Performance analysis laughable

Article: Medfield, Intel's x86 Phone Chip
By: Wilco (Wilco.Dijkstra.delete@this.ntlworld.com), January 24, 2012 8:21 pm
Room: Moderated Discussions
David Kanter (dkanter@realworldtech.com) on 1/24/12 wrote:
---------------------------
>Wilco (Wilco.Dijkstra@ntlworld.com) on 1/24/12 wrote:
>---------------------------
>>David Kanter (dkanter@realworldtech.com) on 1/23/12 wrote:
>>---------------------------
>>
>>"Realistically, Medfield will not have a decisive performance advantage over platforms
>>like TI’s OMAP5 or the Snapdragon S4. At best, Medfield will be slightly ahead of
>>the competition; but in many cases Intel’s performance may >lag by 10-30%."
>
>Actually, I updated the article...the new language you might find more reasonable.

The new version is more accurate, but still suggests that Medfield could be faster than an A15 or Krait. There is no way Medfield would be close, whether you compare at max frequency, equal frequency or equal watts. Unless of course you're talking about Intel marketing numbers, not actual products.

>>I can't believe how anyone could seriously suggest that Medfield will be competitive
>>or even have a performance advantage over 1.5-2GHz dual core Krait or A15, especially
>>given the fact that a 1GHz A9 outperforms an 1.6GHz Atom.
>
>We've been over this before...there's no way that an A9 has a 30-40% advantage
>in terms of effective IPC. Most A9s have a crappy memory hierarchy and lose in a variety of benchmarks.

We've seen a few benchmarks showing how well the A9 does vs 1.6GHz netbook Atoms, which may not be perfect, but they tell a different story than what Intel claims. Unless Medfield has significantly improved IPC, I'd expect it to be a little slower than the netbook variants due to a slower memory system in mobiles. Also it can't run indefinitely at 1.6GHz.

If anything, I expect Medfield running at lower frequencies, having lower memory bandwidth, and using smaller L2 caches than the 28nm SoCs coming out this year.

>>Even in the few cases where x86 software is more finetuned, ie. JIT compilation, the unreleased Medfield shows no advantage over shipping mobiles: http://androidandme.com/2012/01/news/intel-medfield-vs-nvidia-tegra-3-performance-preview/
>
>>Assuming there are no further JIT improvements on the ARM side (unlikely), at the
>>end of 2012 Medfield will be 50% slower on Caffeinemark. And that is Intel's best
>>benchmark... Medfield will do far worse on everything else, >especially on mulithreading.
>
>So what is your estimate for the performance of Medfield on SPECint2000 relative to say, OMAP5, Tegra3 or S4?

Are we comparing compiler tricks or micro architectures? Assuming similar compiler technology, I'd estimate A9 to be 20-30% faster, and A15/Krait to be 60-100% faster at the same frequency - obviously varying considerably with the L2 size and memory of the SoC.

>>So how could you possibly justify the claim that a 1.6GHz ?2-way in-order single
>>core is faster or only a little slower than a 2GHz 3-way >OoO dual core?
>
>Because the performance gains for ARM licensees are limited by power consumption.

Sure, but the efficiency (performance per Watt) isn't the same. Eg. Tegra3 can run 2 cores at 1GHz using less power than Medfield at 1.6GHz (http://www.nvidia.com/content/PDF/tegra_white_papers/Variable-SMP-A-Multi-Core-CPU-Architecture-for-Low-Power-and-High-Performance.pdf). That's approximately twice the perf/Watt despite being 1 process node behind. Obviously these are NVidia vs Intel marketing numbers, so not necessarily reliable, but it gives an idea where things stand.

>You don't get 70% more performance while still having something recognizable as a cell phone.

A process node change can give that kind of improvement due to faster and lower power transistors.

>I'm willing to accept 30% higher performance at the same power consumption, although that sounds generous.

There aren't any numbers available for 28nm, but I don't think things will change in Medfields favour - overall performance/Watt will improve further in the next generation ARM SoCs.

