← All Oral Arguments

INTEL CORPORATION v. PACT XPP SCHWEIZ AG

Oral Argument — 12/07/2022 · Case 22-1037 · 38:10

Appeal Number
22-1037
Argument Date
12/07/2022
Duration
38:10
Segments
1,234
Panel Judges
  • Judge Judge Newman high
  • Judge Judge Prost high
  • Judge Judge Hughes high
Attorneys
  • Appellant Appellant Attorney (Robert Alan Appleby) high
  • Appellee Appellee Attorney (Sanford Ian Weisburst) high
Space Play/Pause   Seek   Prev/Next   [] Speed
0:00 Judge Newman The next target case is number 22, 1037, Intel Corporation against PACT, XBP, Schweitzer AG.
0:08 Mr. Appleby.
0:10 Judge Prost Could I ask a question before the clock starts running?
0:14 Yes, okay.
0:15 Okay, I don't know if the other side has it.
0:17 I'm just curious, are there a lot of other IPRs involving these parties and related patents still going on?
0:24 We've got three.
0:25 We know there's a fourth.
0:26 Are there just a lot more cases in the pipeline?
0:30 Appellant Attorney (Robert Alan Appleby) I do not believe so.
0:31 There were a total at one time of 12, but I believe all but, if somebody can correct me if I'm wrong,
0:41 but all but four, including these three, have been resolved.
0:44 That's good news.
0:45 There is a rehearing pending on a fourth that's not at issue here today that's pending at the PTAB,
0:54 and then there is a fifth patent that's under ex parte re-exam,
0:59 but I believe the other proceedings.
1:01 The other proceedings have been resolved.
1:03 That's my understanding, anyway.
1:05 Sorry.
1:06 Judge Newman Okay.
1:07 All right, thank you.
1:08 Let's start the clock.
1:09 I think we're already in the strap.
1:11 Appellant Attorney (Robert Alan Appleby) Thank you, Your Honor.
1:13 Good morning.
1:13 Robert Appleby on behalf of Appellant Intel.
1:16 May it please the Court.
1:18 In our briefing, Intel raised two errors that occurred below,
1:22 and that led to the Board to find that claim five, the only challenge claim,
1:26 was not unpatentable over the combination of Kabamoto and Bauman.
1:31 In finding that the combination,
1:32 that it did not render claim five unpatentable,
1:34 the Board went astray in two ways.
1:36 First, the Board found that Intel had not shown that one aspect of the interconnect system,
1:44 claim limitation of claim five,
1:45 was met by crediting an argument that PACT now admits it never made below,
1:50 which led the Board to ignore the evidence that Intel introduced on the issue.
1:54 Second, in assessing Intel's argument and evidence that the combination of Kabamoto and Bauman
2:01 represented known techniques,
2:02 to improve similar devices in the same way,
2:06 the Board, in our view,
2:07 did not follow the flexible and expansive analysis that's outlined in KSR,
2:12 but instead improperly cabined the inquiry to whether either reference expressly taught
2:17 that the combination would be an improvement over the prior art.
2:21 Judge Prost Can I ask you, this permeates the other cases,
2:25 and if I start talking about the wrong patents in the wrong case,
2:27 please stop me right away.
2:29 But the question I have is, hypothetically,
2:32 if we agree with your APA stuff,
2:34 what do we do with a case like this?
2:36 Is a remand necessary,
2:38 or are we free, or required, or not allowed to then look at the merits argument ourselves?
2:46 Or do we have to give the Board another chance to say,
2:49 you looked at the wrong argument,
2:51 you were mistaken in terms of what PACT was arguing,
2:53 you came up with some new stuff, do it over.
2:56 Or, given the state of the record at this stage,
3:00 do we just review it,
3:02 and say, yeah, as a matter of law,
3:03 or as a matter of substantial evidence,
3:06 they didn't have enough on this record,
3:08 and review it on this record?
3:11 Appellant Attorney (Robert Alan Appleby) So, in our view,
3:13 with respect to the interconnect subsystem limitation,
3:17 in our view that the evidence was uncontested below,
3:22 that that limitation was met,
3:24 and so we believe that this Court could find that limitation met
3:28 and send, and remand back to the Board,
3:32 for consideration of the obviousness issues.
3:34 With respect to the obviousness issues,
3:36 and the motivation to combine the Board,
3:38 in our view, never did the appropriate analysis,
3:41 under KSR, the flexible and expansive analysis
3:45 that would be required.
3:47 Judge Prost But, so you're not saying that on this record,
3:49 there's sufficient basis to conclude that the Board erred
3:52 in its arguments regarding motivation to combine?
3:56 I mean, you point out all these errors.
3:58 Appellant Attorney (Robert Alan Appleby) No, so maybe I misunderstood your question, Your Honor.
4:00 So,
4:01 I do believe that the evidence on motivation to combine
4:05 only points in one direction,
4:06 that the use of a known alternative of a shared segmented cache,
4:11 a second level cache,
4:13 in place of an individual second level caches,
4:16 that under KSR,
4:18 the evidence only points to one conclusion,
4:22 that there is obviousness in this case,
4:24 and I believe the Court could reach that conclusion.
4:26 Judge Prost Okay, so when you say to send it back for some other obviousness things,
4:29 are there other obviousness things?
4:31 Like secondary considerations or stuff?
4:33 Appellant Attorney (Robert Alan Appleby) There are not, in this case, Jim.
