[00:00:00] Speaker 04:
Our next case is Google versus Singular Computing LLC.

[00:00:05] Speaker 04:
Councillor Speed, I see you've reserved five minutes in front of you both.

[00:00:12] Speaker 04:
Just a second, let's go.

[00:00:13] Speaker 03:
Yeah, definitely.

[00:00:14] Speaker 04:
Okay, you may proceed.

[00:00:17] Speaker 03:
May it please the court, Nathan Speed on behalf of Google.

[00:00:20] Speaker 03:
Singular's patents describe a computer that includes what the patents call Low Precision High Dynamic Range Execution Unit or LPHDR execution unit.

[00:00:28] Speaker 03:
The board found unpatentable, singular claims to a computer that had one LPHDR execution unit.

[00:00:38] Speaker 03:
The only remaining issue is whether or not the board correctly found that Google did not prove it would have been obvious to have a computer with more than one LPHDR execution unit.

[00:00:47] Speaker 03:
On that issue, the board committed legal error.

[00:00:51] Speaker 03:
Before the board, our argument was that Opposer would start with what was called the Doxa reference.

[00:00:57] Speaker 03:
Doxa includes an execution unit called an FPP.

[00:01:00] Speaker 03:
There's no dispute at this point that Doxa's FPP is an LP-HDR execution unit.

[00:01:05] Speaker 03:
The advantage that Doxa describes for its FPP is that it allows for lower power consumption, which is ideal for laptops because they run on batteries.

[00:01:13] Speaker 03:
We then return to McMillan.

[00:01:15] Speaker 03:
McMillan is a reference that describes how to bring supercomputing power to laptops, just like Doxa describes.

[00:01:21] Speaker 03:
And it describes a parallel processing architecture that has many processing elements.

[00:01:25] Speaker 03:
Each of the processing elements includes a floating-point accelerator.

[00:01:29] Speaker 03:
Our argument was that Oppoza, looking at both references, would see how they could be fit together, and they would take Doxa's FPP and use that as the floating-point accelerator in McMillan's device.

[00:01:38] Speaker 03:
The resulting device would be a computer with many LPHDR execution units, rendering the claim obvious.

[00:01:43] Speaker 03:
There was no argument below that the POSIT could not have achieved this invention, and there was no dispute below that the resulting invention would have achieved both of those benefits.

[00:01:52] Speaker 03:
A high-performance laptop that consumed less power.

[00:01:56] Speaker 03:
The board rejected it solely on the finding that the combination would have cost more money to manufacture than the McMillan laptop alone.

[00:02:04] Speaker 03:
Respectfully, that was legal error.

[00:02:06] Speaker 01:
I didn't read the board as saying nothing but cost.

[00:02:12] Speaker 01:
I thought there was this mix of things related to cost, one of which was complexity and maybe an occasional reference to science.

[00:02:22] Speaker 03:
So the board did try to align this decision with the case law, because the case law is pretty clear that economic costs stand alone do not negate an otherwise technologically motivated combination.

[00:02:32] Speaker 03:
What the board did was point to the various discussion in McMillan that said that

[00:02:37] Speaker 03:
We want to avoid complexity in size, not for technological reasons, but because of the economic cost that complexity in size would create.

[00:02:45] Speaker 03:
Macmillan, column one, starts to introduce the invention, and this entire invention is about the fact that supercomputers at that time, this is the 1990s when Macmillan was written, they were really good at performing a lot of calculations.

[00:02:57] Speaker 00:
But what if we were to conclude that the board was reasonable in its reading of the Macmillan reference?

[00:03:06] Speaker 00:
to be that Macmillan is trying to look for the most simple stripped down design possible to accomplish supercomputing inside of a personal laptop computer.

[00:03:22] Speaker 00:
And so its objective, yes, is to keep it as low cost as possible, but

[00:03:29] Speaker 00:
blended with that is this strong objective to use as many simple elements as you can in order to accomplish this supercomputing goal.

[00:03:43] Speaker 00:
I guess the question is, if we were to say that that's a fair reading or a reasonable reading of the reference, then would you not agree that we would have to affirm

[00:03:57] Speaker 03:
I would disagree.

[00:03:58] Speaker 03:
In addition to the case law on economic costs not providing a motivation, undercutting a motivation divide, this court has also had a series of cases where you've made clear that the primary purpose of the prior art reference does not control.