Wilco
< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
Medfield article onlineDavid Kanter2012/01/23 01:51 PM
  server errorbakaneko2012/01/24 03:00 AM
    FixedDavid Kanter2012/01/24 04:02 AM
      FixedJoel2012/01/24 07:43 AM
      FixedRicardo B2012/01/24 11:25 AM
        FixedDavid Kanter2012/01/24 05:29 PM
      FixedGabriele Svelto2012/01/24 01:07 PM
        FixedDavid Kanter2012/01/24 05:30 PM
  Reference platform battery lifeDoug Siebert2012/01/24 02:03 PM
    standby timeFoo_2012/01/25 06:58 AM
      standby timeAnon2012/01/26 03:42 AM
        standby timeFoo_2012/01/26 04:02 AM
          standby timeDoug Siebert2012/01/26 12:39 PM
            standby timeAnon2012/01/26 01:22 PM
              standby timeanon2012/01/26 02:08 PM
                standby timeAnon2012/01/26 06:03 PM
                  standby timeanon2012/01/26 08:57 PM
                    standby timeanon2012/01/26 09:01 PM
                    standby timeAnon2012/01/27 09:32 PM
                standby timeDoug Siebert2012/01/27 02:15 PM
                  standby timeanon2012/01/27 02:41 PM
    Reference platform battery lifeDavid Kanter2012/01/27 10:09 AM
  Performance analysis laughableWilco2012/01/24 03:23 PM
    Performance analysis laughableDavid Kanter2012/01/24 05:19 PM
      Performance analysis laughableIntelUser20002012/01/24 07:30 PM
        Performance analysis laughableIntelUser20002012/01/24 07:32 PM
        Performance analysis laughableDavid Kanter2012/01/24 11:34 PM
          Performance analysis laughableIntelUser20002012/01/24 11:56 PM
            Performance analysis laughableDavid Kanter2012/01/25 02:07 AM
              Performance analysis laughableAlberto2012/01/25 12:54 PM
          Atom HT gainWilco2012/01/25 05:43 AM
            Atom HT gainIntelUser20002012/01/25 06:53 AM
              Atom HT gainnone2012/01/25 07:04 AM
                Atom HT gainIntelUser20002012/01/25 07:35 AM
            Atom HT gainFoo_2012/01/25 07:06 AM
      Performance analysis laughableWilco2012/01/24 08:21 PM
        Performance analysis laughableDavid Kanter2012/01/24 10:13 PM
          Performance analysis laughableWilco2012/01/25 04:30 AM
            Performance analysis laughablenone2012/01/25 06:14 AM
              Performance analysis laughableWilco2012/01/25 07:18 AM
                Performance analysis laughableobserver2012/01/26 04:17 AM
                  Performance analysis laughableWilco2012/01/26 06:25 AM
            Process numbersAlberto2012/01/26 09:29 AM
            Performance analysis laughableDavid Kanter2012/02/02 12:38 AM
          Performance analysis laughabletupper2012/01/25 04:27 PM
            Performance analysis laughableLinus Torvalds2012/01/25 08:37 PM
              Performance analysis laughableDoug Siebert2012/01/26 02:12 PM
  Medfield article onlineAndreas2012/01/25 03:10 AM
    Medfield article onlineAlberto2012/01/25 09:44 AM
    Medfield article onlineIntelUser20002012/01/25 10:24 AM
    Medfield article onlineDavid Kanter2012/01/25 09:58 PM
      Medfield article onlineDoug Siebert2012/01/26 01:20 PM
        Medfield article onlineEric2012/01/26 06:10 PM
          Medfield article onlineDoug Siebert2012/01/27 02:40 PM
  64-bitIngeneer2012/01/25 09:28 AM
    64-bitFoo_2012/01/25 10:23 AM
      64-bitIngeneer2012/01/25 02:34 PM
        64-bitUngo2012/01/25 04:08 PM
          64-bitEduardoS2012/01/26 12:55 PM
  Saltwell memcpySHK2012/01/26 02:41 AM
  Medfield WiFi & BluetoothRob Thorpe2012/01/26 03:09 AM
    Medfield WiFi & BluetoothDavid Kanter2012/01/27 05:54 PM
      Medfield WiFi & BluetoothRob Thorpe2012/01/28 02:22 PM
  Medfield article online (NT)Anil2012/01/26 05:57 PM
  Medfield article onlineAnil2012/01/26 06:11 PM
    Medfield article onlineMr. Camel2012/01/26 06:26 PM
    Medfield article onlinenone2012/01/27 01:41 AM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell avocado?