4:36 Judge Hughes So, do you, I think,
4:40 what I'm still a little confused about your answer is,
4:43 is if we find that the Board just erred as a matter of law,
4:47 looking at their reasons for rejecting the combination,
4:51 and their reasons, assuming, I think,
4:54 that, you know, their reasoning of,
4:57 you wouldn't show how you would attach the,
5:00 I guess, the Bowman,
5:02 Global SLC to the Kappa Moto Snoop bus,
5:05 and that it didn't show improvements,
5:07 that we think both of those are just legally incorrect reasonings
5:10 on the known technique,
5:12 and given everything else in the record,
5:14 do we still have to send it back to the Board
5:17 for them to determine whether there's still a motivation to combine,
5:20 or is it that on the record,
5:23 when you have two references in the same field,
5:25 and it's a known technique to improve a device,
5:28 that there's enough to just reverse
5:30 and find a motivation to combine?
5:32 Appellant Attorney (Robert Alan Appleby) So, in our view, given this record,
5:37 this court could find that as a matter of law,
5:43 that the substitution of a segmented second level cache
5:48 as a known alternative for individual second level caches
5:52 was obvious as a matter of law,
5:54 and could just reverse the Board at that point.
5:56 Okay.
5:57 If that answers your question.
5:58 Right.
5:59 Judge Hughes I have a similar question on the interconnected issue.
6:02 Because I see there that the Board is,
6:06 even though there was no real construction,
6:08 they seem to have implicitly required
6:10 that somehow the interconnect system,
6:14 which you, and tell me if I've got the Board's theory
6:19 of your theory of the case incorrect,
6:21 because I have a feeling that the Board may have
6:22 done something that you don't agree with.
6:25 But it seems like the Board thinks that
6:27 you pointed to the Snoop bus from Kabamoto.
6:31 Snoop bus is a great example.
6:32 It's a great term, by the way.
6:34 I hate these computer things I hate,
6:36 but I love Snoop bus.
6:37 That you pointed to that as the interconnect system,
6:42 and that that doesn't show where the segmented caches
6:46 are interconnected, because that's in Bowman
6:49 and not in the Kabamoto Snoop bus.
6:52 And so, if we, and so it sounds to me
6:56 like the Board is saying the interconnect system
7:01 itself has to connect,
7:03 the segmented cache parts,
7:05 and that Kabamoto Snoop bus clearly doesn't,
7:09 because the Snoop bus is that one line,
7:12 and then the connections between the segments
7:15 in Bowman are done on something that's not the Snoop bus.
7:19 I'm sorry if that doesn't make any sense,
7:21 but it seems to me that essentially the Board's saying
7:23 the Snoop bus itself has to make direct connections
7:26 between the segmented global cache.
7:30 I mean, it's clear to me at least that Bowman
7:32 has connections between all the segmented stuff,
7:35 and when you connect it to the Snoop bus,
7:38 it's all connected.
7:39 But if the Board's construction of interconnect
7:42 requires direct interconnections by the interconnect system,
7:46 and Kabamoto's, the only interconnect system
7:50 is the Snoop bus,
7:51 then that doesn't necessarily show it.
7:53 If we disagree with the way the Board has construed
7:55 what interconnect system means,
7:58 or interconnectedness means,
8:00 and think it can be proven,
8:02 or broader, do we have to send that back?
8:04 Or is there any argument for the Board to make new findings
8:07 under kind of a more correct interpretation?
8:09 Appellant Attorney (Robert Alan Appleby) So, and we outlined this in our brief,
8:12 we do believe that the evidence below was uncontested,
8:17 that through the combination using the Snoop bus
8:21 as an interconnect system,
8:22 it would meet the three relationships
8:27 that are required by the interconnect system of Claim 5.
8:31 To be quite candid,
8:32 I'm not exactly sure what the Board envisioned
8:37 was the difficulty here,
8:39 because it simply states in the final written decision
8:43 that it found the explanation inadequate.
8:47 But its reasoning is based on the belief
8:50 that the interconnect system was challenged by PAC below,
8:55 and that we had not responded to that argument,
8:58 which PAC now admits was not the case.
9:00 The combination of Kabamoto,
9:02 and Bauman,
9:03 the Snoop bus itself is a shared bus
9:06 that interconnects everything that's attached to it.
9:11 And in Kabamoto itself,
9:13 it interconnected processors to processors,
9:17 processors to caches, secondary caches,
9:20 and secondary caches to secondary caches.
9:23 In our proposed combination below,
9:26 we were just simply replacing those secondary caches
9:30 with the cache segments of Bauman,
9:32 and with everything being connected to the Snoop bus.
9:37 And that's made clear in our expert's declaration
9:42 at Appendix 935, Paragraph 135,
9:45 where the expert outlines that each of those segments
9:48 would be connected to the Snoop bus.
9:50 Judge Prost I have the same concerns,
9:52 and I'm not sure you're really directly answering Judge Hughes.
9:56 I apologize if I missed every other question.
10:00 It's hard to get through a lot of questions,
10:02 and there are a lot of pieces of this case.
10:03 It's not just the technology.
10:05 It's kind of the arguments.
10:06 But I think maybe my analysis was like Judge Hughes.
10:11 The only way I could figure out
10:13 or try to make sense out of what the board said
10:16 with regard to Snoop bus
10:17 was to assume that it was requiring direct connection
10:21 and was not allowing for indirect connection.
10:25 So if that is the case,
10:28 and do we here say,
10:30 that's wrong, if that's what we believe,
10:33 that indirect connection was okay,
10:35 and if we reach that portion of it,
10:38 is there a reason to send it back to the board again,
10:42 or do we just say,
10:43 we agree with Intel.
10:45 The Snoop bus was what was connecting it.
10:48 We don't think that direct connection was required.
10:51 Full stop, end of story.
10:53 Appellant Attorney (Robert Alan Appleby) So I would agree with that, Your Honor.
10:54 If this court would include that connection,
10:57 direct or indirect,
10:59 meets the interconnection system,
11:01 the three relationships of the system,
11:03 that's clearly shown in the two references,
11:06 because Kabamoto has the three relationships
11:09 if we look at cash to cash,
11:11 and Bauman expressly,
11:14 and there was no dispute about this below,
11:15 expressly shows cash segment to cash segment interconnection.
