[00:00:00] Speaker 01: Our next and last case this morning is netlist Inc vs. Samson Electronics Company LTD 24-2240 You don't have to move Councilor Lemkin you have four minutes of rebuttal time. [00:00:19] Speaker 03: Yes, thank you. [00:00:20] Speaker 01: All right All right [00:00:25] Speaker 03: the court. [00:00:26] Speaker 03: It's good to be back. [00:00:29] Speaker 03: The invention here allows more memory to be jammed. [00:00:31] Speaker 01: What took you so long? [00:00:36] Speaker 03: The invention here allows more memory to be jammed into a single stacked package critical for burgeoning AI while resolving load and power problems that would otherwise slow those structures down. [00:00:47] Speaker 03: I'd like to begin with claim one of the 060 patent where the court adopted a combination that required abandoning the references principle of operation before turning to claim seven [00:00:58] Speaker 03: which is reducing a difference in load by selecting the number of array dies, and claim 15, which requires multiple drivers, different sizes, not taught in the prior arts version of a single variable power driver. [00:01:09] Speaker 03: For the 060 and 160 patents, PTAB committed legal error in ruling that skilled artisans would be motivated to combine Kim, which does not have their shared TSVs across array dies, and warns that [00:01:25] Speaker 03: and warns against that, with Rajan's figure four to arrive at something with shared TSVs. [00:01:30] Speaker 03: And the problem is that it's fundamentally at odds with the nature of Kim's invention. [00:01:35] Speaker 03: If you turn to the Kim patent on page 3629 of the appendix, there's no real dispute that Kim teaches having a D round. [00:01:44] Speaker 03: where you take out the DRM array dies, you take out the IO, the data input-output circuit that controls timing, and instead you move it out and you have them share that timing circuit. [00:01:56] Speaker 03: So your different array dies are gonna be sharing that IO. [00:01:59] Speaker 03: And it explains that without those individual, this is paragraph 42 on 3631, without those individual IOs, one for each array die, you're gonna have a problem with unavoidable data collisions. [00:02:10] Speaker 03: Because the write takes a while to be effectuated. [00:02:14] Speaker 03: The read is immediate. [00:02:15] Speaker 03: So if you do a write followed by a read, the data is going to collide. [00:02:18] Speaker 03: As Linkim teaches, a remedy for that. [00:02:21] Speaker 03: It has what's called a rank selection circuit. [00:02:23] Speaker 03: Basically a switch that's going to be connected to one radii or another. [00:02:27] Speaker 03: It chooses one. [00:02:27] Speaker 03: Select the rank. [00:02:28] Speaker 03: Which one are you going to be connected to? [00:02:31] Speaker 03: That switch [00:02:32] Speaker 03: That whole invention is the exact opposite of our invention. [00:02:35] Speaker 03: Our invention requires the dies, multiple dies, to be an electrical connection. [00:02:40] Speaker 03: It requires a group of array dies, an electrical connection. [00:02:43] Speaker 03: The rank selection switch is teaching the opposite. [00:02:46] Speaker 03: Connect to one and want only one at a time to avoid data collisions. [00:02:50] Speaker 03: And the board's treatment of Kim basically abandons the entirety of its principal operation. [00:02:58] Speaker 03: Kim is about, hey, share IO circuits. [00:03:01] Speaker 03: and then have a fix because you're gonna have data collisions. [00:03:04] Speaker 03: The fix is one TSV per IO, excuse me, one TSV per array die, so you don't have any sharing, and then when it comes time to hit the IO circuit that's shared, have a switch, connect only one at a time. [00:03:18] Speaker 03: And the only way to connect, to combine Kim with Rajen is to say, we're gonna throw out everything that Kim taught as an invention, we're gonna throw out everything that Kim taught as a solution to the problem, we're gonna ignore Kim's problem of [00:03:30] Speaker 03: these unavoidable data conflicts. [00:03:32] Speaker 02: Now the CHEM can never expressly limits itself to one-to-one relationships, does it? [00:03:38] Speaker 03: Oh no, it does, actually. [00:03:39] Speaker 03: It uses the phrase, and this is on page 3629, paragraph 17, [00:03:44] Speaker 03: It says the TSVs are respectively connected to the dies, each die. [00:03:48] Speaker 