NVIDIA shows signs ... [2008 - 2017]

[satein]

FYI, an updated news on the legal case battle between NV & Rambus, it seems now the judge rules in favor of Rambus, not NVIDIA.

http://www.engadget.com/2010/01/23/court-case-shocker-judge-rules-in-favor-of-rambus-not-nvidia/

... A judge at the U.S. International Trade Commission issued a preliminary determination finding that NVIDIA had indeed violated three of the five patents -- ruling that the other two had not been violated. Now, NVIDIA says that the whole patent mystery continues to be a subject ripe for rexamination by the Patent and Trademark Office -- which has consistently found the claims of infringement to be invalid. NVIDIA says it will take the claims to a full commission for a final decision.

The source is from WSJ, which I am not subscriber.

 

[Groo The Wanderer]

I can give you some average point per day figures for my various NV cards with current projects over the last 6 months or so. Keep in mind all cards are overclocked by anywhere from 5-15% on the ALU domain, which has the greatest impact on F@H performance:

GTX 285: 9000-9800 PPD
GTX 275: 7800-8600 PPD
8800 GTS 512: 4800-5900 PPD
8800 GT: 4500-5500 PPD

My roommate's 4870x2 gets so hot and the fan winds up so high when folding that I actually had to stop because of it, but when it was running it would pull about 3000-4000 PPD.

Cool, thanks. Time to dig, travel season is good for white papers.

-Charlie

 

[ShaidarHaran]

Revised numbers on the 275 after o/c'ing to 675/1512/2484:
8000-9200 PPD

 

[dizietsma]

The *only* reason I've been buying NV cards lately is for Folding @ Home. Between my overclocked 285 (702/1584/2754) and 275 (648/1440/2304 - factory OC, I plan on OCing further) I get somewhere between 17,000-19,000 points per day. If ATi's cards could produce the same amount of points per day as Nvidia's, I would buy them instead. It's pretty sad when my roommate's beastly 4870x2 gets outperformed by my old 8800 GT which I'm passing on to a friend. If and when Pande Group ever gets around to releasing the GPU client version 3 written in OpenCL, ATi cards should finally start performing closer to their NV counterparts at which point I can look at replacing my 285, 275, and 8800 GTS 512 in my 3 gaming PCs. Hopefully by that time Northern Islands will be out and I can upgrade to DX11.

I still have something of a soft spot for AMD & ATi parts. I used to be something of an outspoken FanATic over at Rage3d but have moderated my views quite a bit since G80 and Conroe came out in '06. Sadly, I haven't owned an ATi card nor an AMD processor since the release of the aforementioned architectures. Holding out hope for Bulldozer and Northern Islands though.

For gaming purposes its all a wash right now as the games I play aren't that demanding (Titan Quest, UT3, COD4, L4D) and 3 monitors I use are a 1360x768 32" LCD powered by an 8800 GTS 512, a 1680x1050 22" LCD powered by the GTX 275, and a 1920x1080 42" plasma powered by the GTX 285. If I were running multiple displays or higher resolutions I could see the need for more horsepower for gaming.

That's interesting, I used to be a complete nvidiot, well at least I always used to buy their cards, but since the G80 I have been MR ATI man with this 4850 and like it a lot. I'll see how Fermi does but I can see myself going ATi again for the next upgrade. So I am now neutral Funnily enough I play a lot of Titan Quest now and with nvidia I always got problems with corruption, now with the Ati 4850 I get slow downs with high quality shadows so have to turn them off, even at a modest resolution of about 1600x ...

Apart from Crysis which I love my current card is up to the job it has to be admitted.

 

[willardjuice]

Let's try to stay closer to the topic of Nvidia's financial viability and less about the current status of support for OpenCL. A few users have voiced their concerns (through various means) about where this thread is headed and I am inclined to agree with them.

Again to make things perfectly clear, I am not against the discussion itself, but rather the place it is being discussed. If users feel the discussion needs to go on, they can start a new thread (or perhaps post in OpenCL thread in the GPGPU sub-forum).

