 international publication, which was published on October 19,
2000, more than one year prior to the ’053 patent’s effective filing date,
disclose a plurality of processor blocks, each capable of processing both
vertex and pixel operations. Ex. 1005, 8:34–10:47, Fig. 3; Ex. 3003, 17:26–
22:32, Fig. 3.
Where the offered secondary consideration actually results from
something other than what is both claimed and novel in the claim, as here,
there is no nexus to the merits of the claimed invention. Cf. Tokai Corp. v.
Easton Enters., Inc., 632 F.3d 1358, 1369 (Fed. Cir. 2011) (“If commercial
success is due to an element in the prior art, no nexus exists.”); see also
Ormco Corp. v. Align Tech., Inc., 463 F.3d 1299, 1312, 1313 (Fed. Cir.
IPR2015-00325
Patent 7,742,053 B2

79
2006) (“[I]f the feature that creates the commercial success was known in
the prior art, the success is not pertinent.”) (reasoning that success that is due
“‘partially’ to claimed features” and to unclaimed features and/or other
features already in the art lacks the requisite nexus to show unobviousness)
(citations omitted). In the absence of an established nexus with the claimed
invention, secondary consideration factors are given little weight and
generally have no bearing on the legal issue of obviousness. See In re
Vamco Machine & Tool, Inc., 752 F.2d 1564, 1577 (Fed. Cir. 1985).
Accordingly, ATI’s objective evidence is accorded little weight. Based on
the evidence as a whole, we determine that ATI’s evidence does not
establish that the claimed system of the ’053 patent gained industry
acceptance.
We have weighed objective evidence proffered by ATI that allegedly
demonstrates nonobviousness against the evidence of obviousness in the
present record. For the foregoing reasons, we conclude that, on balance, the
strong evidence of obviousness outweighs the weak evidence of
nonobviousness. See Leapfrog Enters., Inc. v. Fisher-Price, Inc., 485 F.3d
1157, 1162 (Fed. Cir. 2007) (holding that the objective considerations of
nonobviousness presented, including substantial evidence of commercial
success, praise, and long-felt need, were inadequate to overcome a strong
showing of primary considerations that rendered the claims at issue invalid).
IPR2015-00325
Patent 7,742,053 B2

80
Conclusion on obviousness based on Moreton and Whittaker
For the foregoing reasons, we conclude that LG has demonstrated by a
preponderance of the evidence that claims 1 and 2 are unpatentable over the
combination of Moreton and Whittaker.
H. Obviousness Grounds Based on Lindholm and the Admitted Prior Art
LG asserts that claims 1, 2, and 5–7 are unpatentable under § 103(a)
as obvious over Lindholm and the Admitted Prior Art. Pet. 22–24. As
support, LG provides detailed explanations as to how the combination of
Lindholm and the Admitted Prior Art meets each claim limitation. Id. LG
also relies upon Dr. Bagherzadeh’s Declaration. Ex. 1003.
ATI counters that LG fails to show Figure 1 and the discussion in the
Background of the Invention Section of the ’053 patent are admitted prior
art. PO Resp. 38–39. ATI also argues that the combination of Lindholm
and the Admitted Prior Art does not disclose certain limitations, and LG fails
to provide a sufficient reason to combine the teachings of Lindholm and the
Admitted Prior Art. PO Resp. 40–50. As support, ATI cites to Dr. Wolfe’s
Declaration. Ex. 2151.
In our discussion below, we address the parties’ contentions after
providing a brief summary of Lindholm and the Admitted Prior Art.
Lindholm
Lindholm describes a multithreaded graphics system that includes a
programmable graphics processing pipeline. Ex. 1004, Abs., 4:4–12.
Figure 2 of Lindholm is reproduced below.
IPR2015-00325
Patent 7,742,053 B2

81

As depicted in Figure 2, programmable graphics pipeline 150 includes
pixel input buffer 215, vertex input buffer 220, execution pipelines 240,
vertex output buffer 260, pixel output buffer 270, and texture unit 225.
Admitted Prior Art
Figure 1 of the ’053 patent is reproduced below.

