Google could give ARM big data center boost

Google could give ARM big data center boost

Summary: Here's a look at the fallout should Google go with custom ARM chips in its data centers.


Google is reportedly looking at ARM-based processors for the custom servers that power its data center and cloud.

According to Bloomberg, Google may design its own server processors using ARM technology. If Google follows through it would indicate that the search giant plans a larger rollout. It's unlikely that Google would manufacture ARM chips and then stick with Intel for most processors.

There's an impact on Intel's revenue stream for sure. Bloomberg said that Google is Intel's fifth largest customer.

Here's a look at fallout should Google go ARM.

  • ARM gets data center credibility. ARM executives are very clear about wanting more data center market share. The issue is that gaining share will take time.
  • White-box server growth continues. Google is only going to need more servers so what helps its data centers are going to boost contract manufacturers. Google would likely buy both ARM and Intel servers for a while.
  • Intel takes a hit---maybe. Intel could lose its standing to some degree, but the chip giant would need to be hit by far more than Google to lose significant share. Enterprises would have to go ARM en masse. Another wrinkle is that Intel could theoretically become the foundry for Google's ARM chips. 
  • Enterprises evaluate ARM more. Google would be an excellent proof point that would likely lead to more adoption of efforts like HP's Moonshot server line.

More reading:

Topics: Data Centers, Google, Processors, ARM

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


Log in or register to join the discussion
  • Understanding Microsoft's RT Play

    MS's play into ARMs and Windows RT begins to make sense when you see the market movement. Windows RT is build from the Windows NT core and everything learned from RT can be applied to running Windows Server on ARMs as well. And having a ARMs play hedges MS's bets on future technology and keeps a foot where Android and iOS are.
    Rann Xeroxx
    • But Windows RT failed

      It is now indisputable that Microsoft's Windows RT is dead in the water.

      It's gone.

      This leaves Microsoft stuck in the Intel x86 world, and vulnerable. If Google gives a boost to ARM processors, then Google's competitor Microsoft loses.

      That's another big incentive for Google to support ARM.
      • Being "stuck" in the x86 platform ain't a bad thing, especially considering

        that it's still the superior platform or CPU, and will get better with time, while ARM will be forever stuck in catch-up mode.

        Heck, it's the ARM platform that is beginning to worry, since the x86 platform is as energy efficient as ARM cpus, and is much faster, and already matching ARM cpus with prices.

        Even non-MS tablets are being made with Intel's architecture. Google might be looking for diversification in their data platform architecture, but not because they'll be saving much in energy or price, and they will, for sure, not be gaining any speed.

        Having said that, I don't fault any company for diversifying their products' architecture, or their in-house computing platforms. That's why RT, or Surface 2, will still be around for a while.
  • Not everything... The device I/O would be different.

    One thing a server needs is decent I/O capability. The existing ARM architectures do not have it.

    It needs to support 100+MB/sec I/O, and support that from multiple controllers.

    One chip implementations aren't going to be able to do that. Having memory external to the CPU allows for different paths to the memory from what the CPU uses - allowing for a more flexible and faster I/O capability. Each controller could have its own port into memory (multi-path).

    Server systems also need larger memory than a 32 bit data path allows. Again, having memory external to the CPU allows for NUMA architectures that CAN handle the bandwidth, and available memory. Current ARM designs don't support this either.
    • While I agree with you

      on this to some extent, I differ from saying ARM is not suitable as Server processor. They could be clustered and scaled to do single serving functionality like web pages, file and document services etc. Or they could just do serving ads to one subset of one particular demographics. There are so many. The servers can be hosted on big cluster server powered by better CISC processors like Intel x64 etc.
      Ram U
      • It becomes inefficient to do that.

        It is one of the reasons associative memory processors failed.

        Like other servers, it becomes more efficient to use one system - like the IBM z machines that can host 1500 such items for half the cost, and use only 4 CPUs.
        • Now that said...

          The ARM architecture is better suited to large systems than the intel architecture is (due to the multi-layer implementations needed to get it fast - 1. base RISC processor, 2 - reordering optimizer, 3- translator for IA instruction set into RISC base, 4 - cache processing... all three levels of it... to finally get to a single path to memory.