[00:04:11] Speaker 03:
Prior art references are good for everything they disclose.

[00:04:13] Speaker 03:
So in this case, McMillan, whether or not Mr. McMillan wants to keep this laptop as cheap as possible, and I'll talk about that in a second,

[00:04:20] Speaker 03:
That's his invention, and that's what he disclosed to the public.

[00:04:23] Speaker 03:
But he also, in doing that, described a laptop that had parallel processing and numerous processing elements, each of which has a floating-point processor.

[00:04:32] Speaker 03:
And therefore, Oppoza looking at that and not concerned as much about cost as Macmillan might have been, but concerned more about power consumption and high performance computing.

[00:04:39] Speaker 00:
But we have case law that says we're not going to adopt a proposed combination if that combination would

[00:04:47] Speaker 00:
Defeat of references objective like the reference you are about to modify if that reference is essentially saying this I have a core objective here and then that core objective would be somehow erased by the combination then our case law says and that's a fair basis to reject the proposed combination

[00:05:07] Speaker 03:
So those cases, every single one of the boards cited, and that's a good sign, it's great.

[00:05:11] Speaker 03:
They've all related to technological aspects of the computer, or of the prior reference.

[00:05:16] Speaker 03:
So it's removing a technological feature of the reference.

[00:05:19] Speaker 03:
So if someone wanted to take McMillan to remove its parallel processing architecture, that would be contrary to its inventive purpose, which is adding a parallel processing architecture to it.

[00:05:29] Speaker 00:
But what if it's also true or fair to say that Macmillan was looking for the most elemental simplified architecture possible to do what it wants to do?

[00:05:42] Speaker 03:
So that's where we would have to engage on the facts of McMillan, and I would argue that substantial evidence doesn't support that reading of McMillan.

[00:05:48] Speaker 03:
McMillan, it starts with a conventional laptop, and it's adding a supercomputer architecture to it.

[00:05:54] Speaker 03:
It is indiscriminately making that laptop more complex and more costly than it was before McMillan came up with his invention.

[00:06:01] Speaker 00:
At column eight of McMillan, he says that... A complicated machine, to be sure, but the goal is the least complicated version of that complicated machine.

[00:06:11] Speaker 03:
But he does talk about allowing scalable performance at various price points.

[00:06:16] Speaker 03:
So he's not saying, this version that I have here is the only version I have.

[00:06:19] Speaker 03:
And indeed, the embodiment we relied on had 256 processing elements, but he talks about having thousands of these processing elements.

[00:06:26] Speaker 03:
So he's open to having a device that's much more complex than the embodiment, the laptop embodiment that's disclosed in the specification.

[00:06:34] Speaker 03:
But he clearly is, when he talks about cost,

[00:06:37] Speaker 03:
He says, you can have high-performance computing at moderate cost.

[00:06:40] Speaker 03:
And so the question is, in context, what type of cost is he talking about?

[00:06:44] Speaker 03:
And he's saying it's an additional cost that doesn't approach the exorbitant crisis of then-existing supercomputers, which is 100.

[00:06:50] Speaker 03:
He literally, in column one, says $100,000.

[00:06:52] Speaker 03:
So that's the cost that he's concerned about.

[00:06:54] Speaker 03:
All of his discussion of complexity and size, it all goes back to his overall purpose here, which is to bring supercomputer power to the everyday person so that they can have it in their laptops.

[00:07:05] Speaker 03:
It's going to be more expensive, there's going to be a moderate cost to this laptop, but he still saw that as a desirable feature.

[00:07:11] Speaker 03:
And I think Oppozo, desiring that feature and then wanting to have their laptop last a little bit longer on their plane ride to California, what have you, would look to Docter's FPP and see that that is providing a solution to a potential problem there.

[00:07:26] Speaker 03:
Lest there's more questions on the cost issue, if your honors were to reverse the board on that issue, remand would be necessary, except for this other ground that we have, which is the customization ground.

[00:07:38] Speaker 03:
On the customization ground, it started from the same basic premise as the first one, where we have Dr. McMillan.

[00:07:44] Speaker 03:
But in this one, instead of focusing on McMillan's laptop, we're focusing on another disclosure of McMillan, which is that you can use his invention in an embedded computer that supports application-specific processing, like signal processing.