IPR2015-00325
Patent 7,742,053 B2

82
As shown in Figure 1 of the ’053 patent, prior art sequencing system
100 includes arbiters 101–103 and buffers 104, 106, 108, 110. Id. at 1:32–
64. In a typical embodiment, the buffers are first in and first out (“FIFO”)
buffers, each stores multiple command threads. Id. System 100 is divided
into resource divisions: ALU resource division 120 and texture fetch
resource division 112. Id. Command thread 118 is withdrawn from
reservation stations 104, 108 and provided to an ALU, and the command
threads within texture fetch resource division 122 are withdrawn from
reservation stations 106, 110 to be provided to texture fetch processors. Id.
In short, the prior art sequencing system stores different types of command
threads in different portions of the memory.
LG contends that Figure 1 and the Background of the Invention
Section of the ’053 patent are Admitted Prior Art. Pet. 22–23, 36–37;
Ex. 1001, 1:22–2:6. ATI counters that LG fails to show the relied upon text
in the ’053 patent is actually admitted prior art. PO Resp. 4–5, 38–39. LG
disagrees, arguing that ATI’s assertion is contrary to the controlling case
law. Reply 18 (citing Application of Nomiya, 509 F.2d 566, 570–71 (CCPA
1975); Riverwood Int’l Corp. v. R.A. Jones & Co., 324 F.3d 1346, 1354
(Fed. Cir. 2003)). LG further alleges that “nothing here suggests that the
prior art described in the ’053 patent was the work of its inventors.” Id.
Upon review of the parties’ contentions and supporting evidence, we
agree with LG. In Nomiya, when the appellants’ patent application included
two figures, Figures 1 and 2, that were labeled as “prior art” and described
as such in the specification, the Court held:
IPR2015-00325
Patent 7,742,053 B2

83
We see no reason why appellants’ representation in their
application should not be accepted at face value as admission that
Figs. 1 and 2 may be considered “prior art” for any purpose,
including use as evidence of obviousness under § 103. . . . By
filing an application containing Figs. 1 and 2, labeled prior art,
ipsissimis verbis, and statements explanatory thereof appellants
have conceded what is to be considered as prior art in
determining obviousness of their improvement.
Nomiya, 509 F.2d at 570–71 (citations and footnote omitted).
Here, like in Nomiya, Figure 1 of the ’053 patent is labeled as
“PRIOR ART.” The Background of the Invention Section of the ’053 patent
also describes Figure 1 as showing a prior art sequencing system. Ex. 1001,
1:32 (“FIG. 1 illustrates a prior art sequencing system 100.”), 49 (“In the
prior art embodiments of FIG. 1, the first buffer 104 receives an input
command 124 and outputs a completed command thread 126 to the second
arbiter 102.”), 2:13–14 (“FIG. 1 illustrate[s] the schematic block diagram of
a prior art command thread processing system.”). ATI does not allege that
the prior art embodiment was the named inventors’ own work. PO Resp.
38–39. Nor does ATI contend that Figure 1 or the Specification of the ’053
patent contains any error, or that the named inventors sought correction to
the ’053 patent. In fact, the same drawing with the “PRIOR ART” label and
disclosure also were submitted in application No. 10/673,761, filed on
September 29, 2003, now U.S. Patent No. 7,239,322 B2.
In view of the foregoing, we determine that LG has demonstrated by a
preponderance of the evidence that Figure 1 and the related discussion in the
IPR2015-00325
Patent 7,742,053 B2

84
Background of the Invention Section of the ’053 patent are Admitted Prior
Art against the claims at issue. Ex. 1001, 1:22–2:6, Fig. 1.
Memory device
Claims 1 and 5 recite “at least one memory device comprising a first
portion operative to store a plurality of pixel command threads and a second
portion operative to store a plurality of vertex command threads.” Ex. 1001,
7:1–15, 8:5–8. LG takes the position that Lindholm’s instruction cache unit
and thread control buffer, as modified by the Admitted Prior Art, correspond
to the claimed memory device. Pet. 13, 22. LG relies on Lindholm to
disclose a graphics memory for storing pixel and vertex command threads,
and relies on the Admitted Prior Art to disclose storing different types of
command threads in different portions of the memory device. Id. at 22–24
(citing Ex. 1001, 1:42–64, Fig. 1). LG submits that it would have been
obvious to modify Lindholm’s multithreaded graphics system to store
different types of command threads in different portions of memory, in light
of the Admitted Prior Art. Id. at 24.
ATI counters that the combination of Lindholm and the Admitted
Prior Art does not disclose the claimed memory device because the
Admitted Prior Art does not disclose explicitly storing pixel and vertex
command threads, and an ordinarily skilled artisan would not have modified
the Admitted Prior Art. PO Resp. 40–47. ATI also contends that such an
artisan would not have combined Lindholm and the Admitted Prior Art. Id.
at 44–47. ATI further alleges that combining Lindholm with the Admitted
IPR2015-00325
Patent 7,742,053 B2

