Advanced Micro Devices, Inc.

Patent Trials and Appeals Board

May 28, 2020

15156658 - (D) (P.T.A.B. May. 28, 2020)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
15/156,658 05/17/2016 Michael J. Mantor AMD-080006-US-CNT[2] 4186
25310 7590 05/28/2020
VOLPE AND KOENIG, P.C.
DEPT. AMD
30 SOUTH 17TH STREET -18TH FLOOR
PHILADELPHIA, PA 19103
EXAMINER
HOANG, PETER
ART UNIT PAPER NUMBER
2616
NOTIFICATION DATE DELIVERY MODE
05/28/2020 ELECTRONIC
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
eoffice@volpe-koenig.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

Ex parte MICHAEL J. MANTOR, JEFFREY T. BRADY, and
ANGEL E. SOCARRAS
____________

Appeal 2019-003232
Application 15/156,658
Technology Center 2600
____________

Before CAROLYN D. THOMAS, BRADLEY W. BAUMEISTER, and
ADAM J. PYONIN, Administrative Patent Judges.

THOMAS, Administrative Patent Judge.

DECISION ON APPEAL

Pursuant to 35 U.S.C. § 134(a), Appellant1 appeals from the
Examiner’s decision to reject claims 1–11. We have jurisdiction over the
appeal under 35 U.S.C. § 6(b). An Oral Hearing was held on May 15, 2020.
We AFFIRM.

1 We use the word “Appellant” to refer to “applicant” as defined in
37 C.F.R. § 1.42. Appellant identifies the real party in interest as Advanced
Micro Devices, Inc. Appeal Br. 3.
Appeal 2019-003232
Application 15/156,658

2
The present invention relates generally to transferring data to a
redundant shader pipe array when at least one defective shader pipe is
identified. See Spec., Abstract.
Claim 7 is illustrative:
7. A method, implemented in a sequencer, the method
comprising:
identifying at least one defective shader pipe of a
plurality of shader pipes in a shader pipe array; and
generating a signal directing a redundant shader switch to
transfer data destined for each shader pipe identified as being
defective independently to a redundant shader pipe array.

Appellant appeals the following rejections:
R1. Claims 1–5 and 7–11 are rejected under 35 U.S.C. § 103(a) as
being unpatentable over Mantor (US 2006/0053188 A1, Mar. 9, 2006),
Luick (US 2005/0066148 A1, Mar 24, 2005), and Bastos (US 2008/0094405
A1, Apr. 24, 2008). Final Act. 7–15.; and
R2. Claim 6 is rejected under 35 U.S.C. § 103(a) as being
unpatentable over Mantor, Luick, Bastos, and Smith (US 6,785,840 B1,
Aug. 31, 2004). Final Act. 15–16.
We review the appealed rejections for error based upon the issues
identified by Appellant, and in light of the arguments and evidence produced
thereon. Ex parte Frye, 94 USPQ2d 1072, 1075 (BPAI 2010) (precedential).

ANALYSIS
Rejections under § 103(a)
Claims 1 and 4–10
Issue 1: Did the Examiner err in finding that Bastos teaches and/or
suggests independently transferring data, as set forth in claim 7?
Appeal 2019-003232
Application 15/156,658

3
Here, the Examiner admits that “Mantor and Luick do not explicitly
teach wherein the transfer of data is performed independently” (see Final
Act. 9), but instead relies upon Bastos to disclose this feature. Id. at 9–10.
As such, we shall look for error in the Examiner’s interpretation of Bastos.
As a preliminary matter of claim construction, we broadly but
reasonably construe the claim term “independently” as denoting not relying
on anything else.
Appellant contends that Bastos’ “distribution of segments occurs the
same way (e.g., in a round robin fashion) . . . Accordingly, in Bastos,
segments are simply distributed and not independently transferred to a
redundant shader array. Accordingly, Bastos does not make up for
deficiencies of Mantor and Luick.” Appeal Br. 10–11.
The Examiner starts by relying on Luick to teach “in response to
receiving the generated signal, transfer the data destined to each shader pipe
identified as being defective to the redundant shader pipe array.” Ans. 13
(citing Luick ¶ 15). The Examiner imports Bastos to merely teach “the
transfer [of data] is performed independently.” Ans. 13 (citing Bastos
Abstract; ¶ 29).
Specifically, Bastos discloses:
The shader includes a shader distributor that processes rasterized pixel
data and then selectively distributes the processed rasterized pixel data
to the various shader pipelines, beneficially in a manner that balances
workloads. . . . A shader instruction processor (scheduler) programs
the individual shader pipelines to perform their intended tasks.
Abstract (emphasis added).
Beneficially, the shader pipelines are identical, they can be
programmed independently, individual shader pipelines can be
Appeal 2019-003232
Application 15/156,658

4
disabled, and shader processing operations can be performed even if
one or more shader pipelines is disabled.
Bastos ¶ 15 (emphasis added).
Each shader pipeline 302 can be individually programmed via
program information stored in the frame buffer memory 120 to
perform mathematical and other processing operations on rasterized
pixel data to produce shaded pixel data. . . . Furthermore, such
scalability enables shader pipeline 302 redundancies, which, in turn,
enables the ability to functionally disable defective shader pipelines
302, while still maintaining the overall operation of the shader 300.
Bastos ¶ 29 (emphasis added).
In other words, Bastos discloses the ability to selectively distribute
data and individually/independently program shader pipelines to perform
their intended tasks. As a result, we find unavailing Appellant’s contention
that “in Bastos, segments are simply distributed and not independently
transferred to a redundant shader array” (see Appeal Br. 11), given the
aforementioned teachings that clearly suggest the capability of transferring
data “independently.”

