Fresh out of the box gpu and been using it on Freebsd releng 11.2 for almost a week. Absolutely no hitchs.
drm-stable-kmodgpu-firmware-kmodxf86-video-amdgpu - all working fine.
Been on Freebsd only for a few months - Arch for a few years prior. Figured I try Fedora for needed things Freebsd can’t do - yet.
So minimal install then boot into it - dmesg has a line about polaris firmware loaded and amd"something" taking over efifb.
I can get off a couple of commands then screen blackens and the kernel panicks and the card gets a hot spot at the edge opposite the power connector. Way hotter than when booted in Freebsd.
Does this card run hot for anyone else? Just Fedora?
This didn’t happen while booted in the installer.
Userspace stuff doing power mngmnt for the card?
Can you run sensors
It’s part of the lm-sensors package and can be used to read GPU temperature.
Run watch -n2 sensors
and look at the GPU temperatures
You can also use the following script to look at the GPU clock states and see if power management is working correctly
sudo watch -n2 'cat /sys/kernel/debug/dri/0/amdgpu_pm_info'
This script will read the 1st GPU’s power management info every 2 seconds.
Gpu temp was 45c. Then black screen. Its a dpms thing maybe. It’s occurring in both installers as well. Temp is the same for both distros(linux). I was in the arch installer about to reuse the volume group set down by fedora when it rebooted - into - fedora 28. Was about to run the watch command ( I get 2 sec but not much more,) when it locked and started heating up.
If the card is running 45c at idle under linux before it locks, then by “feel” its running 10c cooler under Freebsd.
3770k, Gigabyte GA z77x-up7 w/F5 bios
What kernel are you using?
I hope it’s not related to an adverse effect of:
https://www.phoronix.com/scan.php?page=news_item&px=AMDGPU-4.18-Power-Draw
I’m going to set it aside. I’ll keep fedora and fix it in time, chroot - whatever. For now I’ll boot freebsd, leave the updates alone, and do my research/reporting from there.
The kernel version is latest for each, honestly I don’t even know. I expected to just run through this (amdgpu 2+yrs old, 6+yrs exp on arch.) I’m guessing new kernel and iso’s anytime will fix it, so I’ll watch for that to.
Now I’m gonna click your link.
Bug. In 4.19 too.
What I have now on freebsd: drm corresponding to Linux 4.9, gpu-firmware-kmod version g20180825, and xf86-video-amdgpu-18.1.0 - when in X.