Tests and benchmarks


#21

Has somebody done a comparision between the small heatsink, delivered with the Performance Set

and the big heatsink?

How great or small are the differences in temperature when idle and under full load?


#22

Hi. I can tell you how it is with the small heatsink.
This just isn’t sufficient. It overheats immediatly without a fan.
Wth the big heatsink you can do normal tasks without a fan, but for heay load I’d still use a fan.
I still don’t have the big heatsink, I need to order one. I’ve got the NanoPi M4 with such a big heatsink, it’ a lot more useable since I don’t need a fan.

Temperatures

Debian armhf
No fan idle : 50°C
No fan max load : 85°C throttle keeps rising to +90°C
With fan idle : 37°C
With fan max load : 77°C

Ubuntu arm64
No fan idle : 56°C
No fan max load : 85°C throttle keeps rising to 95°C
With fan idle : 38°C
With fan max load : 83°C


#23

If have two boards, one with small one with big heatsink
The big heatsink’s mass acts in two ways: First like a capacity, to quickly absorb the generated heat and dissipate it with a time delay.
Second with a larger surface to the surrounding air.

This means in a pure passive setup it can help to go through longer peak phases, and also give off more energy to the surrounding air. It depends on the surrounding air temperature, humidity and airflow for how long this works, without CPU throttling.

The small cooler should work fine with a solid active fan, or in aircooled invironments with some “natural” airflow and only sporadic short CPU spikes.


#24

Just look a bit up (just a bit before ask), i typed full set of tests about this.


#25

Thanks for the detailed tests. It’s really help to understand the difference compared with other boards.

Can you make one more test to check WebGL performance in Firefox and Chrome? There are two most popular pages for this:

  1. https://webglsamples.org/aquarium/aquarium.html
  2. https://threejs.org/examples/#webgl_decals

It’s very interesting what FPS rates you’ll see there :grinning:
If WebGL support is disabled in the browser, you can easily enable it:

And thanks for your work. We appreciate it.


#26

Ubuntu
(ATTENTION: Testing was done on custom build xserver from rockchip with this merge request and with modesetting.conf)

Ubuntu

Chrome detected ARM Mali-T860. Started with taskset -c 4-5 (flags in spoiler at the end)
webglsamples Canvas 1024x1024
On 500 Fishes - 20-21 fps
On 1000 - 11-12 fps
On 30000 - 1-2 fps
threejs - 9 fps when i throw paint ball, 30 fps otherwise

Firefox don’t have egl without rebuilding it from the scratch, so it’s use VMware. If i try to use gl4es it’s just fails to detect WebGL.

webglsamples
On 500 Fishes - 3 fps
On 1000 - 3 fps
On 30000 - less than 1 fps
threejs - 3 fps when i throw ball, 10 fps otherwise

Chromium flags

–disable-low-res-tiling
–num-raster-threads=6
–profiler-timing=0
–disable-composited-antialiasing
–disk-cache-dir=/tmp/
–no-sandbox
–test-type
–show-component-extension-options
–ignore-gpu-blacklist
–use-gl=egl

Debian

Debian

Fresh debian with update&&upgrade with radxa’s apt

Chrome detected ARM Mali-T860. Started with taskset -c 4-5
webglsamples Canvas 1024x1024
On 500 Fishes - 12-28 fps (unstable)
On 1000 - 14-17 fps
On 30000 - 1 fps
threejs - 22 fps when i throw a ball, 30-38 fps otherwise

And firefox ESR just crash on me


#27

Chrome results looks good :grinning: Many thanks


#28

I’ve changed the script used to test write speed. Added “progress” and “direct”.

#!/usr/bin/env bash
echo === WRITE 1GiB ===
echo -n "sync..." ; sync ; echo "ok"
echo 3 > /proc/sys/vm/drop_caches
time {
	dd if=/dev/zero of=temp conv=fdatasync bs=1024k count=1k status=progress oflag=direct
	echo -n "sync..." ; sync ; echo "ok"
}
rm -f temp
echo -n "sync..." ; sync ; echo "ok"
echo Done.