[00:07:55] Speaker 03:
We then pointed to this reference called Tong, which is a research paper.

[00:07:58] Speaker 03:
And the whole purpose of Tong was to test a bunch of benchmark signal processing applications to figure out, do they really need full precision, or can they operate at lower precision?

[00:08:08] Speaker 03:
And he concluded that five of those applications can work with varying degrees of low precision.

[00:08:12] Speaker 03:
So he said, or disclosed to the public, that these five applications don't need the overhead of full precision.

[00:08:21] Speaker 03:
The person that's reading all these references together would see,

[00:08:25] Speaker 03:
that you could piece them together to create a better computer that's application specific to support the signal processing applications that Tom discloses.

[00:08:34] Speaker 03:
And when you're doing that, you would just customize Doxer's FPT because at that point you're never going to use its circuitry for full precision.

[00:08:40] Speaker 03:
You're going to only use between 11 and 5 minutes of precision.

[00:08:44] Speaker 03:
And at that point, if you know you're only going to use that, there's no need to create the full Doxer because it's going to have circuitry that's just never going to be used.

[00:08:50] Speaker 03:
The board rejected that by finding that full precision was fundamental to both Doxer and Tong, and respectfully the record just doesn't support that.

[00:08:59] Speaker 03:
The paragraph that they cited in Doxer, paragraph three,

[00:09:03] Speaker 03:
It says, while some applications may require full precision, other applications may not.

[00:09:09] Speaker 03:
This is paragraph three of the one that the board relied on.

[00:09:11] Speaker 03:
It admittedly says that for a general purpose processor, a greater precision may be needed, but our combination was not a greater, a general purpose processor, but an application specific processor designed to support

[00:09:22] Speaker 00:
The sentence you're just quoting about, some need greater precision, others don't.

[00:09:27] Speaker 00:
But then the rest of that sentence says, and that's why we want selectable precision.

[00:09:32] Speaker 00:
That's why we want the ability to have the flexibility to fine tune the precision for any given application that's operating.

[00:09:41] Speaker 03:
When it's used in a general purpose processor that you know you might need full precision, in the embedded context where you know you're never going to use that full precision.

[00:09:49] Speaker 03:
So if you're reading Pong,

[00:09:51] Speaker 00:
Isn't the whole point of DOCSIS to have selectable precision?

[00:09:55] Speaker 03:
And our combination maintains selectivity if you just were selecting between the various low-level precision that Tom taught that its five signal processing applications could be used for.

[00:10:04] Speaker 03:
I had ones that worked well at 11, but others, Alvin and Spinks, I believe they were called, worked well at as low as five.

[00:10:12] Speaker 03:
So you'd be able to switch between those varying levels depending on what signal processing application you were supporting at that time.

[00:10:20] Speaker 03:
And important, I think it's also important in Doxert's paragraph 3, it also says, if it is known that an application always requires a certain reduced precision, the floating point processor can be designed and built to that reduced precision.

[00:10:32] Speaker 03:
That's what our combination was.

[00:10:34] Speaker 03:
It is an embodiment in which you know that you're always going to have reduced precision, and so Doxert's paragraph 3 is telling you that you can

[00:10:41] Speaker 03:
design a processor specific to that and we would just maintain the selectivity of Docter so that it could switch between the different signal processing applications that it's supporting this embedded application specific and body.

[00:10:53] Speaker 03:
As for Tong, the board's decision on Tong

[00:10:58] Speaker 03:
itself at page 81 explicitly identifies that Tong says not all programs need the precision provided by generic floating point hardware, the fine precision of the 23-bit mantissa is not essential, and a single custom floating point format may be a viable option for particular applications.

[00:11:14] Speaker 03:
Tongue, in no uncertain terms, is saying that full precision is not fundamental.

[00:11:19] Speaker 03:
It's suitable for certain applications.

[00:11:21] Speaker 03:
And so with that teaching, having looked at DoxR, I suppose you would understand it's better to customize this chip, reduce circuit space, because otherwise I'm going to have circuitry that's just never going to be used.

[00:11:34] Speaker 04:
Kelsary, you're into rebuttal time, do you want to?

[00:11:37] Speaker 03:
I'm happy to answer any further questions.