03: Each TSV to each respectively connected. [00:03:51] Speaker 03: That phrase, respectively, means one to another. [00:03:54] Speaker 02: And in addition to that... We don't have a claim construction here of respectively, but for you to prevail on this issue, do we need to agree with you that Kim [00:04:06] Speaker 02: and I recognize all the disclosed embodiments are one-to-one. [00:04:09] Speaker 02: Do we have to agree with you that Kim only teaches a one-to-one relationship? [00:04:14] Speaker 03: Well, I don't think you need to teach that Kim only teaches a one-to-one relationship, but Kim definitely has a solution. [00:04:20] Speaker 03: Its invention is to maintain that one-to-one relationship because it has the rank selection switch. [00:04:25] Speaker 03: If you have one line coming in, [00:04:28] Speaker 03: You choose that, and then you might choose another one. [00:04:30] Speaker 03: But that rank selection switch, the entirety of its invention, if you're going to have a shared IO circuit, you need to have the selection switch, so you're only connecting to one at a time. [00:04:39] Speaker 03: That is impossible to reconcile with the notion of sharing TSVs. [00:04:44] Speaker 03: It is impossible to reconcile with combining it with Rajan in order to come up with this shared TSV version that is our invention. [00:04:54] Speaker 03: You're going to have to have the motivation to combine Kim itself when Kim tells you unavoidable data collisions if you're sharing, when Kim tells you use a rank selection slip to avoid it. [00:05:03] Speaker 03: And it's especially problematic because if you look at Rajan itself, figure two, [00:05:08] Speaker 03: 3644. [00:05:09] Speaker 03: Rajan has multiple different ways where you can hook things together, but it's figure two on 3644. [00:05:16] Speaker 03: Doesn't have shadow ray dyes. [00:05:18] Speaker 03: it's figure six on the next page, doesn't have those shared TSVs again. [00:05:25] Speaker 03: So Rajan gives you a menu of options, and I don't understand how you could start with Rajan in these figures where you have one array die, one TSV as an option, and say, well, I'm being motivated to combine it with Kim, which teaches exactly the opposite of that. [00:05:40] Speaker 03: If I could turn briefly to claim seven then, unless the court has more questions on Kim and the combination of Rajan. [00:05:47] Speaker 03: The basic idea is you can't combine Kim when Kim teaches the exact opposite. [00:05:51] Speaker 03: You can't combine with Kim when Kim's principle of operation would have to be abandoned. [00:05:55] Speaker 03: So on claim seven, and claim seven requires selecting the number of array dies to each CSV, quote, to reduce a difference in load. [00:06:06] Speaker 03: The PTAB construed that to Encomus Rio, which teaches having equally TSVs, so there's no difference in load to begin with. [00:06:14] Speaker 03: That omits both elements of claim seven. [00:06:17] Speaker 03: Claim seven says, because Rio doesn't teach selecting the number of radii. [00:06:21] Speaker 03: It says nothing about the number of radii at all. [00:06:24] Speaker 03: And it doesn't do it to reduce the difference in load, because Rio has no difference in load. [00:06:28] Speaker 03: It's simply not possible to reconcile that with the actual construction of the claim. [00:06:34] Speaker 03: And Rio's pretty much the opposite of mention. [00:06:36] Speaker 03: Rio's fundamental insight is have equal linked TSVs so you don't have a difference in load. [00:06:42] Speaker 03: That's one way to solve a problem. [00:06:43] Speaker 02: I understand why you're saying construe, but here I think that's a problem for you. [00:06:48] Speaker 02: There is no construction of reduce a difference. [00:06:52] Speaker 02: So while you argue that reduce a difference means you always every time to practice this claim must have some difference that's non-zero, nothing actually reads out the embodiment where [00:07:07] Speaker 02: Occasionally, there is zero difference, so we don't need to reduce it. [00:07:11] Speaker 03: No, I think it does for two reasons. [00:07:13] Speaker 03: First, it is a claim construction here. [00:07:15] Speaker 02: Appendix 123, it says... Sorry, you didn't ask for this construction that you're now implicitly asking for from us, correct? [00:07:24] Speaker 03: Well, we did argue exactly what we thought the