Results 1 to 7 of 7

Thread: Samsung Arndale Board Exynos 5 Dual Benchmarks

  1. #1
    Join Date
    Jan 2007
    Posts
    14,566

    Default Samsung Arndale Board Exynos 5 Dual Benchmarks

    Phoronix: Samsung Arndale Board Exynos 5 Dual Benchmarks

    The Arndale Board is a dual-core ARMv7 development board built around the Exynos 5 Dual SoC, which features the new ARM Cortex-A15. As shown in yesterday's Samsung Chromebook benchmarks on Linux, the Exynos 5 Dual packs very good performance for being a low-power ARM chip...

    http://www.phoronix.com/vr.php?view=MTIzMzU

  2. #2
    Join Date
    Oct 2010
    Posts
    91

    Default

    why's it so much faster? aren't they the same clock rate and same cpu? is that just linaro improvements?

  3. #3
    Join Date
    Dec 2010
    Location
    MA, USA
    Posts
    1,254

    Default

    Quote Originally Posted by mercutio View Post
    why's it so much faster? aren't they the same clock rate and same cpu? is that just linaro improvements?
    I have a feeling its because the arndale has additional things like USB3 and USB is a CPU intensive bus. Busses like FireWire can process information by itself to offload CPU work.

  4. #4
    Join Date
    Jan 2012
    Posts
    111

    Default

    Quote Originally Posted by mercutio View Post
    why's it so much faster? aren't they the same clock rate and same cpu? is that just linaro improvements?
    We can't be sure about the same clock rate, because this is a typical problem which haunts a lot of Phoronix benchmarks on ARM hardware. The Chromebook processor is reported as "ARMv7 rev 4 @ 1.70GHz (2 cores)". The Arndale processor is reported as "ARMv7 rev 4 (2 cores)". Looks like the Arndale was just running some dodgy software and it's results can't be taken seriously.

    But it was a great catch for Michael. Results differ for some reason? Does not matter, let's quickly post the news!

  5. #5
    Join Date
    Nov 2012
    Posts
    1

    Default

    A small mistake. According to the ArndaleBoard.org it's 2GB board, not 1GB: "1Gbytes x 2"

  6. #6

    Default

    Quote Originally Posted by ssvb View Post
    Looks like the Arndale was just running some dodgy software and it's results can't be taken seriously.
    It happens whenever there is not a cpufreq driver for the hardware, same reason up until few days ago Calxeda benchmarks don't have frequencies shown...

  7. #7
    Join Date
    Jan 2012
    Posts
    111

    Default

    Quote Originally Posted by Michael View Post
    It happens whenever there is not a cpufreq driver for the hardware, same reason up until few days ago Calxeda benchmarks don't have frequencies shown...
    It's not just the frequencies not shown. You don't know the clock frequency the board is running. You don't know whether the clock frequency was ever throttled down because of overheating (and I don't see a big heatsink on the Arndale board picture), etc.

    In order to get reliable benchmark results, you need proper cpufreq driver with cpufreq statistics reporting enabled (CPU_FREQ_STAT). The benchmark tool needs to set the cpufreq governor to "performance", take the snapshot of cpufreq stats and run the tests. After the benchmark has finished, check the cpufreq stats again. We expect that only the time spent on the maximum clock frequency should increase in the statistics. If you see any counter increases for the other frequencies, then there was thermal throttling kicking in during the benchmark and you know that your results are not trustworthy because the time and duration of throttling intervals is a bit less predictable.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •