Recent

% | $
Quotes you view appear here for quick access.

Intel Corporation Message Board

  • ltisteve@verizon.net ltisteve Oct 30, 2011 2:15 PM Flag

    ARM servers

    Okay, the old thread got a bit too long.

    It was a really good thread. What I can say is that you can't discount the ARM processor just because it emerged from the PDA world as a serious contender in the server world. But as Blue said, it's results that were first released are somewhat doubtful. They are segmenting the low performance server world. In other words, weak file servers.

    The R in ARM stands for RISC. The R stands for RISC. You can read more about it there, but I would like to call your attention to a section at the bottom.

    Diminishing benefits for desktops and servers
    Over time, improvements in chip fabrication techniques have improved performance exponentially, according to Moore's law, whereas architectural improvements have been comparatively small. Modern CISC implementations have implemented many of the performance improvements introduced by RISC, such as single-clock throughput of simple instructions. Compilers have also become more sophisticated, and are better able to exploit complex as well as simple instructions on CISC architectures, often carefully optimizing both instruction selection and instruction and data ordering in pipelines and caches. The RISC-CISC distinction has blurred significantly in practice.

    The point about RISC is that it takes a run at Intel once every ten years. There was a piont where it was clearly superior to Intel's offerings. It was so good that Microsoft introduced Windows NT just to have code to support RISC. It flopped,and six years later they rolled the code base back into Windows XP.

    So, the ten years is up, and from what I can tell this is the weakest of attempts at Intel I've seen. The metric of performance per watt is a good one. I agree it is important. Intel also agrees. They are working hard to maximize performance per watt across the board.

    The larger question is in the real world, how much power savings can you get on a processor that is perhaps is 20% more efficient (and that is arguable). Would you see a 5% electrical savings in your plant?

    ARM has security built into it's processors to some extent. Intel's is better. One cyber attack could cost a company far more than they save on electricity. There are also long term reliability issues. Anybody who buys this is buying generation one of a new system. You get to find all the bugs.

    The larger issue with ARM goes like this. If I gave you $20 million to make a Cola people prefer more than Pepsi, could you do it? Maybe, but lets say you can. Does this mean you'll take over Pepsi's empire? If I gave you Steve Job's six billion dollars and told you to knock them off the block, you probably still couldn't do it. You would be huge thorn in the side of Pepsi, but it would take decades to replace Pepsi's market share.

    Come on folks, were talking about cola here. People will buy more of it next week, and it doesn't have to be Pepsi.

    Unlike cola servers are deep rooted in their environments. Some of them have been in play for over a decade. A paradigm shift is unlikely in most organizations. The costs savings alone isn't worth the expense of switching, nor the risc (punny!) of problems.

    None-the-less ARM may gain some ground in servers. I don't expect it to sell like hot cakes, but they may find a few good applications. I don't know how viable it will be for the long run.

    I do welcome good competition. But just the same, Intel is going after ARM market in phones and tablets while ARM is going after Intel's market of servers and low end laptops. Who do you think could inflict more damage on the opposing side? Due to fabrication and marketing muscle I am thinking Intel has more to gain than lose.

    SortNewest  |  Oldest  |  Most Replied Expand all replies
    • I dont think it's a case of Intel OR ARM. They have yet to cross swords in any meaningful way.
      -

      Very good insight. That contest is yet to come. These are two wonderful companies a factor of 50 diferent in size and totally different business models.
      From an investment (or in my case, gamble) standpoint, I look at Intel as being significantly underpriced today and ARMH being significantly overpriced today.
      I can buy out-of-the-money Intel calls for 2013 for 1% of the stock price. There are no options on ARMH with the same leverage. And, of course, I believe the manufacturing expertise of Intel will ultimately win the day.
      ARM might have something of an architectural advantage, but all the X86 stuff is more RISC than CISC now anyway.
      From a battery operated user's point of view power is paramount. As long as function, price, delivery is ok, processor architechure is relatively unimportant.
      Given that most of the power dissipation of a wireless device is due to transistor off state leakage, I feel the TriGate process will far more tha offset any architectural advantage that ARM might have.
      I also believe that planar process scaling below 30nm will prove to be much more difficult than expected and that performance might decrease and power might go up on sub 30nm processes.
      I further think that Intel will beat TSMC and others in process capability for the next 2-3 years....and the next 2-3 years will be critical in determining the winners for the next decade.