plain meaning was, and the plain meaning says you have two things, you have a reduction in load, and you adjust the number of radii in order to reduce the load. [00:07:36] Speaker 01: But now you're arguing something other than plain meaning. [00:07:38] Speaker 03: No, I don't think so, because the words on the claims are to reduce the difference in load, [00:07:42] Speaker 03: Right? [00:07:43] Speaker 03: And selecting the number of array dies. [00:07:45] Speaker 03: Rio doesn't have a difference in load and it doesn't select a number of array dies to correct it. [00:07:50] Speaker 03: It chooses an entirely opposite solution. [00:07:53] Speaker 03: You can't get there from here because Rio teaches, hey, let's just not have a difference in load to begin with by having equal length TSVs. [00:08:02] Speaker 03: And then it doesn't teach you anything at all about selecting a number of array dies in view of load. [00:08:07] Speaker 03: It's indifferent to that. [00:08:09] Speaker 03: In fact, when it gets there, when it talks about the possibility of load differences, it teaches a different solution entirely. [00:08:16] Speaker 03: If the court turns to page 3688, paragraph 99, and this is [00:08:30] Speaker 03: It talks on paragraph 99 about manufacturing variations about four lines down in TSVs. [00:08:37] Speaker 03: And it says, oh, there might be slight differences in load because of these manufacturing variations. [00:08:41] Speaker 03: And then it proposes a solution. [00:08:45] Speaker 03: Test them, and the resistance values, the resistance elements, can be finely adjusted by the use of trimming elements. [00:08:52] Speaker 03: So it's thinking the exact opposite. [00:08:54] Speaker 03: It's not saying, oh, adjust the number of array dies. [00:08:56] Speaker 03: It's saying, let's add more resistance. [00:08:59] Speaker 03: That's, again, the exact opposite. [00:09:01] Speaker 03: The words of the claim themselves require selecting the number of array dies. [00:09:05] Speaker 03: not in Rio, and it needs to reduce the difference in load. [00:09:09] Speaker 03: There's no difference in load in Rio. [00:09:10] Speaker 03: It's really quite impossible to reconcile these two. [00:09:13] Speaker 02: Can you help me with table one in your specification, for instance, in the 060 at A187? [00:09:20] Speaker 02: And I'll admit, this technology I find very complicated, but some of the numbers on that table are identical to each other if I go across [00:09:30] Speaker 02: different entries and plus it's all talking about changing the number of array dies or the length of the TSVs. [00:09:40] Speaker 02: Wouldn't one of skill in the art looking at that table contemplate that sometimes the inventor had in mind that there would not be a difference? [00:09:47] Speaker 02: that needed to be reduced occasionally. [00:09:50] Speaker 03: Well, so if you didn't have a difference and you didn't adjust the array dies, you wouldn't be practicing the invention. [00:09:56] Speaker 03: That's correct. [00:09:57] Speaker 03: But if you have a difference and you do an adjustment, as table one shows, to reduce the difference, then you have practiced the invention. [00:10:05] Speaker 03: So if you just practiced Rio, you wouldn't be infringing. [00:10:08] Speaker 03: But the problem is Rio doesn't show and doesn't anticipate this because Rio doesn't have a difference in resistance, a difference in load, and it doesn't have an adjustment in the number of radii. [00:10:20] Speaker 03: It's simply missing the two critical components of this claim. [00:10:23] Speaker 03: And it's really hard to say that Rio somehow shows that this was already out there and skilled artists knew exactly what to do. [00:10:32] Speaker 03: If I could turn then, time remains to Wyman. [00:10:38] Speaker 03: Yes, Wyman. [00:10:40] Speaker 03: This is the invention that says that where you move to different multiple size drivers as opposed to having multiple different size drivers as opposed to having a single variable driver. [00:10:52] Speaker 03: And so the claim is that a first data conduit comprises a first driver having a first size and a second data conduit having a second driver size. [00:11:01] Speaker 03: And the board said, well, we're going to say that Wyman discloses that. [00:11:06] Speaker 03: But Wyman is, again, the opposite. [00:11:07] Speaker 03: Wyman talks about, its title is Variable