          ARM has a better instruction set that doesn't need all the complexity. Now if Intel dropped the x86 required layers...
    • Arm already has a 64 BIT data/address paths to memory.

      It also supports PCI-e with bandwidths exceeding 100MB/sec. Where Arm lacks is extensive 3rd party support and highly tuned drivers.
      • What ARM 64 lacks is external memory.

        ONE PCI E connection sucks for data throughput.

        And you can't put the CPU + 32GB of memory in one chip.

        Many servers have 8-64 CPU cores, and 32-128GB of memory (if not 256GB).

        For scaling purposes... putting it all in one chip sucks big time.
  • Web 2.0 competitors Facebook, Amazon, Baidu are working with AMD

    Interestingly, on the same day that the Google rumor was floated one of AMD's employees issued this:
    If Google started today, they could homebrew their own server chip in 3 to 5 years.
    What the rumor concerns is Google is threatening Intel, that they are considering working with AMD to create a Semi-custom chip designed specifically for their data centers.
    >>>” According to Feldman, web giants are looking at co-developing ARM-based chips that will take advantage of the greater levels of customization offered outside of the CPU so they can optimize for their own applications’ needs.”>"ARM chips power most of the world’s smartphones, but many vendors in the data center IT food-chain are working around the clock to crank out servers, storage and networking devices with low-cost and low-power ARM-based SoCs instead of the expensive network silicon or powerful and even more expensive x86 processors.">"“Avoton is better than the existing Atom – better than Centerton – but not a big-enough step forward at all,” he says. “Remember, I’m the only guy with any experience actually building servers with Atom parts.” SeaMicro has been – and continues to do so today as part of AMD – building servers using both AMD and Intel chips.">". “One of the ironies here is that in the ARM space, there are a ton of people who have made ARM SoCs, Qualcomm and Samsung being the best, without question,” Feldman says. “Those guys are monsters at making ARM SoCs.” These are companies that manufacture ARM processors inside the bulk of the world’s smartphones. Apple designs the ARM processors inside iPhones but uses Samsung as manufacturer. None of them, however, have made an ARM server before, Feldman says.">"“While others have more experience with ARM SoCs, none of them have an iota – I mean single. Shipped. Part. Ever – of the 30 IP blocks needed to build a server">"That’s the differentiation that we bring as AMD,” he says. You do not get that kind of experience by building it right once, he adds. It’s also not a one-chip race. “If you want to get in the server race, you need a new chip every 15 months for the next 10 years. And I don’t think anybody thinks Applied Micro can do that.” Applied Micro representatives did not respond to a request for comment.">"This is important because it means companies can make custom chips for big internet companies, who are expected to be the first major microserver users. “What that means is we can do a custom one for Facebook; we can do a custom one for Amazon; we can do a custom one for Baidu – each individually and separately – that has their IP in it, that embed advantages [for] their particular technology,” Feldman says.">"He expects ARM to win not because the end products will have slight advantages on key metrics. “They will,” he reassures us. “They win because it’s a fundamentally different approach to delivering a product.” ARM Holdings amortizes its R&D over its 250-300 licensees, while Intel amortizes its R&D over itself, and that is what makes it so much fun to watch, Feldman says.">"The reasons the big internet companies will be the first to adopt microservers, he explains, are that their businesses are extremely sensitive to cost of computing, and that they write their own software and have no allegiance to an instruction set. “They’re writing in PHP. They absolutely don’t give a shit about special instructions or optimizing. They’re writing at a layer of abstraction so high, that they don’t even know what the CPU’s doing.”
    The big internet companies with massive data centers “want an ecosystem’s worth of alternatives,” and that ecosystem is bound to develop very quickly, Feldman says. “Right now, you’ve got fundamental changes. You’ve got demand going through the roof on one side. You’ve got customers being smart and innovative on the other. You’ve got a threat from the ecosystem against an installed giant. A lot of things happening, and that’s cool.”>"Facebook's monster PHP engine ready to muscle into ARM server chips"
    • Actually, I think it would only take 2 years.

