Submitted by NVIDIA’s aspirations to enter the general purpose processor market may never have been clearly spelled out by the company before, but it was getting more and more obvious as each week passed by. Now, it’s pretty much official: NVIDIA says it’s not a question of “if”, but “when”.
During a Q&A session at the Morgan Stanley Technology Conference in San Francisco, Michael Hara, the company’s senior vice president of investor relations and communications, made it pretty clear that within two to three years, NVIDIA will enter the general purpose processor market. When asked if NVIDIA was was going to enter that market, he responded:
The question is not so much I think if; I think the question is when. I think some time down the road it makes sense to take the same level of integration that we’ve done with Tegra. Tegra is by any definition a complete computer on a chip, and the requirements of that market are such that you have to be very low power, very small, but highly efficient. So in that particular state it made a lot of sense to take that approach, and someday it’s going to make sense to take the same approach in the x86 market as well.
Before everybody gets their knickers in twists, NVIDIA is not going to show up in your general purpose desktop any time soon. Hara explained that it would most likely be a system-on-chip design for MIDs and netbooks. He spelled out a time frame of two to three years.
The big problem is that NVIDIA doesn’t have the license to produce x86 chips. There were some rumours a year ago about NVIDIA buying VIA’s chip division, but those rumours never came to fruition.
We could use the competition with another mainstream/performance CPU option.
PPC64 and ARM are already there. Why bother making x86 processors? MIPS64 is another viable option. I don’t understand them. IBM can provide them processors instead of producing new.
Because 90%+ of desktop users still demand Windows compatibility. Particularly the Nvidia fans (I.E, gamers)
Exactly what license is required? How much of the x86 instruction set is covered by patents? And how does any of the instruction set become patentable?
For SoC, no CPU bus interface license would be required other than memory bus interface, so there should be no problems there. And I’m sure they already have any licenses to produce PCIe interface through their existing chipset arm.
they have to licence a LOT of stuff from Intel, but i would assume tehy would go with the hyper tranport protocal (which has minimal or no licence fees and is an open standard) instead of intels quick path.
their focus would be on very power efficient CPU’s while offloading the bulk of their work to their gpu unit (integrated or not) using custom extensions on the processor level for OpenCL and their CUDA technology. They are licencing some stuff from VIA currently too for the on proc encryption. I can’t say any more i am afraid and as is this is still just all on paper, it will be at least another year before anything remotely tangible shows up.
well one more piece of info i guess couldn’t hurt. Nvidia intends to use TSMC’s or UMC’s 28nm process after the perfect the 40nm process for their GPU’s they are working on now. as for the rest, my lips are sealed.
more of a good read here: http://componentsforlaptop.com/news/2008/10/02/tsmc-bringing-28nm-c…
Edited 2009-03-04 21:41 UTC
Actually, you are incorrect because they already have licensed the necessary components… from transmeta.
http://www.electronista.com/articles/08/08/08/nvidia.transmeta.lice…
Transmeta worked around the key Intel patents by being a firmware-based compatibility system, which can adapt to new CPU cores with relative ease… imagine, if you would, the transmeta x86 translation system paired up to, oh, say a CUDA engine…
yes but that doesn’t mean they have a licence to the x86 instruction set which everyone making an x86 cpu has to get from intel.
Most people are unaware that instruction sets can be copyrighted.
That would be the principal hurdle for nvidia. I am sure they may have figured a way around it.
Copyright for instruction sets do not exist.
It is no more copyrightable than the “copy protection” code in printer cartridges.
Don’t ask me for details, or where to find this, but as far as I remember, the story goes like this:
There was as copyright lawsuit of a printer manufacturer against a cartridge manufacturer. The original printer cartridges had a piece of code, which was necessary for the printer to work. Whenever the cartridge was changed, the printer zeroed this piece of code, and when a “valid” cartridge was inserted, the code was loaded from the cartridge which “unbricked” the printer.
The independent cartridge maker made bit-by-bit copies of this code, and put it into their own cartridges.
The original manufacturer then sued for copyright infringement, because it was HIS code which was copied.
It court it turned out to have been fair use, the argument of the judge why was like this:
– it is necessary to copy the code if a working replacement is to be made.
– the original manufacturer put the code into the cartridge to prevent competition, that is not what copyright is there fore.
– The code is not an intellectually outstanding piece of art, it’s ONLY function at it’s location is, to prevent competition.
The same goes for the instruction set. It is just a list of numbers, with a description what each one does if loaded into the processor.
The list as given by INTEL might be under copyright, but nothing prevents me from looking at that list and writing my own documentation.
Necessarily those EXACTLY SAME numbers would occur, and LOTS of the description wording would be similar or same.
That list is a documentation of FACTS, and the facts themselves are not copyrightable.
Some implementations of these codes in hardware for sure are patented, but that is hardware.
Nvidia might not need to have a license –
From the big book of lies
http://en.wikipedia.org/wiki/Cyrix
Unlike AMD, Cyrix had never manufactured or sold Intel designs under a negotiated license. Cyrix’s designs were the result of meticulous in-house reverse engineering. Thus, while AMD’s 386s and even 486s had some Intel-written microcode software, Cyrix’s designs were completely independent. Focused on removing potential competitors, Intel spent many years in legal battles with Cyrix, claiming that the Cyrix 486 violated Intel’s patents.
