Nvidia vs Intel

[MarcVenice]

In a different thread a discussion sparked about Intel and Nvidia and the lawsuit that will happen next year. It has been said that Nvidia's holds patents that Intel will infringe if they were to produce a graphics chip. Since Intel is stopping Nvidia from producing chipsets for Nehalem cpu's, there's a possibility Nvidia would try to stop Intel from making any graphic chips. Now those who are in the know, what patents does Nvidia hold that will stop LRB dead in it's tracks? Could Intel try to work around this, maybe license some IP from ATI?

 

[MODEL3]

The most probable thing is NV to be right about this one.

They have so many freakin patents regarding graphics related staff that is quite possible.

I guess Intel must have deals with ATI/AMD and Imagination technologies (IT holds some patents also) also but i don't know if their IGPs or the upcoming LRB products can be done with only those agreements.

Here is some info about what Huang?s said refarding the countersuit:

A moment ago, we announced that we have responded to Intel?s lawsuit against us with a countersuit of our own.

In our countersuit, we assert our belief that we are licensed to build chipsets for Intel processors. In a pair of agreements signed in 2004, we negotiated for rights to build chipsets. In exchange, Intel obtained a cross license to our valuable patents. Today, Intel is using technologies that we invented in their integrated graphics chips. And they will soon integrate NVIDIA patented technologies into their CPUs and upcoming Larrabee processors.

We claim in the countersuit that they have breached the chipset agreement by spreading damaging claims that we are not licensed to design chipsets for future processors and by Intel taking other actions in the market. Why are they suing us over products that we have not announced or offered to sell? Because they see the tremendous interest in ION and are resorting to every possible tactic to slow us down. ION not only displaces their chipset sales. ION also promotes the use of less expensive, and adequately sized CPUs. ION is a double whammy for them. So with this legal tactic, they are trying to slow down ION adoption by claiming that ION has no future. Instead of going back to the drawing board to build a better chipset, Intel is using their market power and legal antics like this to keep the world from benefiting from ION.

We also claimed that Intel has breached the implied covenant of good faith and fair dealing. It was clear from our agreement what each party wanted from the deal. We wanted to build chipsets for Intel CPUs, thereby offering products to the single largest segment of the graphics market. In exchange for the opportunity to participate in this multi-billion dollar segment, we granted them a cross license to our graphics technologies, which they desperately wanted. They are surely realizing the full benefits from a license to our patents. Yet, they are doing everything possible to prevent us from realizing the benefits that we negotiated for.

The bottom line is that we are license, and they have damaged us. And they have breached our agreement and should, therefore, lose the benefits of our cross license.

We didn?t start this fight, but we will surely rise to the challenge.

And we will not allow the world to be distracted by this tactic. In fact, we will use this as a platform to shout from the highest mountain why Intel is so afraid of ION ? this is the era of visual computing, and the GPU has become the soul of the new PC. And customers deserve better than Intel?s graphics.

Check the underlined part to see where Nvidia is aiming also.


http://games.venturebeat.com/2...chip-set-countersuit/#

 

[BenSkywalker]

Page 12 frames the basis for where we ended up.

As a bit of a continuation of that thread, the biggest trump card Intel has is-

4,727,365 (a computer video image generating system including a computer memory having 3D object data stored therein employs an advanced object generator for retrieving and processing the object data for output to a span processor for controlling the pixel-by-pixel video output signal for a video display -- the advanced object generator includes a translucency processor, an edge-on fading processor, a level of detail blending processor and a bilinear interpolator for texture smoothing)

There are two major issues with this patent as it applies to nVidia. The first one, this patent only covers parts that handle things between chips(which was common prior to the GeForce), the other major issue is that nVidia has full access to these patents as they were all partly owned by SGi who has a cross licensing agreement with nV.

nVidia's patent collection. I can't find an archive of 3dfx's IP which nVidia also owns(some nasty ones to get around there).

Can't find the exact patent numbers, but to give an idea, nV now owns the patent for multitexturing and for LOD based mip mapping(they have tons, these are some of the old 3dfx IP that is damn near impossible to get around unless you move to a purely procedural rendering environment). Trying to dig them up, it gets a bit tricky when the original company gets consumed by another.