Another Mod's Edit: I spun the OpenCL on ATI OT off into the wild GPGPU yonder. --Pete

 

[Ninjaprime]

They lost their case with Rambus, if its anything like Samsung's case, it could cost $900 million. That could hurt with the recent issues they've been having. Seems to me their case was bigger than Samsungs, but I don't really know full details. If they end up paying $1 billion+ that could potentially sink them fast. The infringment pretty much covers every card NV has sold, so they could claim some kind of back royalties on the majority of NV's sales, if I'm hearing it right.

 

[Sxotty]

It is not in a parasites best interest to suck the host dry. Rambus will want enough money to take Nvidia's profits away, but let them break even. Plus a one time penalty to take away their cash horde. Then hope they can siphon off for ages.

 

[FrameBuffer]

It is not in a parasites best interest to suck the host dry. Rambus will want enough money to take Nvidia's profits away, but let them break even. Plus a one time penalty to take away their cash horde. Then hope they can siphon off for ages.

Seeing as ATI paid 75M for 5 years (15Mper year) I'd guess Rambus would seek 15*8 + punitive + future.. best guess says around 250M (but hey I'm no lawyer..) which I think NV would still be getting off on the cheap considering they went nearly a decade long of producing parts that infringe on rambus IP while others paid.

 

[itsmydamnation]

you always settle out of court for less, Also wasn't that before AMD bought ATI and ATI fell under AMD's agreement? I think RAMBUS will ask for much more from NV. What i dont get is, if intel and AMD have settled with rambus why didn't NV? intel and AMD would only settle if they knew they couldn't win what was so different from NV?

maybe NV see the stakes that high that its win at all costs

 

[FUDie]

What i dont get is, if intel and AMD have settled with rambus why didn't NV? intel and AMD would only settle if they knew they couldn't win what was so different from NV?

intel and AMD settling doesn't necessarily mean they didn't think they couldn't win, it's a matter of risk. They may have chosen to settle because that was what they felt was the lower risk path.

-FUDie

 

[itsmydamnation]

yes which likely implies that you have infringed on IP.

 

[Silent_Buddha]

Not necessarily it could also be that they think they are not infringing but that a judge or a court of law might not view it the same way. Especially if they don't have the technological knowledge to know the intricacies and thus rely on testimony from experts.

Or they could be infringing.

A settlement doesn't equate to guilt. It just means that they have determined that settlement would be less costly than going to court win or lose.

Regards,
SB

 

[Sxotty]

yes which likely implies that you have infringed on IP.

Not really. They actually sit down and say, given that we did not infringe what are the expected values for various outcomes. Then they decide the odds of different outcomes and make a decision.

A company that instead just says "F* this we are innocent and the lawsuit is stupid" is taking a huge risk even if they are right.

 

[FrameBuffer]

you always settle out of court for less, Also wasn't that before AMD bought ATI and ATI fell under AMD's agreement? I think RAMBUS will ask for much more from NV. What i dont get is, if intel and AMD have settled with rambus why didn't NV? intel and AMD would only settle if they knew they couldn't win what was so different from NV?

maybe NV see the stakes that high that its win at all costs

The difference being that Rambus has been trying to "settle" the matter with NV for 8 years now, going back to the GF4 iirc and includes nforce mobos too.. so it's simply not one product line its across numerous products. The only quote I found was ""The license agreement focuses on the memory controller logic needed to connect to XDR, Rambus' eXtended Data Rate memories. The deal also includes Rambus patents covering DDR-2, DDR-3, FB-DIMM, and PCI Express technologies, as well as other current and future high-speed memory and logic controller interfaces....."" dates back to Jan 2006 (signed in 2005), predating the merger/acquisition of ATI. Intel has a cross license deal with Rambus back in 2001. The recent ruling centers around "Barth patents" and gives RAMBUS "patents for DDR3,4,5+ memory (controllers) that last until October 2015 (20 years from Oct 95, the application date)".

The 3 patents NV is found guilty of violating are:

Barth '109 - Method of controlling a memory device having a memory core
Barth '405: Protocol for communication with dynamic memory
Barth '353: Protocol for communication with dynamic memory

EDIT**

Added FTC investigation docket info : Certain Semiconductor Chips Having Synchronous Dynamic Random Access Memory Controllers and Products Containing Same, including Graphics Cards and Motherboards, Inv. No. 337-TA-661

 

[NathansFortune]

yes which likely implies that you have infringed on IP.