11:19 Judge Hughes Well, here's where I think,
11:20 and maybe I'm just misreading the board decision,
11:23 and there's not going to be a disagreement
11:25 on the other side of this,
11:26 but what I'm reading the board's decision is saying,
11:29 the Snoop bus connects,
11:31 it has one line to the Bauman global SLC.
11:36 It doesn't have multiple lines from the Snoop bus
11:38 to each specific segmented cash,
11:41 and that's what the board required
11:43 for an interconnect system,
11:45 that it connects,
11:46 the system connects,
11:49 you know, piece to piece between the caches
11:51 by the Snoop bus,
11:54 and not just the Snoop bus
11:56 connected generally to the GL,
11:59 to the SLC,
11:59 and then within the SLC itself,
12:03 they're connected segment to segment.
12:04 Does that make sense?
12:06 I mean, that's my reading of the board's thing.
12:09 To me, that's what,
12:10 when I say I thought they required a direct connection
12:14 between the Snoop bus and each segmented cash,
12:16 that's how I read their opinion.
12:17 Clearly, the global SLC is connected segment to segment,
12:22 and once you connect it to the Snoop bus,
12:24 it's all interconnected.
12:25 That's the way data flow or whatever this is works,
12:28 I assume.
12:29 But if the board's reading of that
12:32 is what they meant by interconnect,
12:35 I mean, I assume you think that's wrong,
12:38 and if we think that's wrong,
12:40 and that indirect connections,
12:44 which is what this would be,
12:46 is good enough,
12:47 then I take,
12:49 I'm just trying to clarify again,
12:51 your answer is that if
12:52 you would not have a connection,
12:53 there's no dispute that this combination does it,
12:57 assuming we also find a motivation combined.
12:59 There would be no dispute that the combination...
13:01 Judge Newman There's no dispute that the Snoop bus
13:03 is an interconnecting system.
13:06 It's just what kind of direct or indirect?
13:11 Appellant Attorney (Robert Alan Appleby) There would be no dispute.
13:12 If indirect and direct connections satisfy,
13:15 there's no dispute that those three relationships
13:17 are satisfied through the combination
13:19 of Kauffman, Bono, and Bauman.
13:21 I believe that's answering your question.
13:24 So I believe if you find that indirect connections are okay,
13:28 then you do not need to...
13:30 Right.
13:30 If direct connections are required,
13:32 Judge Hughes then there may be a problem.
13:34 You may have arguments on that,
13:36 but let's just assume that that's direct connections.
13:39 I don't think that's right,
13:40 but if it was,
13:41 there is a little bit more of a problem
13:43 about the Snoop bus providing a direct connection
13:46 to each segment.
13:49 Appellant Attorney (Robert Alan Appleby) So let me preface that by saying,
13:51 if indirect connections are okay,
13:53 then there's absolutely no dispute
13:55 that everything's interconnected
13:57 in the way it needs to be in the claim.
13:59 Judge Prost Did the other,
14:00 on this discussion,
14:01 just on this point,
14:02 staying on this point,
14:03 was PACT making the argument,
14:05 or was this something that the board,
14:06 on the Snoop bus stuff
14:08 and the interconnected stuff?
14:09 Was this an argument they adopted
14:11 the patent owner's argument,
14:13 or was this something they did on their own?
14:15 Appellant Attorney (Robert Alan Appleby) So this was not an argument
14:16 that was raised by PACT below.
14:19 PACT never challenged
14:21 that the combination as we proposed it
14:25 would meet the interconnect system limitation.
14:28 Where the board, in my view,
14:30 was led astray is PACT below
14:33 on the issue of motivation to combine
14:36 posited two alternative combinations
14:39 under the argument that we believe,
14:42 this is PACT speaking,
14:44 that if one of ordinary skill in the art
14:47 combined these,
14:48 they would not combine them
14:49 in the way Intel had proposed,
14:51 but would combine in the way that,
14:55 in another way,
14:56 and there was two other ways.
14:57 For those two other ways,
14:59 PACT contested that the interconnect system
15:02 wouldn't be met.
15:03 That was why those alternative ways failed.
15:06 But PACT never made the argument
15:07 on our proposed combination
15:10 that it didn't meet
15:10 the interconnect system limitation,
15:13 and therefore also never made
15:15 this collective versus individual connection argument
15:18 that we've been discussing here.
15:20 Judge Newman Okay.
15:21 Let's hear from the other side.
15:22 We'll save you a little time.
15:25 Mr. Weisburst.
15:26 Appellee Attorney (Sanford Ian Weisburst) May it please the court,
15:27 Sanford Weisburst for appellee.
15:29 Thank you, Mr. Chairman.
15:30 I'd like to begin with this direct versus indirect question
15:32 because I think respectfully we're getting
15:34 a little bit ahead of ourselves
15:35 when we ask that question
15:36 because I think the question is whether the petition
15:38 articulated an indirect theory.
15:40 I think that's the first question
15:42 that this court should ask.
15:43 And the board found that it did not.
15:45 And at Appendix 797,
15:47 one of the judges,
15:48 Judge Barrett,
15:49 said in response to Intel raising this
15:53 so-called indirect theory,
15:54 I don't remember seeing that in your petition counsel.
15:57 And in fact,
15:59 if one looks at the petition,
16:00 at Appendix 216 to 217,
16:03 a couple of important points.
16:05 The only interconnect system that Intel pointed to
16:07 was the SNOOP bus.
16:09 Judge Prost So is it your view that the board decided that issue
16:11 and said direct is necessary
16:14 and they didn't allege direct?
16:15 Did the board deal with that?
