Me for example: https://github.com/ThomasKaiser/sbc-bench/blob/master/results/reviews/Radxa-Orion-O6-1.0.md
But with the preliminary state of software it’s rather pointless to collect any scores yet.
Orion O6 Debug Party Invitation
2 Likes
Radxa, you do know what else is needed to get the iGPU working on mainline linux via ACPI? There is currently a Panthor patch and a mesa patch to enable the G720 but that does not seems to be enoght.
So we can have an idea of the work needed.
Hey,
I am using RX6600 now with ACPI and I tested different upstream kernels.
Please note that you might need to try kernel 6.12 first (vanilla if possible) or in case of any issues try to revert this one:
cf424020e040 drm/amdgpu/hdp5.0: do a posting read when flushing HDP
I raised it in maillist:
https://lore.kernel.org/amd-gfx/D97FB92117J2.PXTNFKCIRWAS@linaro.org/
After reverting it works fine mostly.
Can someone with 9.9.0 bios please upload the uart 2 log. I think my optee is corrupted and I want to compare logs
Current mainline efforts are DT-only. Pretty much all peripherals other than USB, PCIe and UART will require significant work to support ACPI binding.
There’s also the display controller that needs to be brought up. Right now you have to rely on UEFI GOP to get some basic display output.
Looks like my board decided to forget its hardware uuid on v9.9.0 after not turning on for a couple days:
Where normally it says:
------End firmware version dump------
E/TC:?? 00 cix_get_key_info:121 Get device key:
Mine says:
------End firmware version dump------
E/TA: _plat__NvInitFromStorage:159 fTPM: block 1 not found, creating
E/TC:?? 00 read_fat:1052 tee_mm_alloc Fail
E/TA: _plat__NvInitFromStorage:173 Failed to create fTPM storage object
E/TA: _plat__GetEPS:78 fTPM was unable to derive an EPS, falling back to random generation.
E/TC:?? 00
E/TC:?? 00 TA panicked with code 0xffff0006
E/LD: Status of TA bc50d971-d4c9-42c4-82cb-343fb7f37896
and a bit more error stuff
Hoping to be able to get it back working when reflashing the bios via an USB SPI Programmer tomorrow. Until then I would discourage upgrading to v9.9.0
1 Like
FWIW mine is here, in ACPI mode with mainline kernel: https://0x0.st/8fEb.bin
I’m noticing that 3 CPU flags only appear on the original kernel, not in mainline. Your A520 seem to be significantly faster than mine on memory accesses, I’ll run some comparison tests to see if any of the OPP has an effect there.
For people that downloaded the initial 9.9.0 Bios there is an updated one: https://dl.radxa.com/orion/o6/images/bios/SystemReady/latest/
2 Likes
Ill try the new bios and the stock Ubuntu 25.04 in ACPI image today if i can. The 24.10 and the 6.11 kernel has been working really really well for me. Even trought the kernel is a bit old.
Yesterday i even got Steam running after compiling box64, but i was unable to do anything with it.
I was worried it would not work at all because you cant run in 32 bit mode (box86) in Armv9, but box32 mode seems to be good enoght to at least get the UI going.
There is a kernel panic booting Ubuntu 25.04 iso, 6.14 kernel about arm-smmu.
Im 99% sure this is caused by the 9.0.0 bios and it was at least booting on the old 0.2.1 bios.
2 Likes
I was able to install a daily build a few days ago, on the older 9.0 bios. No wifi working, but it seems to be a problem for other 25.04 ubuntu users.
I am using RX6600 now with ACPI and I tested different upstream kernels.
AMDGPU developer is looking for some users to test the fix patch
https://lore.kernel.org/amd-gfx/CADnq5_M=YiMVvMpGaFhn2T3jRWGY2FrsUwCVPG6HupmTzZCYug@mail.gmail.com/
Petty sure i booted the daily live image at least once with the 0.2.x bios, then switched to the first 9.0.0 it the same iso would not boot anymore, same result now with the 2nd 9.0.0 and the release version of 25.04.
Also i see no difference with the 2nd 9.0.0 bios compared to the first, PPTT tables seems to be still missing as Windows still list 96MB L3 with 8 sockets, i can still enable the small cores… etc, i think it is just a re-upload of the first
Have you tried booting from an usb c stick? Had some issues installing ubuntu after a fedora install and booting from usb c seemed to work better.
cix_flash_all.bin has a new date (Apr 11 instead of Apr 7) and new MD5. All other files are identical.
I’m not sure exactly what the implications are of cix_flash_all.bin changing but not cix_flash_ota.bin…
AMDGPU developer is looking for some users to test the fix patch
New patch for testing
https://lore.kernel.org/amd-gfx/CADnq5_NE2M19JdrULtJH-OXwycDpu0hrFHy42YiJA3nMYoP=+w@mail.gmail.com/
Subject: [PATCH] drm/amdgpu: only remap HDP registers on X86_64
We remap the HDP registers to an open part of the MMIO
aperture. This doesn't seem to work properly on non-X86 systems,
so only remap on X86_64.
2 Likes
I upgraded to the latest 9.0.0 firmware, and attempted to install Debian 12 Arm ISO as indicated in the Radxa O6 Wiki page (https://docs.radxa.com/en/orion/o6/bios/boot-debian-official-iso), but when it was installing I couldn’t get the Live CD installer to detect the built-in Ethernet (RTL8126). It seems like that driver is not present.
(This was with ACPI, not Device Tree.) Is there a better guide for Debian 12, or is it recommended to use Ubuntu 25.04 or 24.10 instead for now with ACPI firmware 9.0.0?
I also tried Ubuntu 25.10, and it booted into the live CD environment after a bit, but was stuck with these messages a while:
Then it was stuck a bit after systemd-update-done.service. Not sure if that’s normal, but it progressed to the graphical installer at least.
1 Like
Re-running some benchmarks, it seems that even with the BIOS setting the 2.6 GHz limit instead of 2.5, I’m only seeing 8 cores total in Ubuntu, and Geekbench score is reduced slightly, as well as HPL (going from 31.232 Gflops to 20.161 Gflops). Both of those HPL scores are still lower than a 4-core A76 Raspberry Pi 5…
This does not seem to work, at least you were running at 2.5 GHz: https://browser.geekbench.com/v6/cpu/11703943.gb6
Isn’t there an UEFI setting to bring back the little cores? The changelog.txt contains “Remove little core enablement option in BIOS Setup.”
Asides that when trying to test different firmwares/kernels/whatever (in general: most probably different settings) I don’t understand why incapable tools are used that do not measure both RAM bandwidth and latency individually. Since depending on type of benchmark different memory performance might matter more than CPU ‘horsepower’.
As such: benchmarking gone wrong confirmed? With reliable benchmarks the O6 with all 12 cores active scores three times higher than an RPi 5B. So how should the low HPL score be explained other than something went wrong? IMO a benchmark is reliable when it is not that much affected by dependencies like software libs, OS and kernel version (7-zip’s internal benchmark being a pretty good example)
1 Like