NVIDIA Files Complaints Against Samsung and Qualcomm for Infringing Its GPU Patents

nVidia appears to be aiming very broadly with some of those claims; any validity will of course come down to the specifics of the implementations of each idea.

They'll need to be careful with some of their targets. Imagination has been at graphics as long as anyone (HTC's stab at Apple/ImgTec using S3's old patents didn't amount to much), and the Adreno architecture originated from work done by ATi and potentially some BitBoys involvement. On a relative scale, Mali, originating from the grad school work I think of the founders of the old Falanx is newish, but the details of some of these architectures/pipelines (especially PowerVR) makes them seem fairly unique.
 
Last edited by a moderator:
So that's NVIDIA's idea of licensing GPU IP, to turn into a Rambus shenanigan? :devilish:
 
Well, I suppose they can't outright buy them like they did 3Dfx.

Now, I haven't read the patents so I look forward to a run-down of what actually is at stake here (and why they should pay), preferably from the likes of Ars Technica.
 
They couldn't force Samsung to lisence their IP either obviously. I hate to admit that Charlie might have another of his annoying "told you so's".....:rolleyes:
 
One assumes Samsungs mobile's defence on the Qualcomm chips is that it is Qualcomm's problem. They don't have that defense with their Samsung semi's own Exynos line.

The biggest two elephants in the room are Apple and Intel.

Apple is the majority user of IMG PowerVr, but they aren't cited, neither is Intel.

Yes, the PR does say about being in discussions with others, but I'd be surprised if Intel, or in particular Apple just rolled over to get their belly tickled.

I assume an alleged patent infringement is by a physical implementation, hence why chips are being cited and not IP companies. However,amongst other interesting things that this complaint brings up, is that indirectly, Nvidia are getting embroiled with ARM, their CPU IP provider that is central to their Tegra platform.
 
Last edited by a moderator:
One assumes Samsung's defence on the Qualcomm chips is that it is Qualcomm's problem. They don't have that defense with their own Exynos line.

The biggest two elephants in the room are Apple and Intel.

Apple is the majority user of IMG PowerVr, but they aren't cited, neither is Intel.

Yes, the PR does say about being in discussions with others, but I'd be surprised if Intel, or in particular Apple just rolled over to get their belly tickled.

I'd say you wait just a little bit more, since I have the feeling that especially Apple is NOT going to get left untouched.
 
OTOH, they're a public company and under US law - AFAIK - they have the obligation to pursue revenue opportunities or they could be sued by their stockholders. Crazy legislation.
 
The full complaint can be found here:

http://www.scribd.com/doc/238684238/NVIDIA-ITC-Complaint-Final

It says Intel has paid $1.5B for the use of Nvidia licenses

"Intel is licensed by NVIDIA under the Asserted Patents and also conducts significant domestic industry activities in the United States relating to products practicing the Asserted Patents. As part of a cross-licensing agreement, Intel agreed to pay NVIDIA a total of $1.5 billion over five years for the use of NVIDIA’s patents and its state-of-the-art GPUtechnology"

Would that be via some all encompassing agreement that Intel/Nvidia struck some time back ?

the Citation also has states there is a confidential part of it which lists the companies that have licensed said patents.
 
Last edited by a moderator:
nVidia appears to be aiming very broadly with some of those claims; any validity will of course come down to the specifics of the implementations of each idea.

They'll need to be careful with some of their targets. Imagination has been at graphics as long as anyone (HTC's stab at Apple/ImgTec using S3's old patents didn't amount to much), and the Adreno architecture originated from work done by ATi and potentially some BitBoys involvement. On a relative scale, Mali, originating from the grad school work I think of the founders of the old Falanx is newish, but the details of some of these architectures/pipelines (especially PowerVR) makes them seem fairly unique.

In the case of qualcomm, one would have to dig up the exact details of the sale of the mobile biz to qualcomm. I cannot off the top of my head if the sale to qualcomm included actual patents or only licenses to patents. And even if the ATI/Nvidia patent cross license dated to that period, that doesn't mean that indemnity from that cross license actually transferred to any of the designs sold to qualcomm.

As far as being careful with their targets, Nvidia are being very careful. In the case of Samsung, it is highly unlikely that Samsung has any significant IP with which to attack Nvidia in the SoC market that isn't already covered by existing x-licensing agreements with other semiconductor fabricators. In the case of Qualcomm, most of Qualcomm's IP is actually tied to compulsory licensing infrastructure (AKA modem IP wrt 3G/4G, yes they do make a lot of money off if, but the licenses are generally compulsory due to their standard inclusion). As far as graphics, it is unlikely that Qualcomm has any significant/relevant IP due to how they entered the market, and the low likelyhood that ATI actually included actual patents vs perpetual rights to use.