16:17 Appellee Attorney (Sanford Ian Weisburst) Mr. What the board decided
16:18 was that the only theory in the petition
16:20 is one of connection to the SNOOP bus
16:24 and the internal data paths.
16:27 And this can be seen
16:28 on Appendix 216
16:29 and that there was never an articulation
16:31 of a theory of indirectness.
16:33 This sort of direct versus indirect
16:35 is perhaps shorthand
16:37 now that we're at oral argument
16:38 in the Federal Circuit.
16:39 But it was not something
16:41 that was argued in the petition.
16:42 What the petition argued is
16:44 let's take the SNOOP bus
16:45 and let's take,
16:46 there has to be under the claim
16:48 cash segment to cash segment.
16:50 Cabemotor doesn't have cash segments.
16:52 They're getting that entirely from Bauman.
16:53 And yet the petition did not say
16:56 that the SNOOP bus
16:57 somehow reaches up
16:59 and connects one segment
17:00 within the global SLC
17:01 to a neighboring segment.
17:02 Judge Prost So can you point us
17:03 to where the board discussed
17:04 the SNOOP bus
17:05 and where they,
17:06 did they identify the deficiency
17:09 in the petition?
17:10 Mr. Yes.
17:11 Appellee Attorney (Sanford Ian Weisburst) The relevant parts
17:12 of the final written decision
17:13 are in Appendix 19 through 21.
17:15 And specifically on 19
17:17 it starts with the first full paragraph
17:18 that the word petitioner next argues.
17:21 And it goes through 21.
17:23 And I think there's been some discussion,
17:24 well, PAC didn't make this argument.
17:26 And in all candor,
17:27 PAC did not make this precise argument.
17:29 PAC addressed a different combination
17:31 than Intel had proposed.
17:33 However, at these pages, 19 through 21
17:36 of the Appendix,
17:37 the board does not refer to
17:39 patent owners, PAC's arguments.
17:41 It addresses the petition
17:42 on its own merits
17:43 under this court's decision in FanDuel
17:46 that was within the board's power to do.
17:48 It can judge the petition
17:49 within its four corners
17:50 regardless of what the patent owner argues.
17:52 And, you know,
17:54 even though the board was not required
17:56 to give Intel even an opportunity
17:58 to address this deficiency,
18:00 the board did that at the hearing.
18:01 And that is at pages 774 and 797,
18:05 which I referred to earlier.
18:06 The board said,
18:07 look, I only see the SNOOP bus.
18:09 I don't see how that's connecting
18:10 cash segments to cash segments.
18:12 Please help me out, Intel.
18:13 And Intel then came up with this theory.
18:16 By the way,
18:16 it didn't refer to it
18:17 as expert's declaration,
18:18 which they talk about on appeal.
18:19 They only said,
18:20 well, there's this,
18:22 all of the cash segments
18:24 are somehow interconnected
18:25 with the SNOOP bus.
18:26 That's what they said
18:26 at the hearing at 797.
18:28 And Judge Barrett said...
18:29 Judge Hughes Well, is there any dispute
18:30 that in Bulma
18:31 the segmented cash segments
18:34 are interconnected to each other?
18:36 Appellee Attorney (Sanford Ian Weisburst) There is not.
18:36 But the key question is
18:38 what is doing that interconnection?
18:39 And in Appendix 216,
18:41 as we also explained in our brief,
18:42 it's internal data paths
18:44 that are highlighted in gold on 216.
18:46 That is what is accomplishing
18:47 the interconnection.
18:48 The SNOOP bus
18:49 and the proposed combination
18:50 is entirely outside of that.
18:53 It doesn't somehow sneak its way up
18:55 and do interconnections.
18:56 It connects to the whole global
18:58 second-level cash, certainly.
18:59 Judge Newman I'm looking to see
19:00 where in the claim
19:02 that distinction is drawn,
19:04 and I don't see it.
19:06 Appellee Attorney (Sanford Ian Weisburst) The distinction of cash segment
19:07 to cash segment?
19:08 Judge Newman The direct and indirect.
19:12 Appellee Attorney (Sanford Ian Weisburst) The claim language does not refer
19:14 to direct versus indirect,
19:15 but it does...
19:15 Judge Newman Then why are we arguing about it?
19:17 Appellee Attorney (Sanford Ian Weisburst) Because the claim language
19:19 does say an interconnect system
19:21 that connects, among other things,
19:23 a cash segment
19:24 to a neighboring cash segment.
19:25 That's the language
19:26 of the claim,
19:26 pages 108 to 109.
19:28 The question is...
19:30 Judge Hughes Is the dispute here simply
19:31 that they didn't say
19:32 the interconnect system
19:33 is the SNOOP bus
19:35 attaching to Bowman,
19:37 which itself has interconnected
19:39 data segments?
19:40 Appellee Attorney (Sanford Ian Weisburst) We might have a different case
19:42 if their petition...
19:43 Again, it has to be
19:44 on the petition.
19:45 The petition said,
19:46 we're not going to point
19:47 just to the...
19:47 Judge Hughes Let's just assume
19:48 that I read the...
19:50 That's what the argument is.
19:51 What's wrong with that argument?
19:53 Because if the board
19:54 read that argument,
19:55 that way,
19:56 and still rejected it,
19:58 it would have to be
19:59 because there was
20:00 no direct connections,
20:01 because there's certainly
20:02 indirect connections.
20:07 Appellee Attorney (Sanford Ian Weisburst) The...
20:09 Again, I don't believe...
20:11 I want to answer
20:12 your honest question.
20:13 Okay, I understand
20:13 Judge Hughes you don't believe
20:14 that that's the argument
20:15 that their petition made.
20:16 Let's assume
20:17 that the interconnect system
20:19 is...