By and large, Intel lost the Cyrix case. But the final settlement was out of court: Intel agreed that Cyrix had the right to produce their own x86 designs in any foundry that happened to already hold an Intel license
They might have some wiggle room.
Edited 2009-03-05 18:19 UTC
Instead of forging ahead with novel parallel processor technology, Nvidia thinks that the way to go is to copy last century’s dinosaur CPU. It’s enough to make a grown man cry. Whoever is in charge of research at Nvidia should be given the boot. What a waste of talented engineers! But it’s not too late, Nvidia. Click on the link below and do the right thing. Otherwise, Otellini will tear you a new one and you know it.
How to Solve the Parallel Programming Crisis:
http://rebelscience.blogspot.com/2008/07/how-to-solve-parallel-prog…
Edited 2009-03-04 22:08 UTC
i am not sure what you mean about last generation. are you reffering to the archetecture of x86? we all know thats not the best it could be, BUT its whats out there and if they wan to enter the market thats what they are going to do.
Power: no chance
ARM: maybe but terribly unlikely due to not being able to run and form of windows (CE excluded)
MIPS: hahahaha
SPARC: That would be sooo cool, but totaly useless for teh consumer (though a totally open archetecture so no liecencing)
IA64: it suports windows, just imagine. “Now you can play games with the 8 other hobbiests who have itanuim machines” (though in all fairness mine runs OpenVMS and windows server 2008 like a champ).
___________(insert other not going ot happen CPU here)
x86-64 is the best bet. and while its not the best, its cheap and it meets the users needs.
> MIPS: hahahaha
Why “hahahaha”? Try to google a bit for a term “Loongson”…
yes and its an ok proc (i am looking forward to 3 with has some x86 instruction sets) There are a for other MIPS procs that I like, but aside from a very nitch embeded market (and the once mighty IRIX) MIPS is useless to the desktop user. Unless SGI comes back full glory, and even they went IA64
Incidentally, with you on SPARC there, would be nice. I’m working on the T2 to see about mating it up to a HT + DDR2 bus so as to give a variety of chipset options. A pipedream, I know, but yes, the dream is there.
if you are unable to appreciate the distinction between microarchitecture and instruction set.
Chances are that you are a bit out of your league when tackling problems which are orders of magnitude more complex, like solving the parallel programing paradigm.
A current Core2 or i7 CPU is many things, but its microarchitecture could hardly be classified as a “dinosaur.” Or maybe, the word dinosaur doesn’t mean what you think it means 🙂
Edited 2009-03-05 02:27 UTC
Do you know the distinction between a paragraph and a sentence? Do you even know what a complete sentence is?
Blah, blah, blah.
No amount of sour condescending retorts is going to change the fact that a person, who is unaware of the distinction between microarchitecture and ISA, is a tad out of place in trying to solve one of the biggest challenges facing the architectural and software communities.
There! With plenty of sentences and paragraphs for you to enjoy.
Getting back to topic. It seems that the direction that NVIDIA may be taking is towards including atom cores into their SOC designs. It seems Intel will be using TSMC to second source atoms, and they will be licensing atom IP cores to 3rd parties. I doubt NVIDIA has the resources (and the patience) to do a full IA CPU bring up, especially given the current economic situation.
Edited 2009-03-05 06:16 UTC
Look, jackass. Go read my post one more time. Did I make any reference to architecture or instruction set? No I didn’t. I only mentioned CPU. And yes, the x86 is a dinosaur because the days of the CPU are numbered. Why? Because the computing paradigm is changing rapidly to one where the CPU (a sequential processor) no longer makes sense.
And one more thing; ‘paradigm’ is not synonymous with ‘problem’. Solving a paradigm makes no sense. See you around.
x86 has something no other CPU has – applications. You can actually do useful stuff with it because there is so much written for it. That’s the only thing that matters.
They have nice ARM processor – but are crying for x86… if there were an ATX motherboard for such ARM, I would to test one. If it could give about as much computing power, as – say – Pentium III/700, for today it would be quite enough for me.
http://beagleboard.org/
wait till the end of march for the new revision
“poundsmack; jack of all trades, master of none”
…maybe I should stress
that I would it to be cheaper than x86 motherboard, and not more expensive.
Needless to say, Beagleboard isn’t “desktop-type” ATX motherboard.
true,
but if you did what i did (mount it in the same care as one of my workstation pc’s but on the detachible panel side) then you can just mount it in and route any cables that don’t connect to the mobo directly out through an open PCI slot in the back.
one of these days i need ot take a picture of my den, or as i like ot call it “Vault 101” sooooo many obscure computers running sooo many different OS’s…. scratch that, their are upcomming systems in there and that would end badly for me. well thats what MS paint is for “and put a big black square over this part of the image, perfect” haha
Let the games continue
Nvidia announces x86 chip (better make that in the future dam thous headlines)
http://www.theinquirer.net/inquirer/news/248/1051248/nvidia-announc…
After all, the latest GPUs from NVIDIA have 256 stream processors.
Some would think that NVidia would be smart enough to embrace the present and future and not the past .. aka, should of went 64bit or higher …
last time someone tried to compete with Intel..
AMD comes out with their AM386
TI comes out with a clone 486dx
NEC ships V20 and V30
RISC looks to unseat Intel
and even 10 years before that under the second sourcing by Intel/IBM when dozens also mfg clones.
None today (with a small exception of AMD)
have been able to compete and survive against
Intel. Not that Intel has been evil in its
business plans… its just Semi is a brutal
business to be in.