[00:11:40] Speaker 03:
Okay.

[00:11:47] Speaker 04:
Mr. Lambert-Nakos.

[00:11:54] Speaker 02:
Did I pronounce your name correctly?

[00:11:56] Speaker 02:
Lambert-Anakos, Your Honor.

[00:11:58] Speaker 02:
Lambert-Anakos.

[00:12:02] Speaker 04:
Thank you, Your Honor.

[00:12:10] Speaker 02:
Good morning, Your Honors, and may it please the Court

[00:12:13] Speaker 02:
The board did not err in finding that the teachings of Macmillan, which discourage additional size and complexity in the processing elements of Macmillan for the purpose of minimizing costs, were barred as a matter of law by the Farenkopf case and other cases.

[00:12:34] Speaker 02:
The Frarenkopf case specifically holds that a businessman's concerns over the cost of making a combination are not relevant.

[00:12:45] Speaker 02:
Here, the board did not rely on a businessman's concerns at all.

[00:12:48] Speaker 02:
The board relied on specific teachings in McMillan that in order to make a low-cost computer with supercomputing power, one which could be used by consumers,

[00:12:59] Speaker 02:
that the processing elements should be simplified.

[00:13:03] Speaker 02:
They should not have additional registers, they should not have additional output buffers, additional pins, because those features increase die size, increase packaging size, and result in increased cost.

[00:13:17] Speaker 02:
And so the board found that a combination with DoxR would violate those teachings.

[00:13:23] Speaker 02:
DOXR is a processor that actually contains not only the processor which performs the mathematical operations, but also an additional control.

[00:13:34] Speaker 02:
That additional controller goes with each DOXR processor, and its job is to adjust the precision of the DOXR processor.

[00:13:43] Speaker 02:
And so a combination which would replace each processing element in Macmillan with essentially two processors, the Doxer arithmetic unit and the controller, would clearly violate what Macmillan was trying to accomplish, which was a simplified array of processing elements.

[00:14:01] Speaker 02:
There was nothing in the law.

[00:14:03] Speaker 01:
Yes, Your Honor.

[00:14:04] Speaker 01:
Can you step back for a minute from your description

[00:14:14] Speaker 01:
the first principles are the right way of thinking about the role of cost.

[00:14:29] Speaker 01:
Because I think at the heart of what you're just articulating is that cost, if it is

[00:14:37] Speaker 01:
called out in one of the references as important should be treated differently from cost if it is not called out in one of the references.

[00:14:47] Speaker 01:
Tell me how we're supposed to think about why that should be so, or as I think the other side says, not so.

[00:14:58] Speaker 02:
An invention that has as a goal, creating a device which reduces cost, does not fall within the Farenkopf line of case law, where it is the inventive goal, at least where it is the inventive goal of the patent, and the patent teaches technology

[00:15:19] Speaker 02:
that should be chosen in a way to meet that goal.

[00:15:23] Speaker 02:
And what McMillan does is he discusses particular types of components, such as registers, such as output buffers, as examples of technology which you would not include

[00:15:37] Speaker 02:
in the McMillan device because it would hinder reaching his goal of providing a low-cost computer.

[00:15:47] Speaker 02:
So he has technology teachings which he advocates for in order to reach that goal.

[00:15:53] Speaker 01:
I think you're just redescribing what I want to take as an accurate description of what McMillan says.

[00:16:02] Speaker 01:
The question is, as a matter of first principles, why should that matter?

[00:16:07] Speaker 01:
in light of things like I think you know Mr. Speed referred to references teach whatever they teach even components of what's in a references

[00:16:26] Speaker 01:
ultimate goal of creating something that combines things.

[00:16:30] Speaker 01:
It may be obvious to take one piece that it teaches and shed others and thereby not reach the reference's goal.

[00:16:39] Speaker 01:
I'm trying to understand how to think about why this accurate description you're giving of Macmillan, I'm assuming, should matter for the bottom line conclusion of whether there's a motivation to combine.

[00:16:54] Speaker 01:
Because you're not combining the whole thing, right?

[00:16:57] Speaker 02:
Yes.

[00:16:57] Speaker 02:
The ultimate goal matters because under the case law, under the Kimura's line of case law, a combination which would eviscerate or would force you or would result in your failing to meet the inventive goal is not obvious.