20:20 That disposes
20:21 all these elements
20:24 is a combination
20:25 of Kabamoto and Bowman
20:27 and Kabamoto has basically
20:29 everything except
20:31 the global SLC
20:32 and that Bowman has
20:35 segmented data caches
20:37 that are connected
20:38 to each other.
20:39 And by attaching that
20:41 to Kabamoto
20:42 or through the Snoop bus,
20:44 it has everything.
20:45 Right.
20:46 If that's their theory,
20:47 what's wrong with that?
20:48 Appellee Attorney (Sanford Ian Weisburst) So, what's wrong
20:49 with that theory
20:50 is that we don't have
20:52 a motivation to combine.
20:53 And if I could turn to that...
20:53 Okay, no, but I...
20:54 That's a different issue.
20:55 Okay, sure.
20:56 Judge Hughes I'm talking about
20:57 whether there's
20:57 substantial...
20:58 Assuming that...
21:00 Because I really do,
21:02 I don't see any other reason
21:04 the board could have
21:05 to reject this combination
21:09 and have an interconnect system
21:11 unless it's requiring
21:13 that there be
21:14 a direct connection
21:15 between the Snoop bus
21:17 and each of the segments.
21:19 And so, if indirect
21:22 can be good enough,
21:23 then there's no
21:24 substantial evidence
21:25 for the notion
21:27 that this combination
21:29 doesn't disclose everything.
21:31 Appellee Attorney (Sanford Ian Weisburst) I would concede that
21:32 with the caveat
21:33 that we don't believe
21:34 the petition alleged
21:35 in indirect theory.
21:36 Judge Hughes But I understand
21:37 you have arguments
21:38 about why you wouldn't
21:39 combine that.
21:39 Appellee Attorney (Sanford Ian Weisburst) If I could turn
21:40 to the motivation to combine.
21:41 Now, the board,
21:42 after finding this deficiency
21:43 that we've been discussing,
21:45 went on to say,
21:46 even if Judge Hughes's
21:48 formulation is correct
21:49 with the petition,
21:50 here's why there's
21:51 no motivation to combine.
21:52 I'd like to address that point.
21:53 So, Kabamoto,
21:55 you've got these several
21:56 processor elements
21:57 in sequence.
21:58 Okay?
21:58 And you've got a cache
21:59 that's not a segmented cache.
22:01 It's just a cache.
22:02 So they need to go to Bauman
22:03 to get the cache segments.
22:06 And they rely mostly
22:09 on appeal
22:09 on this known technique idea.
22:11 And they throw in an argument
22:11 about KSR.
22:12 But really,
22:13 the question I'd like
22:14 the court to focus on
22:15 is what is the known technique?
22:17 They talk about, well,
22:18 segmenting a cache
22:19 into cache segments
22:20 instead of just
22:21 an unsegmented cache.
22:22 That's a known technique.
22:24 Everyone does it.
22:24 It's in Bauman.
22:25 It's well known in the art.
22:27 And so,
22:27 the problem is that
22:28 that doesn't get you
22:29 to the combination.
22:30 We'll grant Intel
22:32 that segmentation
22:33 is a known technique.
22:34 But what are they doing here?
22:36 They are taking
22:36 these three or four
22:38 processor elements
22:39 from Kabamoto.
22:40 They're taking the cache
22:41 outside of the processor element.
22:44 Okay?
22:44 It's 14-1 in the diagram.
22:46 They take it outside
22:47 the element.
22:48 Then they segment it.
22:50 And then,
22:50 on top of all that,
22:52 they connect it
22:52 to the snoot bus.
22:53 So all of those things
22:55 are part of the combination
22:56 that they are proposing.
22:57 And I don't think
22:58 there's a dispute
22:58 that that's what
22:59 they're proposing.
22:59 It's at Appendix 217.
23:01 The board quoted their petition.
23:02 So,
23:03 it's not enough
23:04 just to say,
23:05 well,
23:06 it's a known technique
23:07 to segment a cache.
23:08 Because this idea of...
23:09 Yeah,
23:09 Judge Prost but can you look at what...
23:10 I'm looking at
23:11 what the board said
23:12 about that argument
23:13 on known technique.
23:15 And what the board said,
23:17 which I think
23:17 is not correct
23:19 under the law,
23:20 which is that
23:21 if this petitioner
23:22 argues
23:23 Kabamoto
23:24 already addresses
23:25 its problem
23:26 through the use
23:27 of a known technique
23:28 similar to that
23:29 of Bauman's,
23:30 we fail to see
23:31 why one of ordinary
23:32 skill in the art
23:33 would regard
23:34 Bauman's technique
23:34 as an obvious improvement
23:36 to Kabamoto.
23:37 Is that right?
23:38 Is that legally
23:39 sustainable?
23:40 Appellee Attorney (Sanford Ian Weisburst) That is right
23:41 in response
23:42 to the specific
23:43 supposed known technique
23:44 of segmentation.
23:45 I think that was
23:46 what the board
23:47 was addressing.
23:47 And that's at Appendix 28.
23:48 So the board was saying,
23:49 look,
23:50 if the known technique
23:51 is let's segment the cache,
23:53 you could do that
23:54 within Kabamoto's structure.
23:55 You don't have to
23:56 do that.
23:56 It doesn't require
23:57 you to combine
23:58 this global second level cache.
23:59 You could just,
24:00 and in fact,
24:00 Intel suggests,
24:01 well,
24:01 Kabamoto has sublines
24:03 which sort of suggest
24:04 that you could segment.
24:05 Yes,
24:05 you could segment,
24:06 but you would do it
24:07 within Kabamoto's structure.
24:08 You wouldn't pull the cache
24:09 outside the processor elements
24:11 and you wouldn't connect it
24:12 to this new bus.
24:13 Now in Bauman...
24:14 Why?
24:14 Judge Prost I mean,
24:14 I guess I'm trying
24:16 to understand
24:16 under KSR
24:17 and under our law,
24:18 the board admits
24:19 that the two address
24:20 the same problem.
