+- PINE64 (https://forum.pine64.org)
+-- Forum: PINE A64(+) (https://forum.pine64.org/forumdisplay.php?fid=4)
+--- Forum: Linux on Pine A64(+) (https://forum.pine64.org/forumdisplay.php?fid=6)
+--- Thread: Pine 64 benchmarks (/showthread.php?tid=389)
RE: Pine 64 benchmarks - longsleep - 03-07-2016
Thats useful thanks! Is take4 with the updated u-boot/dt? Also your heat sink is very small. It would be interesting if a bigger one could get it to stay below cooling state 3 with this benchmark.
RE: Pine 64 benchmarks - janjwerner - 03-07-2016
Still with the old u-boot/dt. Heatsink is tiny (still better than nothing). I will test new dtb later today.
cheers
RE: Pine 64 benchmarks - Andrew2 - 03-07-2016
(03-06-2016, 11:16 AM)janjwerner Wrote: Michael @ phoronix.com just benchmarked several ARM boards. http://www.phoronix.com/scan.php?page=article&item=raspberry-pi-3
Not worth a look, he uses wrong numbers as usual: http://forum.armbian.com/index.php/topic/789-breaking-news-choosing-armbian-speeds-up-your-orange-pi-multiple-times/
(03-06-2016, 08:04 PM)janjwerner Wrote: http://imgur.com/LyEl9dz
Ahem, the most important information is missing: How low has the clockspeed been reduced? Cooler state 2-3 and reaching 90°C... I would suspect heavy throttling happened.
Using a small fan and '-march=armv8-a -mtune=cortex-a53' it would be interesting how results improve. But unless you also improve heat dissipation it's rather useless to optimise compiler settings.
RE: Pine 64 benchmarks - longsleep - 03-07-2016
(03-07-2016, 11:30 AM)Andrew2 Wrote:(03-06-2016, 08:04 PM)janjwerner Wrote: http://imgur.com/LyEl9dz
Ahem, the most important information is missing: How low has the clockspeed been reduced? Cooler state 2-3 and reaching 90°C... I would suspect heavy throttling happened.
Oh why didn't i see this yesterday .. @janwerner can you post the image with the CPU speed again please. Would also be interesting to see the difference when you updated the device tree.
Thanks!
RE: Pine 64 benchmarks - Andrew2 - 03-07-2016
(03-07-2016, 12:43 PM)longsleep Wrote: @janwerner can you post the image with the CPU speed again please. Would also be interesting to see the difference when you updated the device tree.
Yes, this would really be interesting. But regarding 'raw performance' I made the experience that it's not worth a try to check the Pine64's real performance potential without a fan (I crashed mine yesterday, collision between fan and heatsink
so I'll have to wait for a new one)BTW: Do you provide a new Xenial image soon or better upgrade the old one?
RE: Pine 64 benchmarks - janjwerner - 03-07-2016
constants reminders of my sloppiness. updated the orignal post. sorry guys...
RE: Pine 64 benchmarks - longsleep - 03-07-2016
(03-07-2016, 01:34 PM)janjwerner Wrote:
No worries
Thanks for updating.I will publish a new Xenial image soon, fell asleep yesterday while building ..
RE: Pine 64 benchmarks - Andrew2 - 03-07-2016
(03-07-2016, 01:34 PM)janjwerner Wrote:
Thx, cpufreq not exceeding 1008MHz through the whole test and most of the times even lower (816 MHz minimum). With your current cooling capabilities the A64 can be considered to be clocked with approx. 950MHz. By adding a fan you can improve your benchmark scores by factor 1.2 (when thinking about 1152 MHz max).
If @tllim adopts our ideas regarding power through Euler connector, a perfect heatsink and a fan solution that doesn't suck after an extensive set of tests the Pine64 might also be overclocked up to 1200 MHz, 1248, 1296 or even 1344 MHz (most probably not without tweaking the dvfs settings, increasing VDD_CPUX voltage and therefore massively the need for heat dissipation) benchmark results will improve by factor 1.25 or even 1.4 (at 1344 MHz).
(using appropriate compiler switches and code optimizations might even work better
)
RE: Pine 64 benchmarks - janjwerner - 03-08-2016
OK,
some more, this time according to the new article from Micheal:
http://www.phoronix.com/scan.php?page=article&item=raspberry-pi3-odroid2
phoronix-test-suite batch-benchmark 1603051-GA-ODROIDPI362
RPI-Monitor output: http://imgur.com/8uIJr87
cpu frequency counters: http://pastebin.com/aTEEa8P4
results: http://openbenchmarking.org/result/1603089-GA-1603051GA88
memory card used: Samsung EVO class 10 32gb
This test was ran on @longsleep's 2016_03_06 xenial
If you want to run this tests make sure to create swap partition / enable swap file as FFMPEG will not build with 1GB of ram.
next round with:
export CXXFLAGS="-march=armv8-a -mtune=cortex-a53"
export CFLAGS="-march=armv8-a -mtune=cortex-a53"
Pine 64 benchmarks - Andrew2 - 03-08-2016
In case you want to improve SQLite performance you should buy another SD card that shows high IOPS (or use tempfs)since this is all that's 'tested' here. And with a fan you should already outperform RPi 3