It is also unlikely that either of Samsung's other potential suppliers of GPU IP want to voluntarily enter into the court case with Nvidia. It makes little sense for IMG to voluntarily indemnify Samsung in this case, esp, when IMG and Nvidia don't have an existing x-license and may or may not be negotiating one. We have no clue about the existing x-license relationships of other IMG licensees other than Intel with is fully indemnified by their existing x-license with Nvidia. It is unlikely that Apple is currently in any direct danger with of being sued by Nvidia (due to existing biz relationships) and may already have or be in the process of negotiating various license agreements with Nvidia. And Apple probably actually has some relevant IP making the fight more difficult.

Which leaves ARM. It is unknown what if any x-license agreements exist between ARM and Nvidia, but its unlikely based on reported information that any exist wrt graphics based on the suit and PR. However, it is highly likely that Nvidia already has a full license to particular core designs and a full architectural license which means that ARM has little leverage WRT to Nvidia, esp if Nvidia doesn't directly go after ARM per se wrt Mali. AKA, the ARM/Nvidia license agreement does probably have some legal impact if Nvidia was suing ARM directly, but likely has little relavence WRT to Nvidia suing a third party over the same IP (AKA the ARM lawyers probably didn't think of this scenario).

This suit certainly looks like Nvidia did their homework and decided to sue the two companies with both the largest liability and the weakest ability to defend themselves in a patent battle.
 
The full complaint can be found here:

http://www.scribd.com/doc/238684238/NVIDIA-ITC-Complaint-Final

It says Intel has paid $1.5B for the use of Nvidia licenses

"Intel is licensed by NVIDIA under the Asserted Patents and also conducts significant domestic industry activities in the United States relating to products practicing the Asserted Patents. As part of a cross-licensing agreement, Intel agreed to pay NVIDIA a total of $1.5 billion over five years for the use of NVIDIA’s patents and its state-of-the-art GPUtechnology"

Would that be via some all encompassing agreement that Intel/Nvidia struck some time back ?

the Citation also has states there is a confidential part of it which lists the companies that have licensed said patents.

WRT Intel, yes, as part of the settlement of the previous licensing arrangement between Nvidia and Intel, Intel has an x-license WRT to at least the existing portion of the Nvidia IP portfolio at the time of the settlement plus some number of additional years. I don't know if their are any publicly available documents detailing the forward length of the IP licensing timeframe (aka the year when new Nvidia patent filing are no longer covered by the x-licensing agreement). It is probably important to point out, that from my understanding of the Intel/Nvidia licensing agreement(based entirely on publicly available information and action), that it is a fairly one sides affair. Intel gained right to use Nvidia patents but Nvidia did not gain right beyond those included in the original IP license, aka anything past P6 bus based interfaces nor anything x86 related not directly related to the P6 bus based patents. Hence why Intel ended up kicking into the kitty. That means that in the future, if Intel/Nvidia came to patent blows, it is likely that Nvidia would have significant liabilities.


Also just a note on how x-license/patent license agreements generally work... When a general license is entered into, it typically includes all IP to the current data plus a number of additional years (or up to forever for broad general x-license agreements). The IP licensed under the agreement is generally licensed for at least the validity period of the youngest patents or continuations of those patents (or in effect for infinity, as once the youngest patent expires, there is no need to license). That for instance means that when something like the AMD/Intel license agreement expires, Intel's rights to use AMD's 64bit x86 patents don't expire, it simply means that no new filed patents past that date are included within the terms of the agreement. Of course there can and generally are all types of additional riders as part of patent agreements including things like limitations of transferability, area/patent specific future work rights, indemnity limitations/allowances (you can/cannot sub-license specific IP), restrictions of use(AKA can make this product but can't make that product), etc.
 
I think It's pretty simple, Just like Intel spared itself the headache, and paid NVIDIA the licensing fees. Samsung and Qualcomm will most likely do the same, nobody can escape the fact that NVIDIA is the biggest contributer in graphics, somebody somewhere infringed on something they patented for sure.
 
NVIDIA Files Complaints Against Samsung and Qualcomm for Infringing Its GPU P...

There's another problem with your understanding of patents: even if a general concept already exists as prior art, you can still patent specific applications of that idea. Whether or not is a significant development doesn't matter. (I do think it was a major step forward, FWIW.)

Actually it does matter. Patenting something is one thing. Defending that patent in court is a whole different story. There have been recent high profile cases where the US Supreme Court has invalidated patents because they were simply implementations of an "abstract idea" which is inherently not patentable.

