ARM: You think our servers are good now? Just you wait

ARM: You think our servers are good now? Just you wait

Summary: During its earnings presentation Cambridge chip designer ARM expressed confidence in future servers based on its technology, alluding to new designs that will be integrated

TOPICS: ARM, Data Centers

ARM is confident that servers based on its chips will get ever more capable as partners begin using the company's latest chips designed for enterprise workloads, rather than for mobile phones.

During its earnings presentation in London on Wednesday ARM chief executive Warren East noted that the leader of the server pack for ARM chips, Calxeda, has based its technology on a mobile-focused design.

ARM A15 chip
ARM believes future servers based on its technology will be even better than those at present. Image credit: ARM

"[But] bear in mind this first Calxeda silicon is Cortex A9-based — a core we developed very much with mobile in mind," East said. "When you put the server infrastructure around a core designed with servers in mind -– Cortex A15 or even ARMv8 — then you can see more [performance]."

Calxeda published figures in June that demonstrated its ARM servers had a 15X performance per watt advantage according to the ApacheBench benchmark, when compared with Intel Xeon E3 workloads.

"We're very pleased with the [Calxeda] progress," East said. "The data that's coming out suggests that all the experiments we did before and all the simulation has been proven in silicon."

Confidence in networks

Along with this, ARM also noted in its earnings that it had seen more growth in the use of its chips in networking products, such as Freescale's.

"I'm very encouraged by the reception that our customers' customers are providing to ARM technology in this space," East said. "I don't see why we can't have a very large share of this space.

"The power efficiency benefits of ARM.... apply just as well to this networking space as they do to mobile computing."

ARM's chief financial officer, Tim Score, increased the company's expected profit range from around £60m to around £65m for the next quarter. ARM reported pre-tax profits of £66.5m for the quarter.

Topics: ARM, Data Centers

Jack Clark

About Jack Clark

Currently a reporter for ZDNet UK, I previously worked as a technology researcher and reporter for a London-based news agency.

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
  • Not surprising - ARM is just superb

    Back in the early '90s I did a comparison between an ARM running at 8MHz and an Intel 80486 running at 33MHz. In repeatedly calculating the sine of a gradually increasing angle, the ARM took 1.61 seconds and the '486 took 1.59 seconds. i.e. the '486 had to run at over 4 times the speed of the ARM to beat it in a real processing task. As both Intel and ARM have improved their processors over the years, I'm not at all surprised by the x15 figure quoted in the article. One point not specifically mentioned is that for a given performance, the energy consumption of ARM based servers will be a fraction of that of Intel based servers. This has several advantages:
    (1) reduced running costs, (2) less cooling required so again reduced costs, (3) cooler running resulting in longer life.
    I'm looking forward to Microsoft providing a version of Windows for ARM-based laptops, desktops and servers. Once that happens - and it's on the cards now with Windows RT being its first incarnation, I predict a huge increase in ARM-based computers of all sorts, but especially laptops with the seriously increased battery life.
    And who knows, maybe even Apple might migrate to the ARM - after all, they have changed processors before.
    • Google servers

      I wonder if the likes of Google and Facebook have invested in developing their own Arm base servers.
  • I dont see this taking off

    While ARM servesr will be of low power, they will also lack in processing power. ARM chips are nowhere near as fast as x86 ones. It would be a while before we see any ARM traction in the server marketplace.

    These two companies are on the short end of the technology stick.

    The consumer market has moved to ARM and soon the business market will too.

    Not only does this spell trouble for Intel, but once on ARM, the enterprise can finally ditch Windows on the backend, for a more secure and reliable Linux Distro. They can still keep a legacy environment to run Windows apps, until their processes are re-written for mobile.

    The front end wont require Windows either, as the cloud will relegate the OS to being agnostic, accepting either desktop or mobile OS access.
    • Your drunk.

      No, you have to be absolutly plastered.

      Or maybe just joking.

      It makes me giggle like a child when I hear these morons sitting around predicting the demise of of Microsoft. Its not going to happen. Get it through your head.

      Your living in dreamland extraordinaire.

      Every other day for years on end people have been predicting the fall of Microsoft and often running hand in hand with that is the prediction of the rise of Linux.

      The kind of changes in the IT world you are talking about are earth shaking. Even if they were to happen they would be monumental shifts in the IT landscape and would occur about as slowly in the IT world as the rollback of the ice age occurred in the natural world. Things may happen lightning fast in the IT world, but the point is, the kind of changes you are talking about would take many many years to draw to a conclusion.

      And thats only if Microsoft and Intel just sat around for all those years and used none of their vast resources and monumental wealth to work their way through such problems once they seen the writing on the wall.

      For those who love Linux they always see the sky falling for Microsoft. What they fail to see is the following. Windows is a bit like the modern day airplane in a certain respect. The airplane has never been perfect. They occasionally crash. When they do, it can be in a most unfortunate and spectacular way that will always generate lots of criticism. One day the airplane as we know it today may in fact be replaced by something much safer and even more reliable. Right now, despite the many modes of transport at our disposal, that replacement product just does not exist, and nobody is seriously considering scrapping air travel. Thats because despite the horrible crashes that do happen, they are statistically few given the vast number of flights happening every day around the world and the fact is that in the long run airplanes are generally very very good at what they do and there isn’t anything that is up to the kind of job they can do.