24:22 Why isn't that
24:26 only two different references?
24:27 So they don't,
24:28 it's not an anticipation,
24:29 this is an anticipation case.
24:31 So because Kabamoto
24:33 addresses the same problem
24:35 differently,
24:35 why is that evidence
24:37 that there's no motivation
24:38 to combine?
24:39 Sure.
24:39 Appellee Attorney (Sanford Ian Weisburst) So,
24:41 Bauman is not,
24:42 Bauman does involve,
24:43 does involve segmentation,
24:45 but it's focused on,
24:46 and the board found this,
24:48 focused on a different question,
24:49 which is separating
24:50 the cache segments
24:51 into instruction
24:51 and operand.
24:52 Instruction is sort of
24:53 the formula
24:54 and the operand
24:56 is the data
24:56 that you apply that to.
24:57 That is the focus
24:58 of Bauman.
24:59 But even more important
25:00 than that,
25:00 if you look at Bauman,
25:02 how is the global...
25:03 Judge Hughes Are you saying
25:04 that Bauman
25:04 also doesn't show
25:06 that a global SLC
25:09 is a known technique
25:10 for maintaining
25:11 cache coherence?
25:13 Appellee Attorney (Sanford Ian Weisburst) Yes.
25:14 Certainly,
25:15 certainly not
25:15 in connection
25:16 with a snoot bus
25:17 because there is
25:18 no snoot bus...
25:19 Judge Hughes No, no, no, no, no.
25:19 That's not what they're,
25:20 they don't need
25:21 to get to Bauman
25:22 for the snoot bus.
25:23 The snoot bus
25:24 is in Kabamoto.
25:25 The question,
25:26 as I understand it,
25:27 is the issue here
25:29 is maintaining cache coherence
25:31 in multiprocessor systems.
25:33 Kabamoto does it one way.
25:35 Bauman's global SLC
25:37 is a different way
25:38 of maintaining
25:39 cache coherence.
25:40 If that's what they're
25:41 trying to get,
25:42 why isn't Bauman's technique
25:43 just another technique
25:45 of maintaining
25:45 cache coherence?
25:46 Appellee Attorney (Sanford Ian Weisburst) Respectfully,
25:47 it is in part
25:48 cache coherence,
25:49 but it's not entirely
25:50 cache coherence.
25:50 There's another aspect
25:51 of this which is how...
25:52 Well, who cares
25:52 Judge Hughes that there's other aspects?
25:55 There's a way
25:55 to maintain
25:56 cache coherence
25:57 by having a global SLC
26:00 versus separate caches.
26:02 Appellee Attorney (Sanford Ian Weisburst) Here's the issue,
26:03 is that in order
26:04 for the cache
26:05 to have value
26:05 in the system,
26:06 the processors
26:07 need to be able
26:08 to get to it,
26:09 not just for cache coherence,
26:10 but actually to access
26:11 the data
26:12 and work on the data.
26:13 In Bauman,
26:15 the connection,
26:15 each processor
26:16 has its own connection.
26:18 There's no connection
26:19 between processors.
26:20 The processor
26:21 directly connects
26:22 to the second level cache.
26:23 In Kabamoto,
26:24 the processors
26:25 are all connected
26:26 to each other
26:27 via the snoop bus.
26:27 Judge Hughes Are you saying
26:28 this shows it's inoperable,
26:29 this combination
26:30 would be inoperable?
26:32 Appellee Attorney (Sanford Ian Weisburst) No, we're not...
26:32 Judge Hughes So you're not making that...
26:34 Appellee Attorney (Sanford Ian Weisburst) We're not making
26:34 that argument,
26:35 but what I think
26:35 the board was getting at
26:37 at Appendix 28
26:38 was the only known technique
26:40 that's been identified
26:41 is segment the cache.
26:43 But there's a lot more
26:44 to this combination
26:45 than that.
26:46 You're not only segmenting,
26:47 but you're also pulling
26:48 the cache
26:48 outside the processor
26:49 and then you're connecting
26:50 it to the snoop bus.
26:51 And Bauman does not
26:52 use a snoop bus.
26:53 When you introduce
26:54 a snoop bus to Bauman,
26:55 you introduce a question
26:57 of competition
26:58 among the processors
26:59 for access
27:00 to the global
27:01 second level cache.
27:02 Judge Hughes But you're not saying
27:02 it won't work.
27:03 That's the problem.
27:04 I mean,
27:05 if it's inoperable,
27:07 there's no motivation
27:07 to combine.
27:08 But if it can accomplish
27:10 the goal,
27:11 even if it may be
27:12 not as efficient,
27:13 that doesn't mean
27:14 there's no motivation
27:15 to combine it, does it?
27:16 Haven't we said
27:17 that you don't have to show
27:18 that it's the best way
27:20 of doing it,
27:20 that it's just
27:21 a way of doing it?
27:22 Appellee Attorney (Sanford Ian Weisburst) It doesn't have to be
27:23 the best way.
27:24 They did articulate
27:25 a theory that has...
27:26 I mean,
27:27 Judge Hughes to show no motivation
27:28 to combine
27:29 on this argument,
27:29 don't you have to show
27:30 that it would be inoperable?
27:32 Appellee Attorney (Sanford Ian Weisburst) No, I don't believe
27:33 that's our burden.
27:34 I think in terms
27:35 of a motivation
27:36 to combine,
27:37 we're not trying
27:38 to show it's inoperable.
27:39 What we're trying to show
27:39 is based on the evidence
27:41 in the record,
27:42 was there substantial evidence
27:43 to find that the benefits
27:44 of that combination
27:45 were outweighed
27:46 by the cost
27:47 of that combination?
27:48 And would a skilled
27:49 artisan come to that?