 

[MarcVenice]

I wonder how this license is legally set up. If Intels keeps reaping the benefits, whilst Nvidia can't make any chipsets for Intel cpu's, the license has become futile for one party. Logically, it makes sense to end the mutual license agreement and that it can be ended by one party. Intel must feel confident that this is not the case though. Or Intel must feel confident that they can get the patents they MIGHT infringe, to be taken away from Nvidia. Or perhaps they think court will rule that Nvidia has to sell Intel a license, like Intel had to sell a x86-license to AMD?

Also, Ben, how sure are you that 4,727,365 is owned by Intel and partly owned by sgi, and that Nvidia has a cross licensing agreement with them? Because that trump card migh be all Intel needs?

 

[BenSkywalker]

Also, Ben, how sure are you that 4,727,365 is owned by Intel and partly owned by sgi, and that Nvidia has a cross licensing agreement with them?

100%, I can walk that particular patent back a ways(Intel-Real3D-Lockheed-GE). Real3D had a shared IP agreement with SGi, this gave Intel access to all of SGi's patents and nVidia worked out their own agreement to do the same. In terms of corporate ownership, that one has been around for a while and has been a part of many suits over the years. This would be kind of like Nintendo buying Cell CPUs from IBM. While Sony owns IP rights to the technology, IBM has shared rights(along with Toshiba). The patent is certainly valid, but it is already licensed via SGI(and saying it is valid in as much as it is a legit patent, given the exacting terminology used I am not sure anything post GeForce/Radeon would fall under it).

Or Intel must feel confident that they can get the patents they MIGHT infringe, to be taken away from Nvidia. Or perhaps they think court will rule that Nvidia has to sell Intel a license, like Intel had to sell a x86-license to AMD?

This last option is not unlikely, but it comes with a bunch of benefits for nV still. Because nV no longer competes in the integrated chipset business they can argue that anything Intel makes using their IP is diluting the value of said IP and hence demands a higher licensing cost then would normally be agreed upon in such a case. This is an argument that is easy for them to prove, and should be easy for them to win. Either that, or perhaps nV will argue in court the only IP Intel has of comparable value to their graphics IP is a x86 license. I honestly have no idea what the end result of all of this is going to be, but it is certainly going to be interesting to watch.

 

[Idontcare]

Originally posted by: BenSkywalker
Either that, or perhaps nV will argue in court the only IP Intel has of comparable value to their graphics IP is a x86 license. I honestly have no idea what the end result of all of this is going to be, but it is certainly going to be interesting to watch.

Maybe not the place for me to initiate dialogue on the sub-topic, but let me just ask does NV really have the resources to develop a competitive x86 processor alongside continuing the development of their existing GPU-based product strategy? (i.e. what AMD is doing)

I understand they are in a much better place cash-wise and debt-wise versus AMD, but do they have the revenue and cash-flow necessary to actually pursue creating x86-based products that wouldn't see shipment for revenue until 2014 even if they were given a free and unfettered x86 license today?

If we look at what lengths AMD had to go thru to get where they are in the x86 world, it basically took DEC imploding so AMD could hire up one of their teams (which included Dirk as the K7 project lead, the current CEO) so they could still be here 10yrs later...who is left out there to implode (IBM/Power7? SUN?) that would shed the kind of talent and fresh ideas an Nvidia would need to gobble up so they could dive into the x86 world without having failure be a foregone conclusion?

Asking NV to go from zero to competitive in x86 processor architecture is like expecting Intel to go from zero to competitive with Larrabee in the same timeframe but hamstring their R&D efforts by cutting the development budget to about 10% what it has been resourced at thus far.

 

[JSt0rm]

now if nvidia was making x86 cpus the world would be much more interesting

 

[Wreckage]

I think this is all saber rattling.

If AMD truly does get forced out of the market, Intel and NVIDIA will be a huge target for anti-trust. I guarantee they will point to each other as "competition".

Even though it will be just another monopoly like Creative.

Because dispute this "dispute" the Ion platform is rolling on and SLI now can work on Intel chipsets.

 

[BenSkywalker]

Maybe not the place for me to initiate dialogue on the sub-topic, but let me just ask does NV really have the resources to develop a competitive x86 processor alongside continuing the development of their existing GPU-based product strategy?

