We’ve dealt with this ongoing rumor previously. Because of the required licensing issues involved, I doubt if this will ever happen. My guess is that Nvidia is trying to force Intel to buy it.
The Inquirer – August 20, 2008:
THE HOT RUMOR going around IDF is one that we discussed a long time ago when Nvidia bought Stexar: that the company will do an x86 part. The background whispers say that the part will be announced next week at Nvision, but we don’t see that happening.
As we said here and here, Nvidia’s men in white coats certainly have the brainpower to do it, but they also most certainly don’t have a license to sell such a part. Nvidia may have the arrogance to attempt such a thing while hoping to prevail in court, but I can’t see how they would be stupid enough to believe they have a realistic chance of winning that fight.
I’m surprised they do not merge with VIA. I can’t see how NVidia can remain just a graphics company in light of Intel and AMD complete platforms.
The world is a big place. Who owns what intellectual property in other countries is not the same.
Sound a bit unlikely that intel moves to buy nVidia. intel already is the biggest GPU manufacturer, albeit in the embedded market which is a low margin one. intel also produces it’s own chipsets, so no need of nVidia here either. Larrabee is a long shot. intel is not worried to compete head to head with ATI/AMD or nVidia in this segment just now. Larrabee is the first stepping stone into something that will be very important in the future: the GPU as a dedicated code execution unit. That’s why they made it virtually of old x86 cores… not to play Crysis, but to run “fold @ home”.
OTOH nVidia is getting on the shorter and shorter end of the stick: ATI is kicking it’s butt and if intel won’t buy them, they’ll be forced to develop another x86 contender, to have a complete platform solution. Buying VIA or merging would indeed deliver them the licensing…
who knows?
I never said that Intel needed to buy Nvidia. I think it’ll buy it just to shut it up. Nvidia has been threatening to sell a CPU for years. For the cost of litigation to stop that from happening, Intel could simply buy Nvidia outright and be done with it.
The only other option is that Nvidia is not trying to force a sale, but is actually serious about selling an x86 CPU without obtaining a license, and that’s impossible.
When faced with the improbable and the impossible, I’ll chose the improbable.
I think nVidia can pull it off. They bought Stexar for a reason, and it has to be for more than just (in effect) selling it to Intel. After all, Stexar was created by ex-intel guys.
If I were nVidia, I would put executable microcode into the GPU which would allow X86 code to be executed with the massively parallel GPU pipeline alongside the GPU instruction set.
I bet that would skirt many of the Intel patents. At least those which still are in force. After all, patents are good for 20 years max, and then it’s anybody’s ball game. I bet if we looked hard enough, we would find that any potentially infringing patents have expired.
Here’s some background on the subject. (it’s old, but might be coming to fruition now…)
http://tinyurl.com/6c9ggr
part two
http://tinyurl.com/6zqmom
#7 joaoPT. Good posts. I’m interested to see how this all works out.
I am very interested to see how nVidia skirts the patent issue. To me, the one item not discussed in any article is the sunset of the patents. Patents on the general operation of the microcode have expired or are about to. After all, the 386 CPU and all associated patents have already expired and the Pentium came out in 1993, which leaves it’s patents nearly expired if not already.
Now most of the patents for the x86 architecture seem to be related to manufacturing processes, which nVidia doesn’t have to worry about.
I’m at IDF right now. When you see what Intel has got coming out in the next three years for CPUs and GPUs, they’ve got no reason to do anything with or about nVidia. nVidia should be a lot more worried about seeing their margin eaten away.
Just one thought:
Wonder if nVidia could get away with a CPU, not made to support x86 code, but with optimizations to support a software x86 emulator? Wouldn’t that circumvent x86 licensing?
NVidia owns Transmeta technology. They don’t even need to make an x86 compatible CPU, since it will just go through a code translator. Should be incredibly fast even after having to dumb itself down to x86-64 through software.
nVidia licensing of Transmetas code paired with their purchasing of Stexar says that nVidia might not need a patent from them at all in the end.
Welcome to the new paradigm.
11 & 12
Thanks, my point exactly…