Bug 210787 - amdgpu fan NA on multi gpu R9 nano
Summary: amdgpu fan NA on multi gpu R9 nano
Status: RESOLVED OBSOLETE
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: x86-64 Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2020-12-19 09:20 UTC by Janpieter Sollie
Modified: 2021-12-12 10:47 UTC (History)
1 user (show)

See Also:
Kernel Version: 5.9.12-5.10.1
Subsystem:
Regression: No
Bisected commit-id:


Attachments
dmesg with manual GPU loading (8.48 KB, text/plain)
2020-12-19 09:20 UTC, Janpieter Sollie
Details
dmesg with amdgpu loaded automatically (97.09 KB, text/plain)
2020-12-19 09:25 UTC, Janpieter Sollie
Details

Description Janpieter Sollie 2020-12-19 09:20:00 UTC
Created attachment 294229 [details]
dmesg with manual GPU loading

when running sensors, the 1st GPU fan (this GPU has the display attached) is unavailable:
> amdgpu-pci-0700
> Adapter: PCI adapter
> vddgfx:      900.00 mV
> fan1:             N/A  (min = 1000 RPM, max = 4200 RPM)
> edge:         +23.0°C  (crit = +89.0°C, hyst = -273.1°C)
> power1:       10.23 W  (cap = 150.00 W)
>
> amdgpu-pci-4300
> Adapter: PCI adapter
> vddgfx:      850.00 mV
> fan1:        1544 RPM  (min = 1000 RPM, max = 4200 RPM)
> edge:         +26.0°C  (crit = +89.0°C, hyst = -273.1°C)
> power1:       10.08 W  (cap = 150.00 W)
blacklisting amdgpu did not help, but when breaking the config (specifying an invalid parameter in /etc/modprobe.d/amdgpu.conf), and loading it manually with correct parameters, it works.

We're talking about a ryzen 1950x setup here with 2x r9 nano GPUs
Comment 1 Janpieter Sollie 2020-12-19 09:25:14 UTC
Created attachment 294231 [details]
dmesg with amdgpu loaded automatically
Comment 2 Alex Deucher 2020-12-21 14:26:54 UTC
What parameters are you passing to the driver?  It looks like you are disabling UVD and VCE.
Comment 3 Janpieter Sollie 2020-12-21 15:21:26 UTC
Disabling UVD + VCE is on purpose... gfx8 video coding is simply too bad to be useful. Next to that, this issue still apprears:
https://bugzilla.kernel.org/show_bug.cgi?id=206155
Comment 4 Janpieter Sollie 2020-12-21 15:27:37 UTC
My amdgpu options:
options amdgpu ip_block_mask=0x7f discovery=1 compute_multipipe=1 gpu_recovery=1 mes=1 abmlevel=4 send_sigterm=1 sched_policy=1 dpm=1 aspm=1
Comment 5 Alex Deucher 2020-12-21 16:43:03 UTC
Most of those are not applicable or the default.  Does it work properly without any parameters?
Comment 6 Janpieter Sollie 2020-12-22 04:20:13 UTC
No, I rebooted with no options (aside from IPmask, it is required to bring up the gpu) and it doesn't work either
Comment 7 Janpieter Sollie 2021-12-12 10:47:54 UTC
no longer using 2 GPUs

Note You need to log in before you can comment on or make changes to this bug.