CES 2011: Nvidia is developing its first CPU

CES 2011: Nvidia is developing its first CPU

Summary: Most of Nvidia's press conference at CES was devoted to its Tegra 2 processor for smartphones and tablets, but CEO Jen-Hsun Huang saved one big surprise for the end, a project to develop a high-performance CPU for PCs and servers based on the ARM.

SHARE:
TOPICS: Hardware, Processors
18

Most of Nvidia's press conference at CES was devoted to its Tegra 2 processor for smartphones and tablets, as expected. But CEO Jen-Hsun Huang saved one big surprise for the end, a project to develop what he called the first high-performance CPU for PCs and servers based on the ARM instruction set.

Nvidia didn't provide many details but Huang described Project Denver as an effort to build a new CPU from the ground up, which suggests that the comp any has an architectural license from ARM to build a CPU from scratch. Other companies develop ARM-based chips this way--Qualcomm designed its Snapdragon application processor from scratch, for example--but most chips including Tegra 2 are built on existing ARM cores such as the Cortex-A9 MPCore in order to get to marker more quickly and at lower cost.

There have rumors on and off for years that Nvidia was developing its own x86 CPU to compete with AMD and Intel. With this move, Nvidia seems to betting that the ARM ecosystem--the tools, compilers and applications already widely used in mobile and embedded devices--will gradually take over the rest of the computing world as well.

"The energy around the ARM architecture is now absolutely enormous," Huang said. "This is Nvidia's first CPU development. And the world's first ARM processor targeted at high performance computing."

It is an ambitious strategy, but it faces some big challenges. First, while ARM processors have been growing significantly more powerful, they still can't touch the fastest x86 processors. Second, many of the client and server applications used by organizations and individuals would need to be rewritten to run on ARM-based computers.

A number of major operating systems including Apple's iOS, Google's Android, RIM's BlackBerry and Windows CE already run on ARM-based processors, but there's one big omission: the full version of Windows. Huang noted that there are rumors that Microsoft could announce that Windows 8 will support ARM, as well as x86, as soon as today. If that does happen, it would go a long way towards validating Nvidia's strategy.

[Update: Microsoft has confirmed that the "next version of Windows" will run on ARM and demonstrated an early build of Windows 8 running on ARM-based chips from Nvidia, Qualcomm and Texas Instruments.]

But at this early stage, there are still a lot more questions than answers on Nvidia's Project Denver.

Topics: Hardware, Processors

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

18 comments
Log in or register to join the discussion
  • Hard a'starboard

    Windows on ARM ... NVIDIA GPU on ARM [one could say 'arm in arm' :-) ] ... not good alignments for INTEL.
    I'm enjoying CES this year!

    "First, while ARM processors have been growing significantly more powerful, they still can?t touch the fastest x86 processors."
    Neither can I use much of the-far-less-than-fastest x86 on my desktop ... and its too heavy to carry.

    "Second, many of the client and server applications used by organizations and individuals would need to be rewritten to run on ARM-based computers."
    At work we'll use remote desktop to the private cloud.
    At home we'll use remote destop or CITRIX receiver from a Chrome device to the Internet or our private cloud.
    M$ say they'll do Office on ARM ... what about ADOBE?

    Reducing problem anyway ... but yes, lots of work to do. Get on with it already!
    jacksonjohn
    • RE: CES 2011: Nvidia is developing its first CPU

      @johnfenjackson@...

      It will be interesting to see how Intel will respond.

      . . . and whether Intel will continue to push Atoms or perhaps make their own ARM CPU (they theoretically can - they have a license).

      One thing is for certain, though - this is gonna be an interesting year for CPUs.

      "M$ say they'll do Office on ARM ... what about ADOBE?"

      Considering that Flash is already available on several ARM platforms, it's safe to say that Adobe will be trying to put Flash in whatever new OSes come out.
      CobraA1
  • I'm betting nVidia will go massively parallel.

    While it's way too early to tell whether ARM is the way of the future or not (and how Intel will respond to all of this), this is an interesting turn.

    One thing I'm pretty sure of, though, is that it's very lilkely that it will have multiple cores, and if it's successful nVidia will likely push for many, many cores for a massively parallel CPU. They've already demonstrated that the company is very interested in massive parallelism, and I expect that they'll continue that trend if they try to make any CPUs.
    CobraA1
    • RE: CES 2011: Nvidia is developing its first CPU

      @CobraA1 <br><br>It might be interesting to see if Intel will open up the x86 architecture licencing program. After all, It is better to have a competitor that still has to pay you, then one that doesn't. Currently there are 2 companies that have an x86 licence, AMD and VIA.
      mgaul
    • RE: CES 2011: Nvidia is developing its first CPU

      @CobraA1

      That will be a safe bet; to match the processing power of x86 processors they will have to.