Off-Chip Drive. [00:11:12] Speaker 03: So we have a single variable-sized drive. [00:11:14] Speaker 03: And if you looked at Figure 5, you can see it's on page 3699. [00:11:18] Speaker 03: You can see the driver. [00:11:20] Speaker 03: It's 500. [00:11:20] Speaker 03: And what it teaches is, at these various points along this driver, [00:11:30] Speaker 03: you can tap off and you can get different power sections. [00:11:34] Speaker 03: So if you want very little power, you go at T0. [00:11:37] Speaker 03: If you want more, T1, more T2, more T3. [00:11:41] Speaker 03: That is not a set of, that is not lots of different drivers, it's a single variable driver. [00:11:46] Speaker 03: You simply can't take this claim and get to it from Wyman when Wyman is teaching the opposite, these variable-sized drivers. [00:11:54] Speaker 03: In fact, Wyman actually disparages our clay, these individual ones. [00:11:58] Speaker 03: It describes having individual drivers as wasteful and inefficient, and it describes it as having its own technique of having one variable one as an advantage, a level of design flexibility, because it enables a designer to adjust and minimize power consumption. [00:12:13] Speaker 03: That's again the opposite of what this claim teaches. [00:12:17] Speaker 03: Maybe the board can come up with some other rationality. [00:12:20] Speaker 03: But looking at Wyman as sort of the sole basis, Wyman's teaching the exact opposite. [00:12:24] Speaker 03: There's no way you can get there from here. [00:12:26] Speaker 03: If the court has no questions on this, I will surrender the remaining three minutes I have and thank the court for its time and attention. [00:12:39] Speaker 00: I guess my colleague actually wants those three minutes to rebuttal. [00:12:44] Speaker 00: You're going to give it away. [00:12:46] Speaker 00: I'll give them to you if you want them. [00:12:49] Speaker 00: It's been a little bit longer for me, Your Honors, but I'll use the same order that you heard from my colleague. [00:12:54] Speaker 00: So starting out with claim one, we go straight to Kim in this paragraph 42 and this idea of unavoidable collisions. [00:13:05] Speaker 00: Paragraph 42 is cited over 10 times in each one of their briefs, but I'll note [00:13:09] Speaker 00: Paragraph 44 gets no action whatsoever in the briefs, the net list files, and there's a very good reason for that. [00:13:17] Speaker 00: Paragraph 44 of Kim explains that the way you get out of the unavoidable collision, and this is at lines four through six of paragraph 44, so we're on 3631. [00:13:30] Speaker 00: My apologies for not giving you the page number, and I'll certainly let you get that. [00:13:40] Speaker 00: It's 3631, Your Honor, and that is in the first volume. [00:14:04] Speaker 00: So paragraph 44, if you look at lines 5 and 6, you'll see the way you get around the data collisions as described above is by using chip selection signals. [00:14:16] Speaker 00: That's what it says there. [00:14:18] Speaker 00: So when we keep hearing about unavoidable, unavoidable, unavoidable, the reality is the prior art said they're avoidable and here's how you do it, chip selection signals. [00:14:28] Speaker 00: This is really important because this is the problem with the arguments that were made below by Netlist that the board caught on to. [00:14:35] Speaker 00: And the board analyzed this in its analysis on page 107 of the record. [00:14:43] Speaker 00: So if you'll move with me to 107, which is in the final written decision for the 060 patent. [00:14:54] Speaker 00: You'll see in the first full paragraph there, Dr. Brogioli, who is Netlist's... Oh, sorry. [00:15:00] Speaker 00: You're not there yet, Your Honor. [00:15:05] Speaker 01: What is the page again? [00:15:06] Speaker 00: 107, Your Honor. [00:15:06] Speaker 00: 107. [00:15:13] Speaker 00: So this is the first full paragraph. [00:15:15] Speaker 00: Dr. Brogioli's assumption of no change in operation with the addition of two ranks fails to account for the combined teachings of Kim and Rajan. [00:15:24] Speaker 00: and the knowledge and skill of this person of ordinary skill in the art. [00:15:27] Speaker 00: And then at the bottom of that page, the board points out, Dr. Wolf, however, testifies that Rajan teaches emulation techniques, including rank multiplication techniques, and that Rajan, turning the page to 108, teaches rank multiplication by