27:50 I believe
27:51 that's our burden,
27:52 and the word found
27:53 looking at all this,
27:54 it said, look...
27:55 Judge Prost But you're saying
27:55 that their burden
27:56 for motivation
27:57 is to show
27:58 that the benefits
27:59 outweigh the cost?
28:00 I mean,
28:00 it doesn't have to be better,
28:02 right, under KSI?
28:03 No, it has to be something
28:04 that a skilled...
28:05 It doesn't have to be
28:05 the best.
28:06 Appellee Attorney (Sanford Ian Weisburst) It has to be something
28:07 that a skilled artisan
28:07 would be motivated to do,
28:09 and I'd refer the court
28:10 to the Henny Penny decision,
28:12 which...
28:12 Judge Prost So you're saying
28:13 that because it was costly
28:14 or it was difficult
28:15 or challenging,
28:18 then that destroys
28:19 the motivation
28:20 to combine?
28:21 What we're...
28:23 I mean,
28:23 we see motivation
28:24 to combine
28:25 different references
28:26 and different fields
28:27 of technology and stuff.
28:28 These all seem
28:29 very close
28:30 and very related
28:31 under the same umbrella, right?
28:33 It doesn't seem
28:34 a stretch
28:34 to be picking
28:35 and choosing
28:36 between Bowman
28:36 and Kawamoto, right?
28:38 Appellee Attorney (Sanford Ian Weisburst) Well, again,
28:39 I think,
28:40 to the extent
28:40 reasonable people
28:41 could disagree,
28:42 which we think
28:42 is the case here,
28:43 that a skilled artisan
28:44 would be motivated,
28:46 taking into account
28:46 all of the circumstances
28:47 that the board
28:48 had substantial evidence
28:49 to make that finding.
28:53 I mean,
28:53 it's a similar field,
28:53 perhaps,
28:54 but Bowman
28:55 is using
28:56 an entirely different approach.
28:57 Okay?
28:58 It's got a global
28:59 second-level cash...
29:00 Judge Hughes I mean,
29:00 here's the problem,
29:01 is if you look
29:02 at the board's decision
29:03 for rejecting it,
29:04 it rejects it
29:05 on what seems
29:05 to be improper grounds.
29:07 On 27,
29:08 it says,
29:09 well,
29:09 it doesn't actually
29:10 improve the thing.
29:11 And then on the bottom
29:12 of 27,
29:13 in 28,
29:14 it goes on
29:15 and describes
29:16 exactly
29:17 what I said
29:19 was the...
29:23 Subbing in
29:23 a global segment
29:25 in an SLC
29:26 for Kabamoto's
29:27 is just subbing
29:28 in a different
29:29 technique
29:30 to maintain
29:30 cash coherence.
29:31 And then on 28,
29:33 it says,
29:33 well,
29:34 Kabamoto already does that,
29:35 so you wouldn't need
29:36 to go to Bowman
29:37 for that.
29:37 But that's legally incorrect
29:39 under KSR 2.
29:40 So,
29:41 the problem is,
29:42 I understand
29:42 your argument,
29:43 but the arguments
29:44 that the board
29:45 has articulated here
29:46 don't seem to me
29:48 to be legally correct.
29:49 It's not a question
29:50 of substantial evidence.
29:51 They seem to be incorrect
29:52 under KSR
29:53 as requiring them
29:55 to show an improvement
29:56 or them to...
29:58 Or that the art...
30:01 The reference
30:01 they're teaching
30:02 already does this,
30:03 so why would you
30:03 go to another one?
30:05 We've rejected it,
30:06 particularly
30:06 that latter one
30:07 we've rejected
30:09 other times,
30:10 I'm pretty sure.
30:12 Appellee Attorney (Sanford Ian Weisburst) So,
30:12 KSR does not
30:14 require an improvement.
30:15 I agree with that.
30:16 However,
30:16 if the petitioner
30:17 itself attempts
30:18 to allege
30:19 an improvement,
30:19 this court
30:20 and the ARctic Cat
30:21 decision
30:21 will not
30:21 support
30:21 that we cite
30:22 in our brief
30:22 says that it can hold
30:24 the petitioner
30:24 to that burden.
30:25 But the point
30:26 I'd like to end with here
30:28 is just that
30:28 it's not just
30:30 about segmentation.
30:31 It's about segmentation,
30:32 removing the cash
30:34 from the processor
30:34 and connecting it
30:35 to a SNOOP bus.
30:36 There's no SNOOP bus
30:37 in Bauman.
30:38 And because of
30:39 the obstacles there,
30:40 not that they're
30:41 physically inoperable,
30:42 but they are enough
30:43 that a skilled artisan
30:44 might hesitate
30:45 and not think
30:46 of that way
30:46 of achieving
30:47 the benefit of segmentation.
30:51 Judge Hughes But the court
30:51 say that.
30:53 Appellee Attorney (Sanford Ian Weisburst) The board
30:53 on
30:54 You're giving
30:54 Judge Hughes a pretty good
30:56 articulation of
30:56 a basis.
30:57 I'm not sure
30:58 I agree with it,
30:59 but I didn't read
30:59 the board's opinion
31:00 anywhere saying that.
31:01 Appellee Attorney (Sanford Ian Weisburst) So I read the board
31:02 in 27 to 28
31:03 to say that
31:04 it's not
31:05 That's what we just
31:05 looked at.
31:06 Yes.
31:06 Judge Hughes It didn't say that though.