Issue 2: Does the combination of Mantor, Luick, and Bastos change
the principle of operation of Mantor and/or Luick?
Appellant contends that “Mantor and Luick are both designed to
operate using a simple switching function to shift input data in one direction
of a defective ALU or pipeline . . . To add the functionality of Bastos to
Mantor and/or Luick’s switch logic . . . would change the basic principle
under which the Mantor and/or Luick construction was designed to operate.”
Appeal Br. 11–12. We disagree with Appellant.
Appeal 2019-003232
Application 15/156,658

5
Firstly, we note that Appellant’s aforementioned argument is another
form of a teaching away argument. “If references taken in combination
would produce a ‘seemingly inoperative device,’ . . . such references teach
away from the combination and thus cannot serve as predicates for a prima
facie case of obviousness.” McGinley v. Franklin Sports, Inc., 262 F.3d
1339, 1354 (Fed. Cir. 2001) (citation omitted); see also In re ICON Health
& Fitness, Inc., 496 F.3d 1374, 1382 (Fed. Cir. 2007) (“a reference teaches
away from a combination when using it in that combination would produce
an inoperative result,” but the obviousness analysis must account for
“modifications that one skilled in the art would make to a device borrowed
from the prior art”).
Here, Bastos is merely being relied upon by the Examiner, and we
agree, to teach that it was known to “independently” transfer data from one
component to another. For example, as noted supra, Bastos teaches
independently transferring by at least selectively distributing data and
individually and/or independently programming shader pipelines.
Furthermore, Appellant has not shown that Mantor and/or Luick,
criticizes, discredits, or otherwise discourages selectively distributing data
and/or individually/independently programming shader pipelines. Instead,
Luick merely teaches “a redundant pipeline is shared by multiple primary
pipelines, and in the event that any primary pipeline fails, the redundant
pipeline assumes the failing pipeline’s function.” Luick ¶ 15. Luick further
discloses different embodiments for how the redundant pipeline assumes
such functions, e.g., “[a] pipeline failure causes data to be shifted one
position within the array of pipelines, to by-pass the failing pipeline” (Luick
¶ 16); “selection logic controlling the selection between a primary source
Appeal 2019-003232
Application 15/156,658

6
and a secondary source of pipeline data is integrated with other selection
logic . . . is used for by-passing of certain pipeline logic or registers” (Luick
¶ 17); “it would be theoretically possible to divert the data for the inoperable
pipeline directly to the redundant pipeline” (Luick ¶ 53).
In other words, Luick describes alternative embodiments for
transferring data away from a defective shader pipeline to a redundant
pipeline, including selectively distributing data using selection logic, similar
to Bastos’ aforementioned teachings, and even directly transferring data to
the redundant pipe. Mantor similarly illustrates alternative redundant pipe
selection mechanisms. See Mantor Fig. 11.
As such, we point out that “[t]he prior art’s mere disclosure of more
than one alternative does not constitute a teaching away from any of these
alternatives because such disclosure does not criticize, discredit, or
otherwise discourage the solution claimed . . .” In re Fulton, 391 F.3d 1195,
1201 (Fed. Cir. 2004). “A reference may be said to teach away when a
person of ordinary skill, upon reading the reference . . . would be led in a
direction divergent from the path that was taken by the applicant.” In re
Haruna, 249 F.3d 1327 (Fed. Cir. 2001).
We do not find this to be the situation before this Board. At most it
can be argued that Luick shows that “shifting over by one pipeline . . . is
believed to have certain advantages.” See Luick ¶ 53. However, we find
that being less effective is not a teaching away. “A known or obvious
composition does not become patentable simply because it has been
described as somewhat inferior to some other product for the same use.” In
re Gurley, 27 F.3d 551, 553 (Fed. Cir. 1994).
Appeal 2019-003232
Application 15/156,658

7
As such, we find unavailing Appellant’s contention that modifying
Luick’s and/or Mantor’s device with Bastos’ selective transferring
mechanism would change the principle of operation, given that neither Luick
or Mantor discourage the solution claimed.
Accordingly, we sustain the Examiner’s rejection of claim 7.
Appellant’s arguments regarding the Examiner’s rejection of independent
claims 1 and 9 rely on the same arguments as for claim 7, and Appellant
does not argue separate patentability for the dependent claims, except as
noted below. We, therefore, sustain the Examiner’s rejection of claims 1
and 4–10.

