Spectacular Growth Predicted in Smart-thingies

Tablets and smart-phones are predicted to have huge growth in the coming years.

  • smartphones to 800 million per annum by 2013
  • tablets passing netbook shipments this year
  • smartphones to 440 million in 2011

If people carrying smart-phones develop a fondness for tablets, this could change everything… These numbers mean the monopoly is dead by 2013. Very few of tablets or smart-phones use Wintel. Many of them use GNU/Linux, particularly Android and Chrome OS will surely have a big play. The only crevices in which that other OS can thrive are the vestiges of one licence per PC and the cloud. The one licence per PC thing should be dead next year at this rate. ARMed PCs will appear for sure. The cloud is GNU/Linux territory. Performance is all that matters there. DRM, phoning home and re-re-reboots are a thing of the past.

During this phase of rapid growth, prices will remain a bit too high for my liking but we are already seeing healthy competition in small cheap computers running ARM so I am not worried that price will cause the small computer to be uncompetitive. Head to head against Wintel, they will do very well. You don’t need huge drives and cases and power-supplies with ARM so the price advantage will be leveraged.

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in technology. Bookmark the permalink.

6 Responses to Spectacular Growth Predicted in Smart-thingies

  1. oldman says:

    The real issue with ARM is that they have zero track record building anything beyond CPU’s suitable for use in embedded systems (smart phones, tablets, thin client terminals). There isnt even a 64 bit version of the design. and the (IMHO) kludge design that they have announced for addressing more than 4Gb of RAM is itself several years off.

    The only RISC design with any legs is IBM’s power architecture, but this chip is a power drawing monster (as were all proven server class RISC designs) that will only see the light of day in enterprise class server hardware.

    Although at just over $6K the Power 710 Express IS a very interesting box ;-)….

  2. Ray says:

    I was talking about undervolting the processor, and lowering the clock rate :).

  3. No matter what you do with the clock on x86 it stills draws a lot of power and flips more bits per second than ARM. At one point ARM was static so you could under-clock without limit. For a long time x86 has been dynamic with a fairly narrow range of acceptable clock-speeds. For the light loads that most CPUs have, ARM is just a better way to go.

    Today, the CPU on my terminal server never got above 10% utilization since I compiled a kernel this morning. ARM at 300MHz would have done the job well. One of these quad-core jobs at 1.2gHz would probably do on a busy day.

  4. Ray says:

    Agreed.

    If CPUs and RAM could be dynamically adjusted to provide performance, while still powering down a bit so save energy, no one would have this RISC vs CISC argument. 🙂

  5. … The last x86-64 processor I bought was quad-core and deliver 90W or more. That thing had thousands of millions of transistors on it clocking over at 2+gHz. RISC might have a few million transistors.

    There are numbers for DMIP/W:

    see Slashdot

    The most heavily loaded system here is my GNU/Linux terminal server. Even with 20 students hanging all over it, the CPU usage is 20-30%. ie. 75% of all the power wasted on that x86 CPU is truly wasted. The thing is idling most of the time. I should run some charts.

    vendor_id : GenuineIntel
    cpu family : 15
    model : 2
    model name : Intel(R) Xeon(TM) CPU 2.80GHz
    stepping : 9
    cpu MHz : 2800.312
    cache size : 512 KB

    That’s something like 5600 MIPS but 3/4 are wasted leaving 1400 MIPS something like what a modern ARM CPU can do in a watt or two. Most of the RAM is clogged with user data. I only have 1gB on that machine. I don’t see the load on fast or slow storage being a matter because it will be similar for both. Would “7” run 20 users in a gigabyte?

    It is much more efficient to have your bits/CPUs etc. working hard rather than idling. Having fewer transistors/less microcode being clocked means the ARM chip must work harder more of the time, and being more efficient. There is a point at which the power consumption of RAM will pass that of the ARM CPU. ie. You could make the ARM consume 0 watts and the system would be no more efficient but x86 generally uses more power in the CPU than the RAM and the ARM CPU uses a similar amount because the RAM uses dynamic memory and clockspeeds have been going up over time. For the small units like smart-phones ARM easily wins the comparison. In a server or desktop you may have a point. I keep seeing those hair-drier heat sinks, though… None of my RAM has a fan or radiator yet.

  6. Ray says:

    I don’t think RISC type processors (ARM, PowerPC, etc) are faster per watt than CISC types (x86, 68k, z80). You see, in RISC, even though is simpler to implement, it uses more hard drive space, more memory space, more address space per say, programs, leaving everything else as the bottleneck.While on the other hand, even though CISC is more complex, it puts less pressure and load onto everything else, as it uses less address space per program, leaving the CPU as the bottleneck.

Leave a Reply