31:08 Appellee Attorney (Sanford Ian Weisburst) I think
31:09 the board says
31:11 segmentation
31:11 is not all
31:12 that's going on here
31:13 and the board says
31:13 if you wanted
31:14 segmentation you would do in Kawamoto the board had other evidence and respectfully if the court
31:18 thinks that well the board didn't explain those other aspects enough I think the remedy would be
31:23 a remand as opposed to a reversal although again we would encourage the board to find that the
31:28 board had substantial evidence to find that the petition had a deficiency in its theory before
31:32 Judge Prost you sit down I just have one more process sort of APA question I understood you as not contesting
31:38 one of their main arguments this is just on the process APA stuff which is that the board credited
31:45 your argument this is on the interconnected stuff as being your argument and they're adopting your
31:51 argument and it is that is incorrect in other words you didn't make the argument the board
31:57 came up with this argument and therefore at a minimum uh they didn't have an opportunity to
32:04 respond am I is that correct we agree that the board came up with the argument
32:08 Appellee Attorney (Sanford Ian Weisburst) and did so without referring to the patent owner's filings that's at 19 to 21. we disagree
32:13 we disagree on two points one that the board had to provide an opportunity to respond but
32:17 in any event the board did provide an opportunity to respond at the hearing
32:21 Judge Prost appendix 774 and 797 and intel but didn't the board give you credit for making the argument
32:29 that you did not make and that the board came up with I disagree with that um I think at appendix
32:34 Appellee Attorney (Sanford Ian Weisburst) 19 through 21 the board sets this argument out without any reference to the patent
32:38 Judge Newman owner's filings thank you anything else for this council yeah thank you Mr. West first Mr. Appleby
32:50 Judge Hughes thank you your honor can you go straight to the motivation and come on because I'm still
32:55 struggling a little bit about what we do with this case if we agree that the board's findings
33:01 at 27 and 28 are legally incorrect as a basis for finding no motivation to go by then
33:12 normally I would think that we would have to remand it for the board to make a determination
33:16 under the correct legal standard but I don't I mean once you get these two references in the
33:25 same field using very similar techniques and everything is there any substantial evidence
33:32 that would still support a finding of no motivation to combine I'm not aware of any your honor if we
33:39 Appellant Attorney (Robert Alan Appleby) look at these references
33:42 both references directed to improving coherence mechanisms for caching their very similar
33:50 systems that use multi processors with private first level caches and secondary are second level
33:58 caches and the and the motivation combined here is drawn directly from that tight interrelationship
34:05 between these references in that a person of ordinary skill in the art would recognize that a
34:12 Unknown あります
34:12 Appellant Attorney (Robert Alan Appleby) shared, segmented second-level cash was a known alternative.
34:17 Judge Hughes So then, I'm sorry to cut you off, but I get your argument on that.
34:21 It's helpful.
34:22 Can you address what your friend said about,
34:27 because I think he was reading the board's opinion at 27 and 28.
34:32 I mean, he was rewriting it to me in a way that made more sense.
34:36 But can you address that, that the problem here is not just subbing in
34:40 a Bauman's Global SLC for Kabamoto's separated caches,
34:47 but rather that it requires not just the subbing in of those two things,
34:52 but also using the Snoop bus, and because of the distance between the two,
34:57 that a skilled artisan would not combine them in this particular arrangement?
35:04 Appellant Attorney (Robert Alan Appleby) Right.
35:05 So we would disagree with that.
35:07 A Snoop bus is simply a shared bus,
35:10 and it's called a Snoop bus because it snoops on caches.
35:14 It looks at transactions that move between caches and caches.
35:18 So a Snoop bus, which mutually interconnects everything that's connected to it,
35:24 would be the most natural place to place any caching mechanism.
35:28 So a person of ordinary skill in the art looking at Bauman with its shared second-level cache
35:35 and looking at Kabamoto that already interfaces its caches through,
35:40 a Snoop bus would have been naturally motivated to place that second-level cache on the Snoop bus.
35:45 I think the issue that in Bauman there's a particular one-to-one connection with the processors
35:51 is not material to that question because one of us skilled in the art would understand
35:56 that a shared bus makes those interconnections as well.
36:00 So this is really just a classic case of a known alternative, a shared second-level cache,
36:08 being a substitute for individual...
36:10 individual second-level caches, and this is a straight under the flexible and expansive analysis of KSR.
36:18 Judge Hughes I mean, you would agree that if they had shown subbing in Bauman's SLC made Kabamoto inoperative,
36:26 that that would be a motivation not to combine.
36:31 I mean, I think he said that they weren't making that argument.
36:35 They weren't making that argument, and certainly...
36:36 But if combining two references would make it inoperable,
36:41 not that...
36:41 Not that it doesn't work better, but that it's just inoperable,
36:44 that's a finding against a motivation.
36:46 Appellant Attorney (Robert Alan Appleby) Well, KSR itself says that if we're just rearranging known elements to yield predictable results,
36:52 that's obvious unless that's beyond the skill of one skilled in the art.
36:57 So I think if it was inoperable, you might put that in that basket.
37:01 Judge Prost We've got cases that say that, that if it's inoperable, that could be a basis for rejecting motivation.
37:05 Appellant Attorney (Robert Alan Appleby) Right, and I think that that's what KSR implies if you have...
37:09 You're just rearranging known elements.
37:13 That are doing the same function that they're already performing to yield predictable results.
37:18 It's obvious unless that was something that was beyond the skill of a person of ordinary skill in the art,
37:23 which inoperableness would, I think, would fall into that basket.
37:26 That is not the analysis that the board did.
37:28 The analysis the board did clearly looked for a teaching of an improvement
37:37 and looked for a requirement that the combination was actually improvement,
37:41 over known systems, which just contravenes many cases from this court's precedent,
37:49 including Intel versus Qualcomm, which simply says that the combination just needs to be a suitable alternative.
37:58 And with that, if the rest of the court has any other questions, I will...
38:02 Judge Newman Anything else for counsel?
38:04 Anything else for counsel?
38:05 Thank you.
38:06 Thanks to both counsels.
38:07 This case is taken under submission.