NVIDIA's CEO, Jen-Hsun Huang just announced Project Denver - its first CPU architecture design ever, based on ARM's ISA. This is a custom design done by NVIDIA in conjunction with ARM and targeted at the high performance computing (HPC) market.

This is a huge announcement from NVIDIA, but not entirely unexpected. Prior to Project Denver NVIDIA licensed ARM IP but developed its own IP everywhere else for use in Tegra. Going forward, NVIDIA is turning into a full fledged SoC architecture company. This is a huge step in NVIDIA becoming a major player in the SoC evolution going forward.

Update: NVIDIA provided some more details on the announcement. Project Denver is targeted at everything from PCs to HPC/servers. This is completely a high end play going after the x86 stronghold. Project Denver ties in completely with Microsoft's announcement to bring Windows 8 to ARM next year.

NVIDIA also announced that it will be licensing ARM's Cortex A15 core, presumably for use in lower end devices (e.g. smartphones). I wouldn't be surprised if NVIDIA eventually moves to its own architecture based on ARM across the board.

Comments Locked

47 Comments

View All Comments

  • tipoo - Wednesday, January 5, 2011 - link

    http://www.microsoft.com/presspass/press/2011/jan1...

    These two pieces of news together are rather interesting...I'm thinking of a Cell like processor with one or more cores (the ARM part) that tell everything else what to do, with thousands of CUDA/cores/stream processor units doing what they are instructed. Lets face it, we all thought Huang was a bit nutty when he said x86 processors are near their demise, but with this news...I'm not so sure.
  • has407 - Wednesday, January 5, 2011 - link

    Your description sounds very close...

    "An ARM processor coupled with an NVIDIA GPU represents the computing platform of the future. A high-performance CPU with a standard instruction set will run the serial parts of applications and provide compatibility while a highly-parallel, highly-efficient GPU will run the parallel portions of programs."

    http://blogs.nvidia.com/2011/01/project-denver-pro...
  • tipoo - Wednesday, January 5, 2011 - link

    Huh, I didn't see that before...Exciting!
  • has407 - Wednesday, January 5, 2011 - link

    Agree. And while there's quite a bit of hyperbole, Nvidia's doing what they gotta do, otherwise they'll be forever stuck on the end of a PCIe (or whatever) as the tail on the dog, while Intel, AMD, and Via can tightly integrate with the CPU.

    While the x86 crowd will be around for a long time, ARM's licensing and business model is more robust IMHO. I have to wonder if there are talks deep within Intel about opening up to be more ARM-like; "Trust us, we can deliver the best solution--just buy our chips" has obviously worn thin. Apple made the shift a while ago. Microsoft's announcement makes it clear that is no longer an acceptable MO.

    Exciting times indeed!
  • Lolimaster - Wednesday, January 5, 2011 - link

    When your Zacate awesomeness article? Can you do that or "higher forces" daly that article?

    http://scientiasblog.blogspot.com/2006/09/anandtec...
  • T2k - Wednesday, January 5, 2011 - link

    Ahahahaha, SOOOOO TRUE!

    Go to the AT forum and you'll get served as soon as you bring this up - you even got moderators running around with tags proudly advertising their Nvidia-sponsored status while handing you temporary bans for "moderator attack" if you dare to question their "coincidentally opened" pro-NV topics, full with screaming PR nonsense, right around AMD newsdays...

    ...and if you insist then some half-retarded forum imperator simply tells you in his "high educated (= attack-dog) style to STFU and be happy you weren't permanently banned.

    One has to love the "fair" and "balanced" Anandtech, errr, I mean FauxNews, of course, of course, dear moderator.
  • lchen66666 - Wednesday, January 5, 2011 - link

    These days advanced CPU requires large amount resource and money to develop. Can someday
    ARM really be able to replace x86 in desktop/laptop/service market ? I serious doubt. Simply, because it took INTEL around 30 years to build the eco systems.