      The problem is that there are still too many applications that are not multi-threaded or optimised to run on multiple cores. A push has to be made to get developers on board to start making their applications better, this includes mobile application developers.
      pazmanpro
  • RE: CES 2011: Nvidia is developing its first CPU

    If the ARM CPU is running Linux or (soon) Windows, wouldn't an application just need to be recompiled, not rewritten? The only situation I can think of are things like multimedia encoding/decoding that might have hand-coded calls to SSE instructions.
    jgm@...
    • Generally no

      @jgm@... It depends if on how the application is written. If the code is all strictly portable ANSI C++ and the compilers are fully compatible (not all compilers support all of the ANSI standard) then it is may be doable. Non-portable C++ (assumes size of int or endianess or alignment) may require code changes. What if the code is VB6, VBA, C#, VB.net, delphi or other non-arm based language?
      DevGuy_z
  • Things are going to get exciting!

    There's also the news of the Qualcomm-Atheros merger and Qualcomm's promise "to move beyond mobile" with its SoCs. If VIA can ever get its crap together, we'll have some real competition in the space. Oh, yeah, and there's the XCore86 line of SoCs, too, which run the Edubook.
    daengbo
  • Sounds like the people designing ARM cpus are trying to catch up with x86

    technology as far as speed and processing power.

    If that's the case, why would anybody want to trade in an already blazing fast CPU for something new and which might be competing with the newer and improved and even faster and more powerful Intel processors that come out in the future?
    adornoe
    • RE: CES 2011: Nvidia is developing its first CPU

      @adornoe@... Arm platform is already fast. It's a matter of porting similar operating systems to do a realistic comparison. Between x86 and arm.
      QuimaxW
      • I'm aware that ARM is not new and that it's fast...

        I'm just commenting on the efforts to build ARM processors that could take on the x86 processors for speed and power.<br><br>The article above had the following statement:<br><br><b>First, while ARM processors have been growing significantly more powerful, they still cant touch the fastest x86 processors.</b><br><br>That is what I was referring to.
        adornoe
      • RE: CES 2011: Nvidia is developing its first CPU

        @QuimaxW Fast is relative. It isn't fast when compared to x86. It doesn't have the clock speed and it doesn't do as much per clock. x86 can generally do multiple machine code instructions per clock and I'm not talking threading or cores. And in some cases each instruction may accomplish more than what is possible on a RISC chip.

        So slower clock, one instruction per clock and more primitive instructions will keep ARM behind Intel on the performance side for some time. ARMs advantage is heat and power requirements are much less and cheaper. And don't even talk about 64 bit and Double Precision math.
        DevGuy_z
    • RE: CES 2011: Nvidia is developing its first CPU

      @adornoe@... <br><br>ARM is not a new technology. It was introduced in 1983, and the 8086 (aka x86) was released only 6 years prior. The major difference is the why the technology is licensed. ARM is more open to new licence (as long as you pay the royalties you can use it), whereas, x86 requires you to licence from Intel which can be approved, denied or revoked at Intel's discretion.
      mgaul
      • Read my response to QuimaxW

        I don't want to repeat myself.

        I'm not arguing about the merits of going with x86 technology or with ARM. The point I was making is that, the manufacturers of ARM processors are having to try to play catchup against the more powerful and faster x86 CPUs.
        adornoe
    • RE: CES 2011: Nvidia is developing its first CPU

      @adornoe@... ARM CPUs use less energy.
      david08048
      • That's besides the point.. With more power and speed, they'll consume more

        and that's what Nvidia and other are trying to do, that is, making the ARM processors more powerful and faster. With more capabilities and speed, they'll also be consuming more energy.
        adornoe
      • RE: CES 2011: Nvidia is developing its first CPU

        @david08048 And why? Fewer transistors, which means less capable. Start making em more capable and ramp up the clock speed and they'll suck the power.
        DevGuy_z
  • Very predictable move

    I predicted, and expected, some time ago, that Nvidia would make this move into consolidated, unified CPU architecture.

    It makes complete sense when we take into consideration AMD have somewhat leap-frogged competitors in CPU-GPU integrated conceptualization and design.

    It is almost inevitable now that Nvidia will move toward closing the existing divide that AMD-ATI created (now with ATI absorbed, simply AMD).

    These are truly interesting times in chip development.
    thx-1138_