I don't think they need to. Even if we say nVidia had a x86 CPU that was ten times faster then i7 ready to launch the second they got the license, they don't have close to enough fab available to them to be able to leverage that effectively. They are going to be limited in terms of what quantities they are able to bring to the market and that situation isn't likely to change any time in the near future.

I also don't see them even attempting to build a x86 from the ground up, that would be far too intensive on the R&D front given the amount of engineers they have at their disposal. What I can see them doing is adapting a GPU with a x86 front end and OoO support and quickly making a hack solution that would be capable of dealing with fairly simple scalar code in a reasonable fashion while still being utterly dominant in FP and vector tasks.

Would they be capable of fielding a competitive general purpose CPU to compete with Intel in a timely fashion? Highly unlikely, so what possible benefits could they gain? One of the big ones that I can see having an immediate impact in the current x86 cluster market. Pretty much any task that can be spanned across dozens or in some cases hundreds of machines would likely be significantly more cost effective running on nV's hardware, even when you take into consideration the enormous margins they would charge for such an offering. Take their current approach with CUDA and think of it no longer requiring developer support, you can just take your x86 code base and you are good to go. The potential impact of this for numerous HPC applications would be huge, and could very quickly cement nVidia's dominance in the HPC market. Using this approach they can take their time expanding the functionality of their GPU platform until it becomes viable to compete with AMD/Intel in the general x86 market or at the very least carves out their own niche which they can steadily rake in cash from.

While large corporations are going to look at Sysmark performance and the like, what if nV were to offer a CPU that was utterly destroyed under that metric but was utterly dominant in gaming performance and encoding? Obviously the corporate world would have very little interest, but if nV was able to gain a dominant position in the HPC market and also take over the enthusiast/gamer segment of the market it would afford them the R&D funds to continue to attempt to expand their market position long terms. Obviously this would be a several years, perhaps a decade long trek moving their GPU platform to one that was better suited for the CPU market, but taking their current technology and branching it off into two different directions seems like a realistic approach. We know that nV has already spent enormous R&D efforts in the GPGPU space, that work would almost instantly pay dividends if they were to gain a x86 license.

I'm not saying that any of this is going to happen in any way whatsoever, but looking at the market conditions and what nV has in terms of technology it seems like they could leverage a x86 license rather effectively to expand their current position by a very healthy amount. If they manage to take a mere 5% of the x86 market it would almost double their revenue, given the fab availability that they currently have that would likely be about the most they could reasonably expect outside of market conditions changing considerably.

I think in terms of general purpose CPU Intel will maintain dominance no matter what, they have large advantages in fab process, staggering resources and vastly more experience. What I can see is nV having the potential to carve out a distinct slice of the market without devoting a staggering amount of R&D trying to go head to head with Intel in a market they don't stand a chance of winning or even of serving if they could by some divine act create a superior offering.

 

[Dribble]

I suspect they all rely on each other - pretty well the only way you can start suing for patents is if the company is no longer producing a product.

Assuming Intel are forced to back down on the chipset stuff, their bullying has still managed to cost nvidia about 1 year of chipset sales, which is a big win for Intel. Don't know if there's anything Nvidia can do about this.

 

[Janooo]

Does Intel own any 3D patents that are used by NV and ATI?

 

[ZimZum]

Originally posted by: Janooo
Does Intel own any 3D patents that are used by NV and ATI?

I would assume so. If nothing else than on something like the memory controllers alone.

Thats why I think this is a bad road to go down if your nVidia. Getting into a legal war with Intel is probabably not a good idea in the long run.

 

[Genx87]

Originally posted by: ZimZum

Originally posted by: Janooo
Does Intel own any 3D patents that are used by NV and ATI?

I would assume so. If nothing else than on something like the memory controllers alone.

Thats why I think this is a bad road to go down if your nVidia. Getting into a legal war with Intel is probabably not a good idea in the long run.

Really depends on if one has the cash to outlive them. AMD has had a lot of success in courts against Intel. I actually cant think of a single case AMD lost to Intel. Intels track record in court is actually terrible. They just hope to suck you dry and you give up. But if it actually gets to a jury they dont have a good track record.