85
Prior Art would not achieve faster access to command threads, flexible
processing loads to reduce stalls, or performance improvement. Id.
Having reviewed the evidence in this record, we are not persuaded by
ATI’s arguments. Attacking the Admitted Prior Art individually does not
undermine LG’s showing of obviousness. See Keller, 642 F.2d at 426. The
relevant inquiry here is whether the proposed prior art combination, taken as
a whole, would have suggested the claimed subject matter to one of ordinary
skill in the art. See Merck, 800 F.2d at 1097.
ATI’s arguments with respect to modifying the Admitted Prior Art are
misplaced. LG relies on the Admitted Prior Art to disclose a memory device
that includes a first and second portion for storing different types of
command threads. Pet. 22. Indeed, the Admitted Prior Art discloses that, at
the time of the invention, a computer graphics system typically includes
basic graphical processing elements, such as vertices and pixels. Ex. 1001,
1:24–28. The prior art system disclosed in the Admitted Prior Art has one
set of buffers for storing ALU resource command threads and another set of
buffers for storing texture fetch resource command threads. Id. at 1:42–64.
Based on the evidence before us, we agree with LG that it was known in the
art at the time of the invention to store different types of graphics command
threads in different portions of a memory device.
We further are not persuaded by ATI’s argument that LG has not
articulated a sufficient reason to combine Lindholm with the Admitted Prior
Art. Lindholm already stores pixel and vertex samples separately in
different portions of a memory device. Ex. 1004, 4:35–39, 5:20–27, 7:27–
IPR2015-00325
Patent 7,742,053 B2

86
30 (“Thread Control Buffer 420 includes storage resources for each of at
least two thread types, where the at least two thread types can include pixel,
primitive, and vertex”), Fig. 2 (Pixel Input Buffer 215, Vertex Input Buffer
220, Vertex Output Buffer 260, Pixel Output Buffer 270). Dr. Bagherzadeh
testifies that it would have been obvious to one with ordinary skill in the art
to modify Lindholm’s memory device to store different types of threads in
different portions of memory, in light of the Admitted Prior Art, for “faster
and convenient storage and retrieval functions, and thereby enhancing the
logical layout of the thread processing system,” taking “advantage of
Lindholm’s multithreaded architecture that allows the flexible processing of
data loads to reduce calls.” Ex. 1003 ¶¶ 105–07. Dr. Bagherzadeh’s
testimony is consistent with the prior art of record. Notably, the Admitted
Prior Art indicates that, in a typical graphics processing system, vertices and
pixels were processed through multiple steps providing for the application of
textures and other processing instructions, and that to improve the efficiency
of a graphics system, the control of the flow of the multiple command
threads was preferred. Ex. 1001, 1:22–31. Based on the evidence before us,
we credit the testimony of Dr. Bagherzadeh (Ex. 1003 ¶¶ 105–07) over that
of Dr. Wolfe (Ex. 2151 ¶ 178). See Yorkey, 601 F.3d at 1284 (holding that
the Board has discretion to give more weight to one item of evidence over
another “unless no reasonable trier of fact could have done so”).
ATI and its expert testimony also are conclusory, as they do not
explain specifically why storing pixel and vertex command threads in
different portions of a memory device in a graphics processing system would
IPR2015-00325
Patent 7,742,053 B2

87
not improve the efficiency of thread processing. PO Resp. 55–57; Ex. 2151
¶¶ 208–212. Nor do they provide sufficient or credible evidence that such
an implementation would have been “uniquely challenging” or otherwise
beyond the level of an ordinarily skilled artisan. See Leapfrog, 485 F.3d at
1162. As part of the obviousness analysis, “the knowledge of [a skilled]
artisan is part of the store of public knowledge that must be consulted.”
Randall Mfg. v. Rea, 733 F.3d 1355, 1362 (Fed. Cir. 2013). Here, in fact,
such a system was known in the art at the time of the invention, as evidenced
by Moreton. Ex. 1006, Abs., 6:38–55, 15:53–64, Figs. 1, 5. There is no
dispute that Moreton’s graphics processing system stores pixel and vertex
command threads in different portions of a memory device, as discussed
above. Id.; PO Resp. 29–34.
For the foregoing reasons, we determine that LG has articulated
reasoning with rational underpinnings why an ordinarily skilled artisan
would have implemented, in light of the Admitted Prior Art, a memory
device having a first portion operative to store a plurality of pixel command
threads and a second portion operative to store a plurality of vertex
command threads in Lindholm’s graphics processing system, in a manner to
achieve the subject matter of claims 1 and 5. We, therefore, conclude that
LG has established sufficiently that the combination of Lindholm and the
Admitted Prior Art would render the claimed “memory device” obvious.
Arbiter and a plurality of command processing engines
LG takes the position that: (1) Lindholm’s instruction scheduler and
dispatcher disclose the claimed “arbiter,” as required by claims 1, 2, and 5;
IPR2015-00325
Patent 7,742,053 B2