That's nVidia's first burden of proof - they have to convince the judge that the idea was inventive. They may have a shot with programmable vertex and pixel shaders since they were instrumental in defining the spec and producing the first hardware. Unified shaders in comparison seems like a weak angle.


It is perfectly possible to patent designs that haven't been implemented yet. For something to be patentable, it must have been reduced to practice. You'd think that this makes flying cars unpatentable? Wrong. The very act of writing down certain implementation details as a patent counts as a "constructive reduction to practice" (as opposed to "actual reduction to practice".)


Yes, absolutely. You can patent a design that has no physical implementation as long as you can adequately describe the innovation on paper. However it's not clear that Qualcomm et al infringed on nVidia's designs or simply had the gall to come up with an alternative design for the same idea.

Google can exist today because Yahoo couldn't patent the idea of an internet search engine and they were free to come up with their own implementation.
 
If Apple/Samsung do roll over, what about the multitude of Mali (and to a lesser extend PowerVR) licensees ? In particular Mediatek, forecast to ship in excess of 300M socs this year, all of them with either Mali or PowerVr.

Although Qualcomm is cited as a respondent, the vast majority of the complaint seems targeted at Samsung, which is strange as Qualcomm socs ends up in many more products than samsung sells in total.

On a general point, do the providers of IP indemnify their licensees from any patent issues ?, still looks like a gamble to me that Nvidia are effectively indirectly citing their Tegra CPU supplier for violating GPU patents.

And again, do provides of actual chips (i.e. qualcomm) indemnify their customers from patent issues ?
 
Last edited by a moderator:
The biggest two elephants in the room are Apple and Intel.

Apple is the majority user of IMG PowerVr, but they aren't cited, neither is Intel.

Yes, the PR does say about being in discussions with others, but I'd be surprised if Intel, or in particular Apple just rolled over to get their belly tickled.

Intel already settled with Nvidia in 2011 and licensed Nvidia's GPU patents for $1.5 billion over six years.

So Intel believes that Nvidia's patents are valid and worth $250 million a year in licensing fees.

So we’ve established what NVIDIA gets, but how about Intel? The Intel situation looks to be much more straightforward. As we mentioned previously, NVIDIA and Intel originally cross-licensed in 2004 so that Intel could build IGPs using NVIDIA patented technologies and methods. That agreement was set to expire this year, which would have been a massive problem for a company whose CPUs almost always include a GPU. Today’s agreement with NVIDIA renews and extends that original agreement: Intel continues to cross-license with NVIDIA, allowing them to produce IGPs that use/infringe on NVIDIA patents.

http://www.anandtech.com/show/4122/intel-settles-with-nvidia-more-money-fewer-problems-no-x86/2
 
Basically Nvidia can sue any company that make SoC (with modern GPU)? So Mediatek, Rockchip, Allwinner, Apple, etc can be sued by Nvidia? Since the patents are relatively broad (at least from what I know, which isn't much), why Nvidia didn't sue them in the beginning? It is like they let them succeed and then try to reap the benefit.
 
Basically Nvidia can sue any company that make SoC (with modern GPU)? So Mediatek, Rockchip, Allwinner, Apple, etc can be sued by Nvidia? Since the patents are relatively broad (at least from what I know, which isn't much), why Nvidia didn't sue them in the beginning? It is like they let them succeed and then try to reap the benefit.


Lawsuits are expensive. Why bother unless there's a quantifiable benefit?
 
Nvidia should try to make a deal from the beginning so a company doesn't suddenly get charged with lots of money.
 
Nvidia should try to make a deal from the beginning so a company doesn't suddenly get charged with lots of money.

You must have missed this part:

Nvidia said in the complaint that it’s tried since August 2012 without luck to reach a licensing agreement with Suwon, South Korea-based Samsung, the world’s biggest maker of smartphones.

“Samsung has negotiated based on delay and by pointing the infringement finger at its chipset suppliers, such as Qualcomm, or third parties that supply GPU technology to Samsung, while continuing to reap enormous profits from the Samsung-branded products shipped into the United States and elsewhere,” Nvidia said in the complaint.

http://www.bloomberg.com/news/2014-...-sales-in-fight-with-qualcomm.html?cmpid=yhoo
 
Lawsuits are expensive. Why bother unless there's a quantifiable benefit?
Lawsuits, while expensive, typically don't cost US$1.5Bn, though (if you win, I mean...) So yeah, there can be quantifiable benefits. Even if you lose, a court may decide the suing entity isn't entitled to quite as much money as they would like. It's a roll of the dice; you might lose, and if you rather chicken out and pay the ransom then you surely won't win either of course...
 
Back
Top