    Plus INTEL still has one of the most advanced FAB in the world. They have a lot of R&D in manufacture
    process. Even all software is ported to ARM(total number of software apps reaches the same level of x86) some day(this may never happen), I don't think NVIDIA of its partner can compete with INTEL in the manufacturing process. INTEL can make hundreds dollars for every high end chip sold now. The large part of the profit goes to FAB equipment upgrade. Can Nvidia have that kind of margin ? No way.
    Nvidia mostly will ask TSM or Samsung to make chip, but those FAB companies will charge premium if you want your chip to be manufactured with the most advanced equipment. Just think about even AMD with the 100% compatible x86 chip can not make as much as INTEL because very high end consumers always go with leader, which is INTEL.

    People buy Nvidia/ATI graphics chip because they are the leaders. For a high end CPU, that's definitely not possible. Just look at Sun Sparc chip, and even IBM power chip. Where they are now ?
  • T2k - Wednesday, January 5, 2011 - link

    EXACTLY.

    And on top of all this even Intel ITSELF FAILED TO REPLACE x86 with a new ISA - anyone remember IA64? - and Nvidia is VERY FAR from Intel and Intel's resources to push something like that, let alone Nvidia's usual non-compromising, all-out-dog-of-war, user-can-go-suck-@ss-when-complains business attitude...

    ...let's face it, it's mostly about MOBILE DEVICES + LOT OF EMPTY TALK by JHH, as usual, just to TALK UP STOCK PRICES.
  • has407 - Thursday, January 6, 2011 - link

    Yes, Intel and x86 is a powerhouse, but beware as bigger have fallen. (And I'd argue as to whether Intel's x86 CPU's are necessarily more "advanced" than ARM--different yes, but not necessarily more advanced. But that's another subject for another time.)

    On the x86 side we have Intel, AMD and Via. They make chips, and their business model is based on the assumption that they know what's best ("we make, you buy; leave the details to us"). Want a custom chip with an x86 core and your custom logic surrounding it? You can't get it unless you can convince one of those three to do it for you; there is no other option. Good luck. Since they're producing chips for general consumption and their own markets--not your specific needs--be prepared to bring a boat-load of $$$ to the table.

    On the ARM side we have at least a dozen (?) architectural licensees (Samsung, Qualcomm, Apple, Nvidia, etc.), and many more with lesser IP and fab licenses. Want a custom chip with an ARM core and your logic? No problem. Pay the $ for a license and DIY, or find one of the existing licensees who will do it for you, whether simply fab'ing it or integrating your custom design. Note that there are no such options with x86.

    The Intel model assumes that a very small number of tightly controlled suppliers can and should serve the needs of all customers (as well as those suppliers' own interests); the options are few. The ARM model assumes a variety of providers, and that customers can pick and choose as needed from full DIY to "just fab me a chip with the following specs.."; the options are many.

    Which business model is more robust? I'd place my bet on ARM--especially when a high level of integration/customization is demanded. Intel, AMD and Via simply can not be all things to all people and serve their own vested interests--of which they have built a business around, and IMO they are products and artifacts of an outmoded licensing model. Maybe a change in that model would improve things, but I doubt it could come fast enough to make a difference.

    Does that mean ARM is going to take over the desktop or server market any time soon? No. However, those aren't the major growth areas. So while x86 is going to be around for a long time, it's likely to be an ever diminishing slice of the pie. So Intel et. al. successfully defends those markets. Great. But in the end it will get them nada. Defending such territory would be a Pyrrhic victory at best, and why many of the once mighty who commanded major market share are now relegated to fractional niche market share.

    And yes, Intel has great fab facilities. But it it's not producing what customers want and need, what does it matter?
  • Computer Bottleneck - Thursday, January 6, 2011 - link

    "Defending such territory would be a Pyrrhic victory at best, and why many of the once mighty who commanded major market share are now relegated to fractional niche market share.

    And yes, Intel has great fab facilities. But it it's not producing what customers want and need, what does it matter?"

    Lets hope Intel is thinking Long term. I would hate to see them try to maintain some short term success at the risk of stalling Future technology advances.

Log in

Don't have an account? Sign up now