The Current Linux Performance With 16 ARM Boards

[OMTDesign]

Cool to see that ARM boards are getting some love.

I decided to download and run phoronix-test-suite on my Orange Pi One Plus, which has a Allwinner H6 soc, and running armbian Bionic 4.18RC7. I didn't know that the test needed to run for over 4 hours! Micheal must have some insane level of patience to run tests like these on a near daily basis.

https://i.imgur.com/RFCOrLj.jpg

I have attached link to a photo of some results. I knew that Orange Pi had poor xh264 support, but I didn't expect it to perform so poorly. Especially since it has a Mali T720 MP2 GPU, which on paper should have outperformed the Mali400 MP2 on the H2+ and H3 socs.

[OMTDesign]

Here are the results of my first test.

http://openbenchmarking.org/result/1809180-KH-1809111RA96

There are no results for the Perl and Redis benchmarks, due to me not installing Perl beforehand. Also I'm not sure as to how I can add the cost of the board ($20USD) to the test data.

[Michael]

Here are the results of my first test.

http://openbenchmarking.org/result/1809180-KH-1809111RA96

There are no results for the Perl and Redis benchmarks, due to me not installing Perl beforehand. Also I'm not sure as to how I can add the cost of the board ($20USD) to the test data.

Welcome. Hmmm, PTS should have installed Perl by itself unless it skipped it. Anyhow, for price-per-dollar that would be done via COST_PERF_PER_DOLLAR=20 as an environment variable or after the fact can be added via phoronix-test-suite perf_per_dollar.add 1809180-KH-1809111RA96

[caligula]

Pretty much when it comes to the random ARM boards aside from like Jetson or ASUS, it's just whatever LoveRPI.com ends up sending over... Sadly the likes of ODROID and stuff haven't offered sending over any hardware, so my ARM boards are basically what I get sent out and not much control over that.

That's fine. Anyway, I think the Tritium boards cover most of the Banana/Orange/NanoPi board SoCs. Although the performance figures don't look so good. Not sure, but I think the Armbian community might have some special knowledge on tuning things (cpu freq, dvfs, crypto, video acceleration, compiler switches). For example some of the boards have native gigabit LAN so the server/network benchmarks should be a lot faster than on RPi.

[caligula]

Cool to see that ARM boards are getting some love.

I decided to download and run phoronix-test-suite on my Orange Pi One Plus, which has a Allwinner H6 soc, and running armbian Bionic 4.18RC7. I didn't know that the test needed to run for over 4 hours! Micheal must have some insane level of patience to run tests like these on a near daily basis.

https://i.imgur.com/RFCOrLj.jpg

I have attached link to a photo of some results. I knew that Orange Pi had poor xh264 support, but I didn't expect it to perform so poorly. Especially since it has a Mali T720 MP2 GPU, which on paper should have outperformed the Mali400 MP2 on the H2+ and H3 socs.

They recently had a crowdfunding campaign for the old boards. I'd expect hw accelerated h264 by now.

[ah--]

Some RockPro64 results: https://openbenchmarking.org/result/...RA60&obr_sor=y
As before, ignore anything related to disk as my sdcard is terrible.

Also have you tried just asking for a board? Pine64 have sent out a bunch for OS development etc.

[OMTDesign]

They recently had a crowdfunding campaign for the old boards. I'd expect hw accelerated h264 by now.

Well according to the crowdfunding website, the H6 soc (which has a different GPU) was not a part of their original or stretch goals

Making sure that the codec works on the older Allwinner SoCs that are still widely used: A10 (Cubieboard), A13 (A13-Olinuxino), A20 (Cubieboard 2, A20-Olinuxino), A33 (A33-Olinuxino, BananaPi M2-Magic), R8 (CHIP) and R16 (NES and Super NES classic). Support for the newer SoCs (H3, H5 and A64) requires more work, and is part of our first stretch goal below.

[LoveRPi]

Tritium results should be the same as the other H3/H5 board results by Banana Pi, Nano Pi, and Orange Pi. Libre Computer boards and software run at stable frequencies and do not overclock. For H3/H5, Tritium runs at 1.008GHz. The Chinese SBC vendors will claim H3/H5 will run higher when it really isn't stable. HardKernel also overclocks the ODROID-C2 by 20% which is why you are seeing differences between Le Potato and ODROID-C2. The operating frequency in their images is also not a stable frequency. Michael didn't run an AES/Encryption benchmark. Le Potato wipes the floor on all of the 32-bit ARM boards like ASUS Tinker Board as well as ODROID-C2 and Raspberry Pi 3.

[Degra]

I'm just running some of the benchmarks on a badly cooled RockPro64 with the worlds slowest sd card, some results

Could you tell me what you use for cooling the RockPro64?
I'm thinking about buying one but I'm not sure which cooler / fan to buy.

Some benchmarking results you posted seem disappointing and I wonder how much it can be improved with proper cooling.

[ah--]

Could you tell me what you use for cooling the RockPro64?
I'm thinking about buying one but I'm not sure which cooler / fan to buy.

Some benchmarking results you posted seem disappointing and I wonder how much it can be improved with proper cooling.

I strongly suspect you could improve the results a lot with proper cooling, I'm using the 3cm heatsink without a fan, in the NAS case without any other ventilation and two HDDs in there. And the whole thing is then inside a closed cupboard together with my router.

I like it because it's still easily fast enough for NAS usage and it's pretty silent.