Current Performance and Stability Issues on AllWinner A80 OptimusBoard Development Board

Orange Pi Development Boards

Sine A80 OptimusBoard is the first ARM hardware that supports both USB 3.0 and Gigabit that I’ve ever owned, so I though it might be interesting to see what performance I could get with a USB 3.0 hard drive through the USB 3.0 OTG port of the board. For testing purpose, I bought a USB 3.0 OTG adapter on Ebay, but I did not look closely enough as it turned out to be only a USB 3.0 OTG to USB 2.0 female adapter capable of USB 2.0 speeds… Kind of useless item since you can plug a standard USB 2.0 OTG adapter into a USB 3.0 OTG receptacle. But I tried my hard drive anyway, and quickly realized write performance was very poor at 3 MB/s on the NTFS partition, whereas most other devices can handle at least15 MB/s, and usually over 25 MB/s. So I contacted Allwinner with this issue, and they were kind enough to look into it, and provided an updated firmware.

Today, I tried it, and it seems they’ve selected a more aggressive scheduler for the board, as some benchmarks results are much higher. For example, Vellamo Browser score was over 3,000 today, whereas during my initial benchmarks, it only achieved 2,300 points. So quite a boost. Unfortunately, this also introduced so stability issues (possibly resolvable with a heatsink and a fan), and did not improve NTFS performance in a major way:

  1. Multicore benchmarks will make the system reboot. Reproducible with Antutu Multi-thread and Vellamo Multicore tests.
  2. Riptide GP2 game will freeze after a short while. Very similar to what I’ve experienced with some Amlogic S802 TV boxes, except it happens much earlier (after one race).
  3. NTFS write speed is now 4.6MB/s, so it’s still an issue. FAT32 write speed is 29 MB/s, which is fine.

I’ve recorded these three issues in the video below.

So if you are wondering why there aren’t any Allwinner A80 mini PCs just yet, why Cubieboard8 is nowhere to be seen, and why no SDK has been provided for OptimiusBoard and pcDuino8, this could be could the explanation, and more work is needed at this stage of development. The first two issues were not present in the first firmware, but a more aggressive scheduler may have introduced the reboot/freeze issues. I don’t have a spare heatsink/fan, so I haven’t tried to work around these with passive/active cooling. Another possibility for the reboot could be the 5V/3A power adapter does not provide enough power to the board.

13
Leave a Reply

avatar
13 Comment threads
0 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
7 Comment authors
CodyCubieboard4 Benchmarks in AndroidCubieboard4 (CC-A80) Unboxing and Getting Started GuideJeremyTronsmart Draco AW80 Octa core Android TV Box Review Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
anon
Guest
anon

Only performance+noop is a combination worth using on all ARM platforms so far, maybe AMD Opteron changes things, but even Nvidia K1 (both Cortex-A15 and Cortex-A57) platform still suffers on absolutely insane long latency between frequency/voltage changes of ondemand and other schedulers.

And if you plan to do any serious ARM platform reviews you really should get one of those SMPS lab power supply boxes/modules, really, made a huge difference on stability of function on any type of system.

Peter Bauer
Guest

I have seen some troubles with the kernel 3.10.24 ondemand schedulers on Jetson TK1 (stability)
and indeed setting the scheduler to performance solves that.
http://bitkistl.blogspot.co.at/2014/03/evaluating-jetson-tk1-board.html

Simos
Guest
Simos

I managed to run Antutu (without heatsink) and got a score of around 58000.
I also ran 3DMark and got around 9000 which apparently is a modest result.
Admittedly, the SoC got hot, really hot. You can try to touch the SoC to gauge the temperature.
I am not sure if the Antutu score was reliable.
I’ll avoid running such benchmarks again until I get a heatsink.

What would be good is if the SoC has a feature to obtain the temperature.

ovi
Guest

Hello Jean-Luc

I have brought 2 x Optimus boards from Shenzen 2 weeks ago while i was there for our linux DVB Box which is in project to be released soon (will manage to send you a sample).
So i buyed these boards for a side project , the plan is to build an A80 clustered transcoding cheap machine.
The main issue is the lack of open source drivers for the 64 core PowerVR GPU , PowerVR not very generous with the sources for the GPU.
Had to hire an gpu driver writter to can unleash the power of processing of this board.

WIP at the moment but ill inform you with the progress.

trackback

[…] experienced a slow write speed with A80 OptimusBoard, even worse at 3 to 4MB/s, but never published due to other issues, so I’m simply guessing Allwinner SDK has not been optimized for NTFS performance. To prove […]

Jeremy
Guest
Jeremy

Did anyone have any luck with the stability issues? I’m seeing problems where the board just reboots when you try and run any application that has more than a few threads (Linux).

trackback

[…] regards to Ethernet and storage performance, and see if any progress has been made compared to the disappointing results I got with A80 OptimusBoard last year. I’ll also evaluate Linux performance on the board, and possibly tried out the […]

trackback

[…] results for example for USB storage. I documented my findings in a post entitled “Current Performance and Stability Issues on AllWinner A80 OptimusBoard Development Board” which was written in October 2014. But a few months have passed, and since Cubieboard4 is […]

Cody
Guest
Cody

I’ve been having a issue with the board that has me stumped. I have a PcDuino8 & when the board warms up, I start getting strange input errors. Android OS keeps returning me to the Home Screen & Linux acts like Enter is being held down on the keyboard.
Android is ok if I can keep the board below 40C and Linux is almost unusable.
Any idea on what is happening?