88
and (2) Lindholm’s computation resources within an execution unit disclose
“a plurality of command processing engines, coupled to the arbiter, each
operable to receive and process the command thread,” as required by
claim 5. Pet. 14–16 (citing Ex. 1004, 8:47–53, 9:5–8, 9:24–25, 9:41–47,
12:58–64).
ATI does not dispute that Lindholm’s instruction scheduler and
dispatcher teaches the claimed arbiter. See PO Resp. 40–50. Rather, ATI
counters that each of Lindholm’s computation resources is a type-specific
processing engine, and cannot process both pixel and vertex command
threads. Id. at 47–50. ATI also alleges that Lindholm does not describe a
plurality of execution units coupled to a single arbiter. Id.
We are not persuaded by ATI’s arguments. Once again, ATI’s
arguments and expert testimony are predicated on ATI’s proposed claim
construction to exclude type-specific processing engines and requiring each
command processing engine to have the capability to process all of the
command threads selected by the arbiter. Id. As discussed above, we
decline to adopt that proposed claim construction, but instead, construe the
“command processing engines” limitation recited in claim 5 as requiring
each command processing engine to be coupled to an arbiter and operable to
receive and process a command thread selected by the arbiter.
Based on our review of Lindholm, we agree with LG that Lindholm’s
computation resources within an execution unit disclose “a plurality of
command processing engines, coupled to the arbiter, each operable to
IPR2015-00325
Patent 7,742,053 B2

89
receive and process the command thread,” as required by claim 5. Figure 4
of Lindholm is reproduced below.

As illustrated in Figure 4 of Lindholm, each execution pipeline 240
includes a plurality of multithreaded processing units 400 and thread control
buffers 420 for receiving samples from pixel input buffer 215 or vertex input
buffer 220. Ex. 1004, 7:19–42. Thread control buffer 420 includes storage
resources for each thread type, e.g., pixel or vertex. Id. Instruction cache
unit 410 is configured to read thread entries based on the priority assigned to
each thread type, and it can be shared between multithreaded processing
units 400. Id. at 8:4–20. Instruction scheduler 430 is configured to:
(1) receive the program instructions, (2) determine availability of source
data, and (3) schedule the program instructions for execution. Id. at 1:49–
57. Instruction scheduler 430, working with instruction dispatcher 440, also
schedules and dispatches pixel or vertex instructions for parallel execution.
IPR2015-00325
Patent 7,742,053 B2

90
Id. at 12:58–64. Instruction scheduler 430 evaluates the availability of
computation resources in execution unit 470. Id. at 8:47–53. Each
execution unit 470 is configured to process samples—e.g., vertices and
pixels—“using programmable computation units to perform operations such
as linear interpolation, derivative calculation, blending, and the like.” Id. at
9:41–47 (emphasis added).
Contrary to ATI’s argument and expert testimony that Lindholm does
not describe multiple execution units coupled to a single arbiter (PO Resp.
49 (citing Ex. 2151 ¶¶ 186–88)), Lindholm explicitly discloses a plurality of
computation resources or programmable computation units (command
processing engines) within execution unit 470 that are coupled to a single
instruction dispatcher working with an instruction scheduler (an arbiter). Id.
Nothing in claim 5 requires the command processing engines to be coupled
directly to the arbiter. Therefore, it is unnecessary to modify Lindholm, as
alleged by ATI (PO Resp. 49–50; Ex. 2151 ¶¶ 189–91), for meeting the
“command processing engines” limitation.
Given the evidence before us, we agree with LG that: (1) Lindholm’s
instruction scheduler and dispatcher teaches the claimed arbiter; and (2)
Lindholm’s computation resources or programmable computation units
within an execution unit are command processing engines, coupled to an
instruction scheduler and dispatcher (an arbiter), each operable to receive
and process a pixel or vertex command thread selected by the arbiter, as
required by claim 5. In view of the foregoing, we determine that LG has
demonstrated sufficiently that Lindholm in combination with the Admitted
IPR2015-00325
Patent 7,742,053 B2

91
Prior Art would render the “arbiter” limitation, as recited in claims 1, 2, and
5, and the “command processing engines” limitation, as recited in claim 5,
obvious.
Secondary considerations of nonobviousness
As discussed above in our obviousness analysis based on Moreton and
Whittaker, we have considered ATI’s evidence of secondary considerations
pertaining to industry acceptance. PO Resp. 60 (citing Ex. 2142, 14;
Ex. 2148, 9, 21; Ex. 2149, 12; Ex. 2150, 5). We find that ATI fails to
establish the required nexus between the merits of the claimed invention and
the alleged evidence of secondary considerations. Accordingly, ATI’s
objective evidence is accorded little weight. We conclude that, on balance,
the strong evidence of obviousness outweighs the weak evidence of
nonobviousness. See Leapfrog, 485 F.3d at 1162.
Conclusion on obviousness based on Lindholm and the Admitted Prior Art
ATI has not raised any additional arguments with regard to claims 6
and 7 other than those addressed above. Upon review of LG’s analysis and
supporting evidence, we agree with LG’s showing—and adopt it as our
own—that the combination of Lindholm and the Admitted Prior Art renders
claims 6 and 7 obvious. See, e.g., Pet. 11–24; Ex. 1001, 1:22–2:6, Fig. 1;
Ex. 1004, 4:57–61, 5:36–40, 9:41–59, Fig. 4. For the foregoing reasons, we
determine that LG has demonstrated by a preponderance of the evidence that
claims 1, 2, and 5–7 are unpatentable over Lindholm and the Admitted Prior
Art.
IPR2015-00325
Patent 7,742,053 B2