generating chip select signals for each chip in the stack. [00:15:47] Speaker 00: In other words, [00:15:49] Speaker 00: Samsung proposed exactly what Kim says solves the unavoidable collisions. [00:15:54] Speaker 00: You use chip select signals for each chip in the stack. [00:15:58] Speaker 00: So all these citations to paragraph 42 are ignoring the actual combination as the board realized that Samsung had proposed. [00:16:07] Speaker 02: Doesn't the board also allude, though, to suggesting that want to spill in the art wouldn't really care that much about avoiding collisions, and this really isn't a problem, which would not be supported by the facts, I think. [00:16:20] Speaker 00: Well, that's a separate argument that the board makes. [00:16:24] Speaker 00: They are pointing to the fact that their patent says reduce or prevent, which kind of suggests that, well, maybe it's only OK to reduce. [00:16:32] Speaker 00: But frankly, Your Honor, when you look at that discussion, and that discussion is at, [00:16:37] Speaker 00: 103, so if we could turn there, just a few pages previous. [00:16:45] Speaker 00: So that point is made right in the middle of the page, right after that big block quote from the shared specification. [00:16:53] Speaker 00: And they say, because it mentions reducing or preventing signal collisions, that suggests there is no prohibition on signal collisions. [00:17:00] Speaker 00: Prohibition, obviously, is a really strong term. [00:17:02] Speaker 00: That doesn't mean we want signal collisions, but there's no prohibition. [00:17:06] Speaker 00: But I think really important, Your Honor, is if you move down that page to the bottom, the four lines up from the bottom, furthermore, so this is a separate point being made by the board, the above-quoted passage from the 060 patent does not provide any particular details on how to control timing to reduce or prevent collisions, suggesting that this was within the knowledge and skill of a person of ordinary skill in the art. [00:17:31] Speaker 00: They don't contest that in their briefs. [00:17:33] Speaker 00: The board found, based on their own shared specification, [00:17:36] Speaker 00: which just gave you a black box and said, you know how to do it, prevent collisions. [00:17:41] Speaker 00: The board found that suggests the person with skill in the art did know how to do it. [00:17:44] Speaker 02: But isn't there something odd about that? [00:17:46] Speaker 02: Kim, I think, makes clear that we don't want any collisions. [00:17:52] Speaker 02: We want to prohibit them. [00:17:54] Speaker 02: And yet the board is saying you would use Kim in a combination that might lead to some collisions [00:18:02] Speaker 02: like the challenge patent. [00:18:05] Speaker 02: It seems contrary to what Kim teaches, at least. [00:18:07] Speaker 00: Well, Kim teaches in paragraph 44, you should avoid collision by using read-select signals, which is exactly the combination Samson put forward. [00:18:16] Speaker 00: And the board said, that's the combination here. [00:18:19] Speaker 00: And that was the big problem with the expert testimony from Netlist, because it wasn't addressing the actual combination. [00:18:26] Speaker 00: The expert testimony from Netlist was, oh, you've only got these two read-select signals for four different dies. [00:18:33] Speaker 00: No good. [00:18:33] Speaker 00: The problem was that's not what Samsung was proposing. [00:18:37] Speaker 00: The board found that very clearly as a factual matter. [00:18:40] Speaker 00: And I'll note that the teachings of Kim, the teachings of the shared specification, these are all factual issues for the board. [00:18:46] Speaker 00: The board had support to say, yes, Kim doesn't want the collisions, but Kim says read select signals are the way to avoid them. [00:18:54] Speaker 00: And that's what the combination achieves. [00:18:58] Speaker 00: That was supported also by expert testimony, which the board cites too, which is the Wolf testimony on page 108 and 109. [00:19:06] Speaker 00: So we have a lot of evidence backing up the fact that the read-select signals, which are the actual combination proposed by Samsung, they're taught by Rajan. [00:19:16] Speaker 00: Rajan has its own analysis using timing to avoid collisions. [00:19:22] Speaker 00: And that is in appendix 3663. [00:19:26] Speaker 00: which specifically references that those separate chip select