      As a server chip is actually simpler than a SOC.

      1 year for prototype CPU chips.
      Another 6 months for prototype motherboards... then production chips+motherboards by the end of the second year with quantity available in the third.

      The big question would be how many cores on that chip... Four is common now, 8 would be a possiblity... but if they went with 16 then it would be a stretch to finish in even 3 years.

      But it would be a screamer for speed.
  • AMAZON is clearly working with AMD

    ARM will take a substantially larger portion of the server market in a few years than most people understand.

    Please watch this video:

    Amazon's James Hamilton: Why Innovation WinsAMD SeaMicro

    What was so innovative about Tesla is that Elon Musk chose to string together numerous low cost high volume lithium batteries into a large powerful battery.

    Almost all the initial competitors chose to build ONE large new battery.

    Tesla and SeaMicro's approach is how technology has progressed in the IT industry.

    Instead of a huge mainframe, the next wave was smaller computers from Digital Equipment.

    But, then servers and PCs replaced the DEC approach.

    Servers used extremely powerful, expensive chips from numerous companies like Sun's Sparc.

    Intel and MSFT were able to create a lower cost server solution platform, even though Intel's X86 chips were NOT as powerful as Sun's Sparc, and IBM's Power server chips.

    NOW, AMD's SeaMicro is pioneering the next breakthrough in Server technology through the use of lower power, lower cost ARM chips.

    AMD will start with a few HUGE cloud companies.

    It will take a few years for the full ARM server ecosystem to grow out for smaller players, but the HUGE players are on track to be consuming HALF of all the server chips sold.

    The switch to ARM server chips will shadow the HUGE GROWTH in the cloud computing over the next few years.

    If you are Amazon, Facebook, Baidu, etc, who do you want to partner with for a key infrastructure technology?

    A company that has huge experience, billions of revenues, and is now profitable, that has a DISRUPTIVE technology, such as the SeaMicro architecture, or some small company like AMCC, or Calxeda?

    {“While others have more experience with ARM SoCs, none of them have an iota – I mean single. Shipped. Part. Ever – of the 30 IP blocks needed to build a server.

    That’s the differentiation that we bring as AMD,” he says. You do not get that kind of experience by building it right once, he adds. It’s also not a one-chip race. “If you want to get in the server race, you need a new chip every 15 months for the next 10 years
    NO way that Google alone has the team in place to create successful 64 bit server chip, even if they license from ARM. It would not surprise me that Google could be floating the idea of "rolling their own" in their negotiations with Intel, but they would need an extensive design team, more than just a few ASIC engineers.

    But, ASIC engineers are key to optimizing AMD's SeaMicro dense servers for specific uses.

    AMD's server VP Andy Feldman has been hinting for quite some time that AMD is working with key Cloud companies.

    AMD has said publicly that they are using their semi-custom model in the server area, just as they did with MSFT and Sony in the game console area.

    AMD created the AMD 64, which was so successful compared to Intel's Itanium 64 bit server chip, that Intel forced AMD to license the design to them. AMD 64 has been the basic design for most all of Intel's server chips.
  • Apple leading the way!

    Apple is clearly leading the way, with 64-bit ARM designs* already in consumers' hands by the millions.

    *Customized in-house.
    • Those chips aren't exactly suitable for servers.

      Memory is way too limited.

      IO bandwidth is also too limited.

      It would have to be redone for servers, so the current chips don't count except for a proof-of-concept. And that was already done by ARM itself.
      • What do you expect?

        For an ARM that's tailored to battery operation, Apple's A7 has "only" 9 GB per second streaming memory bandwidth, and servers perform best when all relevant data are cached. ARM is just a design house and the 64-bit A7 is way beyond proof of concept. You can bet Apple is already sampling higher performance A-series chips. Apple has been leaning heavily in this direction since it bought PA Semi in early 2008.
      • In case you missed it...

        Apple shocked the chip world with the 64-bit A7, just like it shocked the world with the iPhone.

        Great for code optimization, 64-bit ARM has twice as many integer and floating point registers as Intel/AMD X64. I don't see how dumping X86 compatibility from current X64 chips will help those architectures.