[00:17:16] Speaker 02:
So if the inventive goal of Macmillan is a low-cost computer and your combination would prevent you from reaching that goal because you violated technological teachings in the reference, then that combination is not obvious.

[00:17:33] Speaker 02:
The same applies for Doxer.

[00:17:36] Speaker 02:
Doxer's inventive concept is to create a general purpose process.

[00:17:41] Speaker 02:
the general purpose processor can execute all the applications that are presented to the processor where it's installed.

[00:17:50] Speaker 02:
Everyone agrees that Doxer has full precision support.

[00:17:54] Speaker 02:
But Doxer has full precision support because his goal was to present something which is general purpose.

[00:18:01] Speaker 02:
He acknowledges that there exists special purpose or low precision processors in the art.

[00:18:07] Speaker 02:
But that's not what he's doing.

[00:18:09] Speaker 02:
He's presenting a general purpose processor.

[00:18:12] Speaker 02:
And so a combination, such as in ground 4A, which would remove that full precision support, would eviscerate Doxer's inventive goal.

[00:18:24] Speaker 01:
So on that one, I think as I understood Mr. Speed's point, was that

[00:18:34] Speaker 01:
grant all of that, but the claims here require...

[00:18:44] Speaker 01:
a processor that does certain things and there are such things which the market is by assumption interested in.

[00:18:53] Speaker 01:
Special purpose processors and why wouldn't a relevant skilled artist and thinking about making a special purpose processor for one of the purposes of

[00:19:05] Speaker 01:
for which precision just isn't needed, not needed.

[00:19:10] Speaker 01:
Make this combination of massively parallel, imprecise owners.

[00:19:22] Speaker 02:
In addition to the size arguments, which apply also to Mr. Speed's argument, a person of ordinary skill in the art, under the case law, does not go to a reference

[00:19:33] Speaker 02:
and modify that reference in a way that would remove the inventive concept.

[00:19:40] Speaker 02:
And so the reason why a person of ordinary skill in the art wouldn't look to a general purpose processor to create a special purpose processor is because the law, in an attempt to remove hindsight combination,

[00:19:53] Speaker 02:
have certain guardrails on how to modify a reference.

[00:19:57] Speaker 02:
And under the Comores line of case law, if the reference is teaching a person of ordinary skill, this is the CPU of a device.

[00:20:06] Speaker 02:
It is to execute all of the applications that are running on this device, a general device, such as a laptop, a PDA, et cetera.

[00:20:18] Speaker 02:
You would not take that reference and strip away

[00:20:22] Speaker 02:
what makes it the invention in the doctor's mind, and it's disclosed in the publication itself, to put it into your combination in order to invalidate a claim.

[00:20:33] Speaker 02:
So it's the line of case law, such as the Comoros case,

[00:20:40] Speaker 01:
petitions I guess here was this idea that the claims cover processors that are not just general-purpose processors and

[00:20:59] Speaker 01:
purpose processors essentially provide the motivation for this, what the Doxer, McMillan, Tong combination.

[00:21:09] Speaker 01:
Is that adequately presented in the petition?

[00:21:15] Speaker 02:
Well, the first part of what Your Honor said about the claims presenting a special purpose computer.

[00:21:22] Speaker 01:
No, no, they don't require it.

[00:21:23] Speaker 01:
They're completely indifferent about whether

[00:21:25] Speaker 01:
whether the processors that are in the claims are for general purpose use or special purpose use, which means if you could anticipate it or make it obvious by its special purpose, it would come with it.

[00:21:39] Speaker 01:
So is that an adequately preserved argument?

[00:21:43] Speaker 01:
I don't mean to be asking us.

[00:21:47] Speaker 01:
difficult question yeah I'm sorry but I didn't follow did the petition set forth this theory of a special purpose processor that isn't that we've heard about this morning

[00:22:01] Speaker 02:
The petition mentioned in round 4A that the modification would be in order to be able to run certain identified low precision applications.

[00:22:13] Speaker 01:
And that there is a skilled artisan who would be interested in creating such a process.

[00:22:19] Speaker 02:
It says, I believe, that a skilled artisan would be interested in doing so and particularly would remove the full precision capability of Doxxer in order to decrease the cost of the computer by eliminating additional circuitry which would take up space and cost money.