Not at all. You can settle for a number of reasons, people do it all the time. Lawsuits by and large generate a lot of bad press, people don't like being associated with someone or a company that has a very public lawsuit in the courts.

If AMD/Intel decided it was less money to settle than the amount of damage a lawsuit would do to their brand then settling is clearly the way forwards regardless of who did what.

 

[Geo]

That's interesting, I used to be a complete nvidiot, well at least I always used to buy their cards, but since the G80 I have been MR ATI man with this 4850 and like it a lot. I'll see how Fermi does but I can see myself going ATi again for the next upgrade. So I am now neutral Funnily enough I play a lot of Titan Quest now and with nvidia I always got problems with corruption, now with the Ati 4850 I get slow downs with high quality shadows so have to turn them off, even at a modest resolution of about 1600x ...

Apart from Crysis which I love my current card is up to the job it has to be admitted.

ZOMG. Dizzy as a fanATIc? I've been away too long. . .

 

[chavvdarrr]

Its disgusting that people like those in Rambus are allowed to "make money" this way.

last I heard a US medical firms already got patents on 20% on human genome... I guess in not so distant future US court may start suing every human on the planet - for patent infringement?

 

[dkanter]

1. Calculations regarding damages that Rambus might seek are totally off. If ATI paid up $15M, then it's not unreasonable that Rambus start with $25-30M from NV (since NV has far more revenues and units sold). Plus they would probably slap on willful infringement (triple damages), so you could be talking close to $1B. That being said, Rambus is unlikely to ever get that, since they will likely settle out of court.

2. Rambus is actually quite an innovative company, and they have done a lot of novel work with high speed I/Os well ahead of the rest of the industry. If you just compare how RDRAM works to things like GDDRx, you'll see they were very much ahead of the curve. So to the extent that they truly invented those techniques, they deserve credit and compensation for it. I think it's clear everyone would love to use their technology...but the price and design changes are huge barriers (plus massively negative publicity).

3. The whole rambus vs. DRAM makers issue is hardly black and white. Without really knowing what happened, there's no way to come to a reasonable conclusion.

David

 

[rpg.314]

1. Calculations regarding damages that Rambus might seek are totally off. If ATI paid up $15M, then it's not unreasonable that Rambus start with $25-30M from NV (since NV has far more revenues and units sold). Plus they would probably slap on willful infringement (triple damages), so you could be talking close to $1B. That being said, Rambus is unlikely to ever get that, since they will likely settle out of court.

I am pretty sure that 8 years of talks and all these lawsuits aren't worth $25-30M to nv. The number is probably in hundreds of millions for them.

2. Rambus is actually quite an innovative company, and they have done a lot of novel work with high speed I/Os well ahead of the rest of the industry. If you just compare how RDRAM works to things like GDDRx, you'll see they were very much ahead of the curve. So to the extent that they truly invented those techniques, they deserve credit and compensation for it. I think it's clear everyone would love to use their technology...but the price and design changes are huge barriers (plus massively negative publicity).

Publicity aside, if GDDRx indeed stalls, I have a feeling AMD just might go their way. After all they were the first to shift over to GDDR5 from GDDR3. Nv has typically been conservative with their memory transitions as well.

 

[Silent_Buddha]

1. Calculations regarding damages that Rambus might seek are totally off. If ATI paid up $15M, then it's not unreasonable that Rambus start with $25-30M from NV (since NV has far more revenues and units sold). Plus they would probably slap on willful infringement (triple damages), so you could be talking close to $1B. That being said, Rambus is unlikely to ever get that, since they will likely settle out of court.

It depends on how many years of violations the settlement was for, and then added to that whatever per year licensing fee's AMD is currently paying.

I'm going to take a guess that the 8 years the lawsuit has been going on + damages sought for any time prior to the start of litigation will factor in heavily into what Rambus is currently seeking. So it has the potential to be far higher than 15 million USD. Especially if the judge finds in favor of Rambus, settlements usually go for less than the potential damages awarded if the plaintif wins.

Regards,
SB