With VP9 the decoder firmware tends to crash easily and then fallback to software rendering until you restart the application. That seems to be a bug in upstream qualcomm venus driver.
Ps. what you linked is for qualcomm iris driver which at some point will have support for this chip according to their plans but that is far in to the future as they work on other SoC’s first.
Geekbench will never tell you when results are invalid due to background activity, throttling, shitty settings or whatever. It simpy uploads garbage numbers to its own database and that’s why this ‘tool’ is popular, It does not annoy clueless users with the most common benchmarking outcome that’s called ‘something went wrong, scores invalid’,
As a side note: the @strongtz scores were made with a different DRAM initialization, so your preliminary blog post claiming to compare active vs. passive cooling is a bit off.
Does anyone know when the Radxa Dragon Q6A SBC will become available for purchase or pre-order?
I haven’t seen any official release date yet, so any info or update would be greatly appreciated.
The PCB fabrication lead time is too long for QCS6490.
The Dragon Q6A is so far the only “single” board device based on QCS6490, all the others are built from existing Qualcomm reference design, which is a SoM.
More feedback. I connected a screen to the board after I upgraded to the t4 image and latest firmware. I started firefox and found some videos on youtube up to 4k 60Hz; I first tried a 2.5k one (that’s the resolution of my screen), and it was quite jerky. I saw that the SoC was super hot (my thermal camera told me around 77°C). I placed a small aluminum heatsink and it worked reasonably well but with some dropped frames (about one frame dropped per second, i.e. 1.6%). Looking at the CPU usage I saw “RDDprocess” taking 100% and firefox taking half of a CPU. I’m now seeing that this thing is a process isolation from firefox, so perhaps there’s a little bit to save by disabling this isolation (on the other hand, it allows to use more cores). Or maybe a slight overclocking to 2.8 would stop the dropped frames. I’ll continue to experiment a bit, but so far I’m pleasantly surprised as I wouldn’t have imagined such a small board playing such videos.
for youtube you’d want chromium from amazingfate‘s PPA that can utilise V4L2 hardware acceleration as well as an extension that forces h264 instead of vp9 on youtube since there is a bug in the venus driver that crashes with vp9. Issues are reported upstream and most likely will be solved with the upcoming iris driver for video acceleration. You probably experienced software decoding
Makes sense. At least it shows that the CPU is powerful enough if everything’s done in software.
Edit: I tried what I think is amazingfate’s ppa (disclaimer: I have zero experience with PPA so I blindly copy-pasted the commands from the repo’s description). I only found “chromium from debian” which I installed and doesn’t seem to provide codecs. It plays slowly for 5s then pauses for 5s. I also noticed some rockchip-specific packages, which make me wonder if this is not what you had in mind (and I tried them just in case but they don’t work here, they’re built for wayland).
No big deal, I’m definitely not a desktop power user anyway and don’t know enough about these environments to understand what I’m doing (my last encounter of KDE was in 1998 just before 1.0 was released). That just reminds me about windows when I was using it a very long time ago, where you would blindly copy-paste stuff and keep fingers crossed, so I must confess I never feel at ease randomly clicking everywhere there and am always wondering if my feedback can really be relevant to targeted users :-/