signals can avoid collisions. [00:19:31] Speaker 00: And just for the record, that's at column six, lines 34 to 38. [00:19:36] Speaker 00: So all of that supports the idea that exactly as Kim says, if you use read select signals, you avoid the collisions. [00:19:44] Speaker 00: The one other thing I'll note is I heard from my colleague that there's this whole issue about IO and that Kim is all about one IO. [00:19:52] Speaker 00: Well, there are a couple of things that you need to look at in Kim to realize that's not the case. [00:19:56] Speaker 00: Kim is not teaching one IO circuit as a essential of the invention. [00:20:03] Speaker 00: First of all, if you look at figure two of Kim, and this is on page 3624 of the record, you'll see there that there is a shared data inputs output section. [00:20:21] Speaker 00: But then there is an input output driving section for each of the ranks. [00:20:26] Speaker 00: Now that is different than figure five, which is three pages later. [00:20:30] Speaker 00: If you look at appendix page 3627, the appendix five identifies 1,000 as the input output. [00:20:37] Speaker 00: Kim taught both options, right? [00:20:39] Speaker 00: It taught you could do it either way. [00:20:41] Speaker 00: And the language of the Kim specification backs that up. [00:20:45] Speaker 00: So if you turn a few pages later to 3631, [00:20:49] Speaker 00: and look at paragraph 48. [00:20:53] Speaker 00: So that's about two-thirds of the way down. [00:20:58] Speaker 00: You'll see that Kim specifically talks about the fact that it is sufficient for a shared data input-output section to be disposed in at least one of the main chip and the first and second die chips. [00:21:12] Speaker 00: So Kim was very open about, you can put the IO in various spots. [00:21:16] Speaker 00: There is no essential teaching of just one IO. [00:21:19] Speaker 00: It's just not here. [00:21:21] Speaker 00: These were factual matters that the board looked at and decided against Netlist. [00:21:24] Speaker 00: They really, on our standard of review, they shouldn't be subject to reweighing of the facts under this court's law. [00:21:32] Speaker 00: I would like to turn to Claim 7, which my colleague also addressed. [00:21:37] Speaker 00: First of all, [00:21:38] Speaker 00: I mean, I think the panel indicated this. [00:21:41] Speaker 00: There was no request for claim construction here. [00:21:43] Speaker 00: And if you look at the board's discussion of claim seven, the board doesn't do any claim construction. [00:21:49] Speaker 00: It doesn't cite the specification. [00:21:51] Speaker 00: It doesn't cite dictionaries. [00:21:53] Speaker 00: It only uses the language of the claim. [00:21:56] Speaker 00: So this is not a claim construction issue. [00:21:58] Speaker 00: It's a substantial evidence issue as to the factual conclusions of the board in applying the claims themselves. [00:22:05] Speaker 00: Secondly, we need to understand that reducing to zero [00:22:08] Speaker 00: is reducing. [00:22:10] Speaker 00: We keep hearing that, oh, well, what we have taught in REHO is how to get to equal loads. [00:22:17] Speaker 00: Well, that is reducing the difference, because it's reducing the difference to zero. [00:22:22] Speaker 00: The real problem with what my colleague is talking about is he's not addressing the combination. [00:22:27] Speaker 00: The combination, remember, is Kim and Rajan, and then adding REHO. [00:22:34] Speaker 00: Kim and Rajan show different length TSVs. [00:22:37] Speaker 00: Right? [00:22:37] Speaker 00: So you would have this difference in load. [00:22:40] Speaker 00: Reho teaches, okay, take that and instead have this reduction to zero. [00:22:46] Speaker 00: That's a reduction. [00:22:48] Speaker 00: And that was a reasonable interpretation, a reasonable application. [00:22:51] Speaker 00: There was no claim construction. [00:22:54] Speaker 00: So all we're looking at is whether the board reasonably interpreted and adapted the language. [00:22:59] Speaker 00: Now, my colleagues briefs point to, well, the board says requires, and the board says scope. [00:23:05] Speaker 00: But if you look at each of the paragraphs they're citing, all the board does is say quotes the claim language. [00:23:10] Speaker 00: There is no construction. [00:23:12] Speaker 00: And frankly, even on appeal, there's no construction really put forward by the other