[00:22:41] Speaker 02:
So what's interesting about their argument for 4a is that they're arguing that a reduction in cost because of less circuitry is a motivation to combine.

[00:22:51] Speaker 02:
But they question whether Singular can rely on the fact that Macmillan teaches that you want to reduce cost and reduce size.

[00:23:02] Speaker 02:
And Singular's argument is that combination in ground three and four would increase cost.

[00:23:09] Speaker 02:
In other words,

[00:23:11] Speaker 02:
It's an adequate motivation to combine for Google that cost reduction would take place by removing circuitry.

[00:23:20] Speaker 02:
But it is not an appropriate argument for Singular to say that the Doxer combination would increase costs and therefore suggest to a person of ordinary skill in the art not to make the combination, even though McMillan explicitly contains those teachings.

[00:23:41] Speaker 02:
Now the combinations that are cited by Google, none of them are supported by the references themselves.

[00:23:48] Speaker 02:
McMillan doesn't suggest that his processing elements can be replaced by low precision processing elements.

[00:23:55] Speaker 02:
Doxer doesn't suggest that his processing element should be used in an array or that it should be limited only to low precision.

[00:24:04] Speaker 02:
Tong doesn't suggest that a computer should use only low precision.

[00:24:08] Speaker 02:
In fact, he suggests that his processor would be used in a computer with a full precision processor and a switch to allow selection between them.

[00:24:18] Speaker 02:
And Tang does not suggest using his processor in an array.

[00:24:24] Speaker 02:
So the only evidence that was before the board on these issues was provided by Google's expert.

[00:24:32] Speaker 02:
Whereas Singular, throughout all of its arguments, relied on the teachings in the references themselves to show that a person of ordinary skill in the art would be discouraged from making the combinations.

[00:24:45] Speaker 02:
that Google advocates because of the increased costs associated with DoxR in contravention of Macmillan's teachings as well as DoxR's teaching that his processor is a general purpose processor.

[00:25:02] Speaker 02:
And the board agreed with our reading of those references.

[00:25:05] Speaker 02:
It agreed with our experts' discussion of those references and how they ought to be interpreted.

[00:25:13] Speaker 02:
And therefore, the board had substantial evidence to find that Google had not met its burden of truth.

[00:25:20] Speaker 02:
The board weighed the benefits that Google cites.

[00:25:23] Speaker 02:
The board also weighed the evidence that we put in front of it.

[00:25:26] Speaker 02:
and decided that for the claims that require a massively parallel group of low precision, high dynamic range processors, that that invention was not obvious.

[00:25:51] Speaker 02:
Now, should the court determine that there was an error below and that the decision should be vacated, the board below expressly limited its findings on secondary considerations to claims with only one LPHDR execution unit.

[00:26:12] Speaker 02:
The board did not address, explicitly did not address the claims to massively parallel arrays of LPHDR execution units.

[00:26:20] Speaker 02:
We presented evidence of industry skepticism.

[00:26:23] Speaker 02:
We presented evidence of industry praise.

[00:26:26] Speaker 02:
And that evidence was specifically focused on the arrays.

[00:26:32] Speaker 02:
And so that evidence has not been considered yet by the board.

[00:26:36] Speaker 02:
And so even in vacating the board's decision, we should have the opportunity to present that evidence and perhaps even obtain further evidence as

[00:26:46] Speaker 02:
discovery is ongoing in a district court case and there may be evidence that would be relevant to the board's finding on secondary considerations.

[00:26:56] Speaker 02:
It's also important to note for the what are called the exceeds claims under ground three and four where the full precision capability is not excised from doxer in the combination.

[00:27:09] Speaker 02:
that the board did not consider the issue that the claims require that in a massively parallel array that there be a certain number of low precision, high dynamic range execution units in the array that exceed the number of units that are capable of doing full precision 32-bit multiplication.

[00:27:30] Speaker 02:
Now that's important because DoxR is capable of doing 32-bit multiplication.

[00:27:36] Speaker 02:
even if you consider it to be low precision, the number of DOCSER units cannot outnumber itself.

[00:27:43] Speaker 02:
Under the claim, Google is relying on an LPHDR execution unit being DOCSER, but then excludes, in a claim construction argument, the ability of DOCSER to satisfy

[00:27:58] Speaker 02:
the claim element that refers to the number of units.