      So, Intel call options seem to be an inexpensive, highly favorable, risk controlled play on the contest that will develop during the next year.

    • What bad news???

      ARM's complete void in fabrication after 20nm.

      That's the bad news.

      And no diversion over servers that may or may not arrive years in the future is going to change that.

      As far as ARM being conservative, don't make me laugh. Saying Intel is a niche player isn't being conservative, it's the worst form of grandstanding.

      And if ARM wasn't concerned, they wouldn't be sending fanboys to the Intel message board to plead their case...
      -----

      :) Fanboy eh? You are right I do know more about ARM than most on this board...

      I have been open about my view: I dont think it's a case of Intel OR ARM. They have yet to cross swords in any meaningful way.

      ARM are a conservative firm...Something is afoot...You could be right and it's nothing but grandstanding. We shall see.

      As for fabrication I have attempted to explain why the ARM world isn't that 'bothered' about tri-gate and the rush to ever smaller nodes by Intel (different transistors/processes for different markets). Regardless of the advantage of Intel's transistor (or not as the case may be) bad CPU design is still a major problem (Atom) which isn't going to be addressed until Atom @ 22nm.

    • Yours is a good read. IMHO, INTC is under valued.

    • No reason to doubt that ARM can come with 64 bit server processors soon. But the claim that it will displace Intel processors goes too far and is a sign of mobile success (accidental) and favorable media going into its head. Forgot that smartphones and tablets are not powered by ARM alone but there are hidden servers behind these devices? If everything was to be done by onboard ARM chips, would those be as power efficient as hyped?
      -----

      Not sure where you get the idea that there are 'hidden' servers doing the heavy lifting in an ARM device? Next time you are playing a game or an app, disconnect from the net and you'll see no difference in performance. Now, if you are expecting your web browser still to work...:)

    • Talking about it is different than doing it. Two years can easily turn to 5 or 6 to never because they have no control in what they are doing. Trigate will be out by that time.

    • ARM is talking about taking two years to fill a gap that won't be there in two years...

    • ltisteve@verizon.net ltisteve Oct 31, 2011 1:31 PM Flag

      Intel is about to go after ARM's market, so ARM has to look strong and appear to be going after Intel's market at the same time. I know that ARM is somewhat conservative, but they also do a bit of grandstanding.

      I don't know how big of a market the multi node processor servers are. I'm sure there is some demand for it. As far as I can tell ARM will be going after specialized segments of the market as will Intel with it's 50 core processors.

      The other segment that may get people's attention are companies that want to project a green image. Apple for example would install these and talk about them in their annual report, but getting exact numbers may be hard. Would they use 5 or ten thousand?

      There is plenty of business out there for ARM and Intel. But, just the same I won't lose sleep over what ARM could do to Intel.

    • "What bad news is this? Could be a publicity stunt, but would be unusual for ARM as they are so conservative."

      ***

      What bad news???

      ARM's complete void in fabrication after 20nm.

      That's the bad news.

      And no diversion over servers that may or may not arrive years in the future is going to change that.

      As far as ARM being conservative, don't make me laugh. Saying Intel is a niche player isn't being conservative, it's the worst form of grandstanding.

      And if ARM wasn't concerned, they wouldn't be sending fanboys to the Intel message board to plead their case...

    • No reason to doubt that ARM can come with 64 bit server processors soon. But the claim that it will displace Intel processors goes too far and is a sign of mobile success (accidental) and favorable media going into its head. Forgot that smartphones and tablets are not powered by ARM alone but there are hidden servers behind these devices? If everything was to be done by onboard ARM chips, would those be as power efficient as hyped?

    • Smartphones and tablets need sevrer power to bring services to the screens with embeded SOCs just acting as display cards and sound cards. I am wondering if the so called energy efficient servers will actually need more servers to function.

    • View More Messages
 
INTC
31.57+0.08(+0.25%)May 27 4:00 PMEDT