side. [00:23:16] Speaker 00: And that's because they didn't have one below. [00:23:19] Speaker 00: Finally, let's turn to claim 15. [00:23:22] Speaker 00: And of course, if the panel has any questions on that one. [00:23:25] Speaker 00: Claim 15, this is the question of the separate drivers. [00:23:29] Speaker 00: The board pointed out how Wyman taught the option. [00:23:34] Speaker 00: Again, this is kind of like the Kim argument they're making, right? [00:23:37] Speaker 00: They cherry pick one part of a prior art reference and say, oh, it's only about this. [00:23:44] Speaker 00: But the board found, reasonably looked at Wyman and found on appendix page 134, if we could turn to 134. [00:23:54] Speaker 00: So this is the back of the 060 final written decision [00:23:59] Speaker 00: getting to one of these later claims. [00:24:06] Speaker 00: They spend several pages looking at Wyman, and they point out on page 134 about seven lines down, Wyman expressly discloses that a suitable option involves disabling unneeded transistors and not using taps. [00:24:22] Speaker 00: If you remember what they focused on were the taps with the different levels of driving. [00:24:26] Speaker 00: Not using taps to provide selectable drive levels. [00:24:30] Speaker 00: Furthermore, we find that Wyman's disclosure of providing different drive strengths with different numbers of transistors would result in drivers of different physical size, consistent with the disclosure of the 060 patent. [00:24:43] Speaker 00: So going to the next paragraph, based on the foregoing, we find that Wyman teaches using different amounts of signal drive for different lengths of TSVs [00:24:52] Speaker 00: and that a person of ordinary skill in the art would have recognized that different physically separate drivers can be used to provide those different amounts of signal traffic. [00:25:01] Speaker 00: The board looked at Wyman as a whole. [00:25:03] Speaker 00: It didn't cherry pick particular phrases and came to a finding based on substantial evidence. [00:25:09] Speaker 00: If there are no further questions, I'll give up the rest. [00:25:12] Speaker 00: I will surrender the rest of my time. [00:25:14] Speaker 02: I just wonder if you would take a few seconds. [00:25:16] Speaker 02: Of course. [00:25:16] Speaker 02: We spent all morning with you nice people. [00:25:19] Speaker 02: Could you help me make sure I understand how these four cases relate together? [00:25:23] Speaker 02: The three IPR appeals cover every claim that was at issue in the trials? [00:25:30] Speaker 00: Yeah, that's my easy answer, right, Your Honor? [00:25:32] Speaker 00: Just affirm the IPR appeals and it goes away. [00:25:34] Speaker 00: Yes, they cover every claim. [00:25:36] Speaker 00: That's one option. [00:25:37] Speaker 00: I understand. [00:25:38] Speaker 00: And we all knew that, but go ahead. [00:25:42] Speaker 02: But I guess, alternatively, if we vacate and remand or reverse in any part of any of the three IPR appeals, then I guess we need to dig into the merits of your appeal from the trial. [00:25:58] Speaker 02: Is that right? [00:25:59] Speaker 00: With regard to particular patents, that would be true, Your Honor, in the sense that the verdict form had [00:26:06] Speaker 00: separate determinations as to patents, but not separate determinations as to claims. [00:26:13] Speaker 00: So for example, my colleague is challenging some dependent claims. [00:26:18] Speaker 00: So if only a dependent claim were found to need to go back to the Patent Office, that would vacate and remand the jury verdict, because there isn't a finding by the jury as to dependent claims. [00:26:29] Speaker 00: So with that tweak, Your Honor, I would agree. [00:26:32] Speaker 01: OK, thank you. [00:26:33] Speaker 01: Of course. [00:26:34] Speaker 01: Was there a discussion of a stay or anything like that at the district court? [00:26:39] Speaker 00: I would actually agree. [00:26:40] Speaker 00: Defer to my colleague, Mr. Cordell, on that. [00:26:43] Speaker 00: I was not at the district court. [00:26:45] Speaker 01: OK. [00:26:45] Speaker 01: It's all right. [00:26:46] Speaker 01: We can find that out. [00:26:47] Speaker 00: OK. [00:26:48] Speaker 00: Any other questions? [00:26:49] Speaker 00: No. [00:26:50] Speaker 00: OK. [00:26:51] Speaker 00: Thank you. [00:26:58] Speaker 01: You have two minutes. [00:26:58] Speaker 03: I suspect you're