      Im not going to sit here and criticize Linux because I did use Linux for awhile severall years back and was very pleasantly surprised at how good it was. The problem is, and will always be, unless Linux adopts a more Windows like way of doing things (I seriously doubt it) Windows is a better choice for very active computer users like the average person. Linux can be made to work for very light computer users; you know, email, chat and some web surfing and photo storage, and Linux can be a fascinating OS for genuine savvy geek types who really want a system that gives them various kinds of control and they have the ability to more easily work through some of the more intensive and laborious parts of Linux that come into play from time to time. But not the average person who is in between those two kinds of users.
    • Why would a large company trust

      the cloud that much?
      Michael Alan Goff
  • Qualcomm has been

    irritated at Microsoft ever since, well it wasn't pretty! When you say ARM, you mean Qualcomm. Take out Intel, the easier mark, and you take out Windows too...
    Tony Burzio
  • Compilers

    Is there anyone here with serious Computer Science creds who knows whether they ever solved the compiler puzzles? The promise of RISC hinged a lot on moving intelligence from the processor (mostly the microcode) into the compiler. That way the chip could be simple, cheap, and low-power, and still deliver comparable performance. Trouble was, the compilers never got smart enough to fulfill the dream. Maybe that's fixed now. I don't know.
    Robert Hahn
    • Compilers for RISC Arcitecture

      Compilers have been "smart enough to fulfill the dream" since the mid-sixties. My department in the Los Gatos ASDD Lab simulated IBMs first RISC architecture in 1967. We had optimizing compilers developed at the IBM research labs in Yorktown. I developed compilers (Fortran & COBAL) for the System/360 in 1964. The original S/360 (1964) was close to a RISC architecture.
      • Certainly you meant COBOL instead of COBAL?

        I remember Formula Translating System (Fortran) from IBM as well as COBOL COmmon Business-Oriented Language.

        The design goal of RISC (Reduced Instruction Set Computer) was to achieve 1 clock cycle per operation. Back then CISC (Complex Instruction Set Computers) such as x86 could take 50+ clock cycles (or more) to conclude one single lengthy operation.

        However this was then. Nowadays, all modern CPUs include both RISC and CISC design philosophies and thanks to multiple-pipelines, predictive branching, Out of Order (OoO) technologies just to mention a few; can now easily relinquish that original design goal of RISC to the trash can since CISC can now do up to five (or more) simultaneous operations per single clock cycle.

        The only advantage RISC has over high-performance contemporary CPUs is simplicity (Reduced Instruction Set Computer) and low-power consumption with corresponding low computational performance and low cost.

        Perfect for battery operated devices. Basically RISC was a mere FAD.

        There is nothing new under the sun, but there are lots of old things we don't know.
        ~ Ambrose Bierce

        Every generation laughs at the old fashions, but follows religiously the new.
        ~ Henry David Thoreau
        • CISC gobbles power

          Trouble with multiple pipelines, predictive branching, out of order etc. is that as soon as they fail to keep the processor fully loaded they are consuming a lot of power for no useful work.

          One of the reasons for Win RT not running existing software is that most older software is not written with power consumption in mind, and since Win RT is intended for tablets, it needs to be more power efficient. Windows 8 introduces several more power modes and asychronous process handling to cut power consumption.

    iOS and Android obviously compile for ARM just fine. It's funny you mentioned that though, because Microsoft is having problems getting RT to run on ARM:

    IOS is compiled from Ojective C
    Android is compiled from Java, with a hokey C wrapper (which is why it's laggy)
    Windows is written in C
    In theory, Windows RT will be a direct port to ARM

    Windows 8 Intel will call Win32 APIs
    Windows RT on ARM will call WinRT APIs (hence its non- support for legacy apps)

    I'm not sure the RT compilation errors are from the Windows X86 port, or from the wrapper Microsoft used so it will call WinRT APIs.

    In any event the ARM compile is not working out so well for Microsoft.
    • Wow, a single article

      that links to another single article where people were having trouble IN MAY when the OS was still in beta. Yep, rock solid evidence.
      Michael Alan Goff
  • Ask Yourself This Question, Michael...

    When was the last time you heard anything about Windows RT?
    • Me personally?

      Not in a while, I'm not the ones they should be caring about anyway. I mean... I probably won't be getting a Windows RT tablet. But I did Google it, and the first article about it (a positive one) was from July 16 saying something about Nvidia defending Windows RT.

      Apparently, they're really adamant about how Windows RT will be great. And I think they knew more than a source from the beta.
      Michael Alan Goff
  • One More Thought...

    @Robert Hahn

    Microsoft is trying to compile RT for ARM on 2 different chipsets (Microsoft always has to do things the hard way); with those chipsets being Qualcomm and Nvidia.

    All I've heard thus far is that it's NOT going well; with conflicting reports about which company is further along.

    My guess is that Microsoft stripped down (cleaned up the code) a version of Windows 8, re-wrote it and then used some kind of either C++ or Objective C wrapper (which would explain why Qualcomm is further along; as their chips are used by Apple).

    The real problems lie in the dirty MS Windows code.

    If MS can't get this to work, and from an engineering standpoint I highly doubt it will run smoothly (if at all...hence the Surface demo fiasco), then Microsoft's mobile future is DEAD IN THE WATER!
  • The opportunity is lost

    Intel Medfield seems to be a total equal to ARM.
    I wouldn't have believed that couple of years ago.

    It it far to much simplification to just use the word server as a general meaning. There is a vast difference for a number churning server, or one used a a file server or a database. The rest of the chips are far more important in the latter case and cpu power far less.