Claim 2
Issue 3: Did the Examiner err in finding that Luick teaches
transferring the data destined for the defective shader pipe without
transferring the data destined to all other shader pipes, as set forth in
claim 2?
Appellant contends that in both Mantor and Luick “when data
destined to any defective pipe is shifted one pipe in a particular direction,
data destined to all other pipes in that same direction is also shifted.”
Appeal Br. 13. In particular, Appellant contends that “neither of these
references discloses or suggests . . . transfer the data destined for each shader
pipe identified as being defective without transferring the data destined to all
other shader pipes . . . not identified as being defective.” Id.
The Examiner finds that Luick teaches that “only the defective data of
an identified defective pipeline is sent to the redundant pipeline.” Ans. 17.
Specifically, Luick discloses “[p]referably, a redundant pipeline is shared by
Appeal 2019-003232
Application 15/156,658

8
multiple primary pipelines, and in the event that any primary pipeline fails,
the redundant pipeline assumes the failing pipeline’s function.” Luick ¶ 15.
Luick’s preferred embodiment describes
[a] pipeline failure causes data to be shifted one position within
the array of pipelines, to by-pass the failing pipeline. The shift
is always one position . . . [i.e.,] if the first pipeline in the array
fails and the fifth pipeline redundant, data intended for the first
pipeline is shifted to the second pipeline, data intended for the
second is shifted to the third, and so forth.

Luick ¶ 16.
Interestingly, we further note that if Luick’s fourth pipeline fails
(instead of the first pipeline) and the fifth pipeline is redundant, then under
Bastos’ method the data destined for the defective fourth pipeline would
shift directly to the redundant fifth pipeline, i.e., transfer data destined for
defective fourth pipeline directly to the redundant pipeline, without
transferring the data from non-defective pipelines.
Furthermore, as noted supra, Luick explicitly discloses that “it would
be theoretically possible to divert the data for the inoperable pipeline directly
to the redundant pipeline (in the example above, divert data for pipeline 302
to pipeline 305).” See Luick ¶ 53. In other words, Luick discloses and
suggests alternative embodiments whereby data from the defective shader
pipe is transferred directly to the redundant pipeline without transferring/
shifting the data destined to all other shader pipes that are not defective.
Accordingly, we find unavailing Appellant’s contention that neither
reference discloses the argued limitation given the aforementioned
disclosure in Luick. We, therefore, sustain the Examiner’s rejection of
claim 2.

Appeal 2019-003232
Application 15/156,658

9
Claims 3 and 11
Issue 4: Did the Examiner err in finding that Mantor and/or Luick
teaches a horizontal path or a direct path to the redundant shader pipelines,
as set forth in claims 3 and 11, respectively?
Appellant contends that both Mantor and Luick “are silent with
respect to a ‘horizontal path to the redundant shader pipe array’ or a ‘direct
path to the redundant shader pipe array.’ . . . the data is shifted rather than
transferred via a horizontal or direct path.” Appeal Br. 14; see also Spec.
¶ 28 and Fig. 1.
The Examiner finds that Mantor’s Fig. 1 describes shifting logic that
shifts in a horizontal direction, i.e., shifting to the left or right. See Final
Act. 11; see also Mantor ¶ 20 (“shifting input data for the defective ALU in
the first direction to a neighboring ALU”). Additionally, as noted supra,
Luick teaches failure causes data to be shifted one position within the array
of primary pipelines and a redundant pipeline that are physically adjacent to
one another. See Luick ¶ 16.
Although Mantor and Luick do not explicitly state that the disclosed
data shifts are along a horizontal path, Appellant’s aforementioned
contentions fail to persuasively distinguish the combined teachings of
Mantor’s Fig.1, which illustrates left/right shifting, and Luick’s pipelines
being physically adjacent to one another, from the claimed horizontal path.
For example, Appellant’s Specification merely illustrates, in Appellant’s
Fig. 1, arrows that point in a left/right direction between the shader pipe
array 140 and the redundant shader pipe array 150. This illustration is what
Appellant is relying on to support the claimed horizontal path. See Appeal
Br. 4 (citing Spec. ¶ 28 and Fig.1). Appellant fails to persuasively
Appeal 2019-003232
Application 15/156,658

10
distinguish the disputed claim limitation, read in light of the Specification,
from Mantor’s right/left shifts. See Mantor, Fig. 1.
Accordingly, we sustain the Examiner’s rejection of claims 3 and 11.

CONCLUSION
The Examiner’s rejections of claims 1–11 as being unpatentable under
35 U.S.C. § 103 over at least Mantor, Luick, and Bastos is affirmed.
In summary:
Claims
Rejected
35 U.S.C. § Reference(s)/Basis Affirmed Reversed
1–5, 7–11 103 Mantor, Luick,
Bastos
1–5, 7–11
6 103 Mantor, Luick,
Bastos, Smith
6
Overall
Outcome
1–11

No period for taking any subsequent action in connection with this
appeal may be extended under 37 C.F.R. § 1.136(a)(1)(iv).

AFFIRMED