92
I. Obviousness Grounds Based on Stuttard and the Admitted Prior Art
LG asserts that claims 1, 2, and 5–7 are unpatentable under § 103(a)
as obvious over Stuttard and the Admitted Prior Art. Pet. 24–37. In support
of its assertion, LG provides detailed explanations as to how the
combination of Stuttard and Admitted Prior Art meets each claim limitation.
Id. LG also cites to Dr. Bagherzadeh’s Declaration for support (Ex. 1003).
ATI counters that LG fails to show Figure 1 and the related discussion
in the ’053 patent are actually admitted prior art. PO Resp. 38–39. We
addressed those arguments in our Lindholm obviousness analysis above and
likewise determine that are unavailing here.
ATI also argues that the combination of Stuttard and the Admitted
Prior Art does not disclose certain limitations, and LG fails to provide a
sufficient reason to combine Stuttard and the Admitted Prior Art. PO Resp.
50–58. In support, ATI directs attention to Dr. Wolfe’s Declaration.
Ex. 2151. We begin our discussion with a brief summary of Stuttard, and
we then address the parties’ contentions in turn.6
Stuttard
Stuttard discloses a multithreaded graphics system that includes a
memory device and processing core. Ex. 1005, 2:48–50, 4:38–5:40.
Figure 3 of Stuttard is reproduced below with our annotation in green added.

6 A summary of the Admitted Prior Art has been provided previously in our
obviousness analysis based on Lindholm and the Admitted Prior Art.
IPR2015-00325
Patent 7,742,053 B2

93

As shown in Figure 3 of Stuttard, the processing core includes thread
manager 102 that provides instructions to a plurality of processing blocks
106 through array controller 104 and channel controller 108. Id. at 4:43–64.
Processing block 106 includes processor units 1061a for executing the
instructions. Id. Thread manager 102 manages instruction streams or
threads to increase efficiency and reduce response time. Id. at 5:19–44.
Figure 4 of Stuttard is reproduced below.

IPR2015-00325
Patent 7,742,053 B2

94
As shown in Figure 4 of Stuttard, thread manager 102 includes thread
scheduler 1025 and thread processors 1026, as well as cache memory
unit 1024 for storing instructions for each thread—having one buffer per
thread. Id. at 5:45–52. The threads are assigned priorities relative to one
another. Id. at 5:56. Thread scheduler 1025 determines which thread should
be active at any one time. Id. at 6:19–21. Thread processors 1026 control
the issuance of instructions from the thread manager to maintain processing
of simultaneously active program threads. Id. at 5:65–6:4.
Memory device
With respect to the “memory device” limitation recited in claims 1
and 5, LG relies on Stuttard’s cache unit for its disclosure of a memory
device for storing pixel and vertex command threads. Pet. 27, 29–31 (citing
Ex. 1005, 5:45–55, 9:50–65, 10:42–47). LG also relies upon the Admitted
Prior Art for its disclosure of storing different types of command threads in
different portions of the memory device. Id. at 36 (citing Ex. 1001, 1:42–
46). LG submits that it would have been obvious to modify Stuttard’s cache
unit to store different types of command threads in different portions of
memory, in light of the Admitted Prior Art, for “faster access to different
types of command threads while also simplifying the storage of threads in
memory of a multithreaded processor.” Id. at 36–37.
ATI counters that it would not have been obvious to modify Stuttard
in view of the Admitted Prior Art. PO Resp. 55–57 (citing Ex. 2151 ¶¶ 210–
11). In particular, ATI argues that LG’s proposed combination is based on
an improper modification of the Admitted Prior Art. Id. at 56. ATI also
IPR2015-00325
Patent 7,742,053 B2

95
alleges that combining Stuttard with the Admitted Prior Art would not
achieve faster access to command threads, reduction of memory, increases in
operation speed, or performance improvement. Id.
ATI’s arguments as to modifying the Admitted Prior Art are
misplaced. As discussed above, the Admitted Prior Art discloses that, at the
time of the invention, a computer graphics system typically includes basic
graphical processing elements, such as vertices and pixels. Ex. 1001, 1:24–
28. The prior art system disclosed in the Admitted Prior Art has one set of
buffers for storing ALU resource command threads and another set of
buffers for storing texture fetch resource command threads. Id. at 1:42–64.
Based on the evidence before us, we agree with LG that it was known in the
art at the time of the invention to store different types of graphics command
threads in different portions of a memory device.
We are not persuaded by ATI’s argument that LG has not articulated a
sufficient reason to combine Stuttard with the Admitted Prior Art. As noted
by LG, Stuttard discloses a graphics processing system for processing pixel
and vertex operations, and the Admitted Prior Art discloses storing different
types of command threads in different memory portions in a graphics
processing system. Pet. 27, 29–31, 36–37; Ex. 1005, 5:45–55, 9:50–65,
10:42–47; Ex. 1001, 1:42–46, Fig. 1. Dr. Bagherzadeh testifies that it would
have been obvious to one with ordinary skill in the art to modify Stuttard’s
cache unit to store different types of threads in different portions of memory,
in light of the Admitted Prior Art, for “faster and convenient storage and
retrieval functions, and thereby enhancing the logical layout of the thread
IPR2015-00325
Patent 7,742,053 B2