[00:28:02] Speaker 01:
Just so I understand, so this is a point that doesn't need and the board did not need to reach given how it upheld claims on another basis, but would have to be reached if what the board did was set aside.

[00:28:19] Speaker 01:
That's correct.

[00:28:24] Speaker 02:
Would you like to conclude, at this point, your hearing for your other time?

[00:28:30] Speaker 02:
Yes, Your Honor.

[00:28:31] Speaker 02:
Thank you very much.

[00:28:39] Speaker 04:
We have a little over three minutes.

[00:28:40] Speaker 03:
Thank you, Your Honor.

[00:28:42] Speaker 03:
Just briefly on the question of whether or not the petition raised it, I point Your Honor to appendix 720 to 723.

[00:28:47] Speaker 03:
That's where we set forth the customized ground.

[00:28:50] Speaker 03:
It's explicitly set forth there that it was about an embedded system supporting Tong's applications and that's there.

[00:28:57] Speaker 03:
The other just back points I just wanted to point out the idea that this adding registers is the the problem that Macmillan was teaching again Macmillan explicitly its invention adds additional registers to the laptop they called address location registers and that's that column 14 I believe and so Macmillan's actually adding registers, which is the only thing that Singler's Council cited as causing the increase in cost of Macmillan

[00:29:21] Speaker 03:
But going back to the question, is this a case of first principles, or how should we be looking at this?

[00:29:26] Speaker 03:
In Ray Farincock, that prior reference in that case, the primary reference that was being modified, said the use of inhibitors is disadvantageous because it's costly.

[00:29:35] Speaker 03:
And the very combination was using inhibitors in that reference.

[00:29:38] Speaker 03:
And the patent owner said, you can't do that.

[00:29:40] Speaker 03:
This reference says it's costly, so why would you ever use inhibitors in view of this reference?

[00:29:45] Speaker 03:
And the federal circuit, and it's

[00:29:49] Speaker 03:
that a given combination would not be made by a businessman for economic reasons does not mean that persons skilled in the art would not make the combination because of sub-technological incompatibility.

[00:30:00] Speaker 03:
Only the latter fact would be relevant.

[00:30:03] Speaker 03:
Economic costs are legally irrelevant.

[00:30:05] Speaker 03:
The hypothetical person ordinary skill is an ordinary

[00:30:08] Speaker 03:
artisan in the technical field.

[00:30:10] Speaker 03:
They shouldn't be burdened by wondering if this particular invention is going to be too costly to actually be commercially viable in the market.

[00:30:17] Speaker 03:
It's a question that you get a patent if you come up with a technological invention.

[00:30:21] Speaker 03:
There's nothing in Singular's patents I should know about cost savings or that this is a particularly a device that he's found a way to achieve a lower cost silicon or anything like that.

[00:30:32] Speaker 03:
He has an idea that you should have low precision, high dynamic range execution units.

[00:30:36] Speaker 03:
The problem for him is that

[00:30:38] Speaker 03:
Doxer had it before, Tong had it before.

[00:30:41] Speaker 03:
And so the only basis that he has these claims now is on the idea that it would be more expensive to manufacture Macmillan's laptop if he used Doxer's FPPs.

[00:30:50] Speaker 03:
That is contrary to this court's law.

[00:30:52] Speaker 03:
You've got Ingrid Farrincock, most recently in grit energy.

[00:30:55] Speaker 03:
In grit energy, this court vacated and remanded on this exact issue because the board found that the combination would have been more costly.

[00:31:02] Speaker 03:
This court said in that case,

[00:31:04] Speaker 03:
Even if we accept the board's factual determination that swapping the primary references components results in a more expensive system, that determination standing alone is insufficient to reject the technological reasons that the petitioner provided.

[00:31:18] Speaker 03:
That's precisely what happened here.

[00:31:20] Speaker 03:
The board allowed the cost concerns of McMillan to stand alone

[00:31:24] Speaker 03:
and never address our technological basis for the motivation.

[00:31:28] Speaker 03:
There is no dispute that POSIS could have achieved this laptop, that it would have been a high-performance computer, and that it would have saved power.

[00:31:37] Speaker 03:
Unless there's any questions, I'm happy to rest on the briefs.