not going to give me that long, are you? [00:27:02] Speaker 01: You have three minutes, so rebuttal. [00:27:04] Speaker 03: Thank you so much. [00:27:07] Speaker 03: So I'd like to begin where my colleague began, which is with paragraph 44, and that's on page 3631. [00:27:14] Speaker 03: And the argument is, oh, the board found, and Kim discloses that your answer is a chip selection switch. [00:27:21] Speaker 03: Chip selection switch. [00:27:23] Speaker 03: But that's not what paragraph 44 is talking about. [00:27:26] Speaker 03: It says that it has a rank selection unit, 1,100, [00:27:31] Speaker 03: which responds to the chip selection switch. [00:27:34] Speaker 03: So we're talking about having a switch between these data lines where it's only connected to one at a time, right? [00:27:41] Speaker 03: Rank selection switch. [00:27:43] Speaker 03: And if you look down further on that page, it talks about, you know, if you get CS0, which is a chip switch, [00:27:50] Speaker 03: The right selecting unit of the rank selecting unit is connected to the first data input output, so it's only connected to one. [00:27:55] Speaker 03: And the bottom, it says if you get the other one, it's connected to the other. [00:27:58] Speaker 03: So it's connecting only to one line at a time with this rank selection switch. [00:28:02] Speaker 03: That is the precise opposite of our invention. [00:28:05] Speaker 03: Our invention requires you to be electrically connected to a group of array dies. [00:28:11] Speaker 03: One group. [00:28:12] Speaker 03: not just one, but a group. [00:28:14] Speaker 03: And so this is teaching the exact opposite of what our invention is. [00:28:17] Speaker 03: Select one array die, one TSV to be connected with through a rank selecting switch, because otherwise you're going to have these data collisions. [00:28:27] Speaker 03: The chip selections, and then the other problem with the chip selection switch argument is [00:28:31] Speaker 03: the board never finds that chip selection switches are actually a solution. [00:28:34] Speaker 03: It never says you can solve this with chip selection switch. [00:28:38] Speaker 03: And it would be odd to do that because Kim didn't think that chip selection switches alone would do it. [00:28:42] Speaker 03: It needed this rank selection switch that is the exact opposite of our invention. [00:28:48] Speaker 03: The board did fault us and say you only looked at two chip selection switches and not four, but it never says you can do it with four. [00:28:55] Speaker 03: So turning then to the fact that [00:29:00] Speaker 03: In our invention, we don't say, oh, you need this special selection switch. [00:29:05] Speaker 03: We don't worry about data collisions. [00:29:08] Speaker 03: There's a reason for that. [00:29:09] Speaker 03: Our invention uses the typical setup where every array die has its own IO. [00:29:14] Speaker 03: So you have your own timing solution built in. [00:29:18] Speaker 03: Raja. [00:29:19] Speaker 03: has its own IO for each of the memory circuits. [00:29:24] Speaker 03: That's built in to solve the problem. [00:29:26] Speaker 03: You get the problem if you start with Kim, or you try to incorporate Kim, where you've taken out the IO. [00:29:33] Speaker 03: and you've shared it, then you're gonna have data collisions. [00:29:35] Speaker 03: And the only way to combine Kim with Rajan is to take all of Kim away, the shared IO circuit, take away its solution, the rank selecting switch, and to then combine it with something else. [00:29:46] Speaker 03: And that's simply not the way it works. [00:29:48] Speaker 03: And then finally, if you were to turn to the picture that my colleague points to, figure two on 3624, that is, if you wanna see if there's actually only one IO, you can tell. [00:29:57] Speaker 03: It says, shared data input output section. [00:30:01] Speaker 03: That is your IO circuit. [00:30:02] Speaker 03: The driving section, that's the driver that puts power in. [00:30:06] Speaker 03: It's not a shared IO circuit that controls timing. [00:30:09] Speaker 03: Finally, I have four seconds left. [00:30:12] Speaker 03: What's missing from Rio is selecting the array die, the number of array dies, in order to reduce the difference. [00:30:17] Speaker 03: There's no selecting of array dies. [00:30:18] Speaker 03: Thank you, Your Honor. [00:30:19] Speaker 03: I appreciate your time and patience.