96
processing system,” taking “advantage of Stuttard’s multithreaded
architecture that can more efficiently and speedily process graphics data.”
Ex. 1003 ¶¶ 142–44. That testimony is consistent with the prior art of
record. Notably, the Admitted Prior Art indicates that, in a typical graphics
processing system, vertices and pixels were processed through multiple steps
providing for the application of textures and other processing instructions,
and that, to improve the efficiency of a graphics system, the control of the
flow of the multiple command threads was preferred. Ex. 1001, 1:22–31.
Based on the evidence in this record, we credit the testimony of
Dr. Bagherzadeh (Ex. 1003 ¶¶ 142–44) over that of Dr. Wolfe (Ex. 2151
¶ 211). See Yorkey, 601 F.3d at 1284.
ATI and its expert testimony also are conclusory, as they do not
explain specifically why storing pixel and vertex command threads in
different portions of a memory device in a graphics processing system would
not improve the efficiency of thread processing. PO Resp. 55–57; Ex. 2151
¶¶ 208–212. Nor do they provide sufficient or credible evidence that such
an implementation would have been “uniquely challenging” or otherwise
beyond the level of an ordinarily skilled artisan. See Leapfrog, 485 F.3d at
1162. As part of the obviousness analysis, “the knowledge of [a skilled]
artisan is part of the store of public knowledge that must be consulted.”
Randall Mfg., 733 F.3d at 1362. Here, in fact, such a system was known in
the art at the time of the invention, as evidenced by Moreton. Ex. 1006,
Abs., 6:38–55, 15:53–64, Figs. 1, 5. There is no dispute that Moreton’s
graphics processing system stores pixel and vertex command threads in
IPR2015-00325
Patent 7,742,053 B2

97
different portions of a memory device, as discussed above. Id.; PO Resp.
29–34.
For the foregoing reasons, we determine that LG has articulated
reasoning with rational underpinnings why an ordinarily skilled artisan
would have implemented, in light of the Admitted Prior Art, a memory
device having a first portion operative to store a plurality of pixel command
threads and a second portion operative to store a plurality of vertex
command threads in Stuttard’s graphics processing system, in a manner to
achieve at the subject matter of claims 1 and 5. We, therefore, conclude that
LG has established sufficiently that the combination of Stuttard and the
Admitted Prior Art would render the claimed “memory device” obvious.
Arbiter
Claim 1 recites an arbiter “operable to select a command thread from
either of the plurality of pixel command threads and the plurality of vertex
command threads based on relative priorities of the plurality of pixel
command threads and the plurality of vertex command threads.” Ex. 1001,
7:16–20 (emphases added). Claim 2 recites “wherein the arbiter is further
operable to provide the command thread to the command processing
engine.” Id. at 7:24–26. Claim 5 recites an arbiter “operable to select a
command thread from either of the plurality of pixel command threads and
the plurality of vertex command threads.” Id. at 8:9–12 (emphasis added).
LG asserts that Stuttard’s thread scheduler and thread processors in
the thread manager, collectively, disclose an “arbiter,” as required by claims
1, 2, and 5, because the thread scheduler, working in conjunction with the
IPR2015-00325
Patent 7,742,053 B2

98
thread processors, schedules and activates the highest priority thread.
Pet. 28 (citing Ex. 1005, 6:16–21, 11:13–12:25). LG also alleges that
Stuttard’s processing blocks correspond to the claimed “command
processing engines.” Id. at 25 (citing Ex. 1005, 9:61–62).
ATI does not dispute that Stuttard’s processing blocks disclose a
plurality of command processing engines, each operable to receive and
process a pixel or vertex command thread, as required by claims 2 and 5.
PO Resp. 50–58. Rather, ATI argues that Stuttard does not disclose an
“arbiter that selects between pixel and vertex command threads because
Stuttard’s system performs graphics-processing operations in discrete,
sequential phases, i.e., a phased system.” Id. at 51–55 (emphasis added by
ATI). ATI contends that Stuttard does not disclose selection between pixel
and vertex threads because, during the vertex processing phase, the arbiter
would select only from vertex threads, and, during the pixel processing
phase, the arbiter would select only from pixel threads, but not select
between pixel and vertex threads. Id. at 53–54.
ATI’s arguments and expert testimony, however, narrowly focus on a
simplified illustration of how each processing block works when a thread
manager is not utilized and fail to consider Stuttard, as a whole, including
the disclosure of a thread manager and scheduler. PO Resp. 51–55 (citing
Ex. 1005, 9:51–10:47, 18:11–23; Ex. 2151 ¶¶ 192–206). “A reference must
be considered for everything it teaches by way of technology and is not
limited to the particular invention it is describing and attempting to protect.”
EWP Corp. v. Reliance Universal Inc., 755 F.2d 898, 907 (Fed. Cir. 1985)
IPR2015-00325
Patent 7,742,053 B2

