7 Comments

Summary:

AMD, which has fallen behind its chief rival Intel in the x86 processor business, announced on Monday plans to make new 64-bit chips based on ARM’s chip technology that will target data center and cloud computing companies. AMD will continue to make x86 processors as well.

AMD HQ
photo: AMD

AMD will license the ARM chip technology as part of a strategy that will bring cell phone chips into its servers. The company on Monday announced that it will design 64-bit ARM technology-based processors in addition to its x86 processors for multiple markets — hoping to cater to the needs of data center and cloud-centric companies looking for low power computing.

The move has been debated within AMD for some time, and represents AMD’s embrace of a heterogeneous computing strategy. The news also shows how AMD is distancing itself from its fellow x86 rival, Intel, and in reality, could prove to be AMD’s best chance to continue on as player in the chip market.

At a press conference on Monday AMD CEO Rory Read said, “Modern cloud is the killer app and it is bringing about the fastest growth across the industry.” He is convinced that ARM and AMD “can change the server and data center landscape.”

In a press statement to accompany the news conference, Reed added:

“Through our collaboration with ARM, we are building on AMD’s rich IP portfolio, including our deep 64-bit processor knowledge and industry-leading AMDSeaMicro Freedom supercompute fabric, to offer the most flexible and complete processing solutions for the modern data center.”

ARM, through its various partnerships, has been slowly gnawing at Intel’s dominance of the chip business, thanks in part to the booming demand for smartphones and other such devices. ARM has been lusting to take a piece of the server business, giving Intel more headaches. AMD, is a perfect partner for such an assault. The chips are likely to be made available in 2014, according to AMD executives.

We’ve been anticipating this move for some time, ever since AMD purchased SeaMicro, a startup that builds ultra-dense low-power servers for cloud computing that use Intel’s low-power Atom chips. SeaMicro uses x86-based chips for its boxes, but it has a technology that enables it to use any type of processor, including ARM-based cores.

The transition to alternative forms of computing in the data center has come about in some market segments, because certain jobs need less computing horsepower to complete their tasks and data center operators are looking for the most energy-efficient processor for the job. Just like you might not take your 12-cylinder Lamborghini to the grocery store to pick up a gallon of milk, the data center guys are increasingly seeing high-end general purpose CPUs as appropriate for some tasks, but overkill for others.

ARM has seen the opportunity for so-called wimpy cores, and has invested in Calxeda, a systems maker that is building a new type of servers using ARM-based SoCs. Dell, HP and others are also getting in on the ARM-server market with new products using chips from Calxeda, Marvell, Applied Micro, and perhaps even Cavium. Now that AMD has jumped on the bandwagon and with ARM servers in production later this year, getting ARM into the data center is looking more and more likely. Your move Intel.

Additional reporting by Om Malik.

  1. Stacey, good work as always. I’d be interested in hearing more of your thoughts on how ‘competitive’ this move truly is given Intel’s work with Xeon Phi / MIC.

    Share
    1. I’m waiting for my Phi briefing actually :) Intel’s been pushing that for HPC and I think of HPC having a need for fast IO and brawny cores, which is a different market than AMD/ARM (although not APUs).

      Share
  2. You forgot Nvidia,they are very much into ARM servers too,just ,maybe, more under the radar for now but they are one of the few players with their own GPU and that can be an asset( not to mention that they are investing in a custom core).

    Share
    1. You are right. I did not mention Nvidia although I don’t recall if they are doing something with ARM in servers. I’m aware of Tegra on the handset and Kepler using GPUs for servers. Did I miss something?

      Share
      1. Gentleman Trader Monday, October 29, 2012

        project denver is their computing platform, not sure if its targeting servers though

        Share
  3. I hope they start using a fixed continuous current like 24V to be easily connected to a CC no-break.

    Share
  4. Are they building it on their own or planning to acquire some firm doing this?

    Share

Comments have been disabled for this post