99
(emphases in original). In an obviousness analysis, we must consider the
combination of references, as a whole, in light of the general knowledge of
an ordinarily skilled artisan. Merck, 800 F.2d at 1097.
Based on Stuttard’s disclosure as a whole, we are not persuaded by
ATI’s argument and expert testimony that Stuttard does not disclose
selection between pixel and vertex command threads. As discussed above,
Stuttard’s cache unit stores a plurality of pixel and vertex command threads.
Ex. 1005, 5:45–55, 9:50–65, 10:42–47. ATI acknowledges that Stuttard’s
system processes both pixel and vertex operations, and selects a pixel or
vertex command thread for processing. PO Resp. 53–54. Indeed, Stuttard’s
thread manager, which includes a thread scheduler, selects a thread based on
the relative priorities to one another from a plurality of command threads.
Ex. 1005, 5:56–59 (“The threads are assigned priorities relative to one
another.”); 10:49–12:54 (“The thread scheduler, when running, recalculates
which thread should be active whenever one of the following scheduling
triggers occur: A thread with higher priority than the current active thread is
READY . . . . [T]he scheduler activates the highest priority READY
thread.”). Nothing in Stuttard suggests that, when selecting a pixel thread,
the thread manager would disregard all vertex threads, or when selecting a
vertex thread, the thread manager would disregard all pixel threads, as
alleged by ATI. PO Resp. 54. Further, as the Admitted Prior Art describes,
an arbiter was known in the art to select a command thread between different
types of command threads. Ex. 1001, 1:32–64, Fig. 1.
IPR2015-00325
Patent 7,742,053 B2

100
In view of the foregoing, we determine that LG has demonstrated
sufficiently that Stuttard’s thread manager, which includes a thread
scheduler and thread processors, would have rendered the “arbiter”
limitations, as recited in claims 1, 2, and 5, obvious.
Texture processing engine
With respect to claim 7, which requires a texture processing engine,
LG asserts that one of ordinary skill in the art would have recognized that
Stuttard’s graphics data processing of lighting and shading would require a
texture processing engine. Pet. 29 (citing Ex. 1005, 9:61–65). As support,
Dr. Bagherzadeh testifies that because Stuttard discloses that the processing
elements perform lighting and shading functions which are considered
texture processing, the processing elements of Stuttard’s processor units
include a texture processing engine. Ex. 1003 ¶ 131.
ATI takes the position that Stuttard does not suggest a texture
processing engine, as texture operation is an optional, but not a necessary
step of lighting and shading. PO Resp. 57–58; Ex. 2151 ¶¶ 215–17. ATI
also directs our attention to examples disclosed in several references that do
not involve texture operations. Exs. 2004, 2005, 2138, 2139.
An obviousness analysis, however, “need not seek out precise
teachings directed to the specific subject matter of the challenged claim, for
a court can take account of the inferences and creative steps that a person of
ordinary skill in the art would employ.” KSR, 550 U.S. at 418. Prior art
must be read in context, taking account of the general knowledge possessed
IPR2015-00325
Patent 7,742,053 B2

101
by a person with ordinary skill in the art at the time of the claimed invention.
Translogic, 504 F.3d at 1259–1262.
There is no dispute that the use of a texture processing engine in a
graphics processing system was known in the art at the time of the invention.
Ex. 1001, 1:32–48 (“FIG. 1 illustrates a prior art sequencing system . . . and
the command threads within the texture fetch resource division 122 maybe
withdrawn from the reservation stations 106 and 110 to be provided to a
texture fetch processor.”). Notably, in a typical graphics processing system
at the time of the invention, basic graphical elements—vertices and pixels—
were processed through multiple steps providing for the application of
textures and other processing instructions; and arithmetic logic units, and
texture processors were utilized for processing those elements and
instructions. Id. at 1:22–48.
ATI acknowledges that Stuttard’s system processes both vertex and
pixel operations. PO Resp. 53; Ex. 1005, 9:50–10:47. Stuttard also
discloses performing “three dimensional geometry, view, lighting and
shading.” Ex. 1005, 9:63–65. ATI’s expert, Dr. Wolfe, testifies that “vertex
commands can also involve texture operations.” Ex. 2151 ¶¶ 159, 165
(“at least 10–20% of vertex command threads involving texture operations
should be processed by the texture fetch processor”). Dr. Wolfe also agrees
with Dr. Bagherzadeh that the majority of pixel command threads involve
texture operations. Id. ¶¶ 161–162. Therefore, an ordinarily skilled artisan
would have recognized that Stuttard’s graphics processing system includes a
texture processing engine to process pixel and vertex command threads.
IPR2015-00325
Patent 7,742,053 B2

102
Given the evidence in this record, we determine that LG has shown
sufficiently that the combination of Stuttard and the Admitted Prior Art
suggests a texture processing engine, as required by claim 7.
Secondary considerations of nonobviousness
As discussed above in our obviousness analysis based on Moreton and
Whittaker, we have considered ATI’s evidence of secondary considerations
pertaining to industry acceptance. PO Resp. 60 (citing Ex. 2142, 14;
Ex. 2148, 9, 21; Ex. 2149, 12; Ex. 2150, 5). We find that ATI fails to
establish the required nexus between the merits of the claimed invention and
the alleged evidence of secondary considerations. Accordingly, ATI’s
objective evidence is accorded little weight. We conclude that, on balance,
the strong evidence of obviousness outweighs the weak evidence of
nonobviousness. See Leapfrog, 485 F.3d at 1162.
Conclusion on obviousness based on Stuttard and the Admitted Prior Art
ATI has not raised any additional arguments with regard to dependent
claim 6 other than those addressed above. Upon review of LG’s analysis
and supporting evidence, and for the reasons stated above, we agree with
LG’s showing—and adopt it as our own—that the combination of Stuttard
and the Admitted Prior Art renders claim 6 obvious. See, e.g., Pet. 24–37;
Ex. 1001, 1:22–2:6, Fig. 1; Ex. 1005, 20:44–47, Fig. 11. For the foregoing
reasons, we determine that LG has shown by a preponderance of the
evidence that claims 1, 2, and 5–7 are unpatentable over Stuttard and the
Admitted Prior Art.
IPR2015-00325
Patent 7,742,053 B2

103
J. Motion to Exclude Evidence
LG filed a Motion to Exclude Evidence, seeking to exclude portions
of Mr. Watson’s Declaration (Ex. 2105 ¶¶ 4, 5, 11–16, 22–135) and the
documents these paragraphs purport to authenticate (Exs. 2007, 2009–18,
2020–42, 2053–104, 2108). Paper 46 (“Mot.”), 1–3. ATI filed an
Opposition (Paper 47), and LG filed a Reply (Paper 54) in support of its
Motion to Exclude.
ATI relies on the documents (Exs. 2007, 2009–18, 2020–42, 2053–
104, 2108) to antedate Moreton, Lindholm, and Stuttard. PO Resp. 1–29.
We need not reach the merits of LG’s Motion to Exclude Evidence because,
even without excluding portions of Mr. Watson’s Declaration and the
documents, we have decided that issue in LG’s favor. Accordingly, LG’s
Motion to Exclude Evidence is dismissed as moot.
III. CONCLUSION
For the foregoing reasons, we determine that LG has demonstrated by
a preponderance of the evidence that claims 1, 2, and 5–7 of the ’053 patent
are unpatentable based on the following grounds:
Claims Basis References
5–7 § 102(e) Moreton
1 and 2 § 103(a) Moreton and Whittaker
1, 2, and 5–7 § 103(a) Lindholm in view of the Admitted Prior Art
1, 2, and 5–7 § 103(a) Stuttard in view of the Admitted Prior Art

IPR2015-00325
Patent 7,742,053 B2

104
IV. ORDER
For the foregoing reasons, it is
ORDERED that claims 1, 2, and 5–7 of the ’053 patent are held
unpatentable; and
FURTHER ORDERED that, because this is a Final Written Decision,
parties to the proceeding seeking judicial review of the decision must
comply with the notice and service requirements of 37 C.F.R. § 90.2.

IPR2015-00325
Patent 7,742,053 B2

105
PETITIONER:
Robert G. Pluta
Bryan Nese
Cliff Maier
Amanda Streff
Michael W. Maas
MAYER BROWN LLP
rpluta@mayerbrown.com
bnese@mayerbrown.com
cmaier@mayerbrown.com
astreff@mayerbrown.com

PATENT OWNER:

Jason D. Eisenberg
Michael Ray
Johnathan Tuminaro
Michael D. Specht
STERNE, KESSLER, GOLDSTEIN & FOX P.L.L.C.
jasone-PTAB@skgf.com
mray-PTAB@skgf.com
jtuminar-PTAB@skgf.com
mspecht-PTAB@skgf.com