Hello all,
Regarding amdgpu, I've been using some Radeon 5700XTs for compute work with kernels through 5.8.14. I recently tried kernel 5.9.0, and found that the following is no longer allowed:
echo "m 1 200" | sudo tee /sys/class/drm/card0/device/pp_od_clk_voltage
Is this an expected change? If so, where should I look for documentation regarding how to manually configure these cards with this kernel. Note, I've had this working with 5.8 kernels for months.
Thanks
Josh
On Sun, Oct 18, 2020 at 4:32 PM Josh Fuhs Joshua.Fuhs@pm.me wrote:
Hello all,
Regarding amdgpu, I've been using some Radeon 5700XTs for compute work with kernels through 5.8.14. I recently tried kernel 5.9.0, and found that the following is no longer allowed:
echo "m 1 200" | sudo tee /sys/class/drm/card0/device/pp_od_clk_voltage
Is this an expected change? If so, where should I look for documentation regarding how to manually configure these cards with this kernel. Note, I've had this working with 5.8 kernels for months.
It was a bug. Fixed in 5.9.1.
Alex
Thanks. I tried 5.9.1 and I think there's still a problem, or at least something different.
Using the same configuration script, I noticed that my cards are running a lot hotter. For example, here's total power consumption of a two-card system with two different kernels:
5.8.14: 460W 5.9.1: 560+W
Memory and system clocks are initially set the same on all cards in all cases.
Josh
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Monday, October 19, 2020 2:22 PM, Alex Deucher alexdeucher@gmail.com wrote:
On Sun, Oct 18, 2020 at 4:32 PM Josh Fuhs Joshua.Fuhs@pm.me wrote:
Hello all, Regarding amdgpu, I've been using some Radeon 5700XTs for compute work with kernels through 5.8.14. I recently tried kernel 5.9.0, and found that the following is no longer allowed: echo "m 1 200" | sudo tee /sys/class/drm/card0/device/pp_od_clk_voltage Is this an expected change? If so, where should I look for documentation regarding how to manually configure these cards with this kernel. Note, I've had this working with 5.8 kernels for months.
It was a bug. Fixed in 5.9.1.
Alex
On Mon, Oct 19, 2020 at 8:53 PM Josh Fuhs Joshua.Fuhs@pm.me wrote:
Thanks. I tried 5.9.1 and I think there's still a problem, or at least something different.
Using the same configuration script, I noticed that my cards are running a lot hotter. For example, here's total power consumption of a two-card system with two different kernels:
5.8.14: 460W 5.9.1: 560+WMemory and system clocks are initially set the same on all cards in all cases.
Can you bisect?
Alex
Josh
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Monday, October 19, 2020 2:22 PM, Alex Deucher alexdeucher@gmail.com wrote:
On Sun, Oct 18, 2020 at 4:32 PM Josh Fuhs Joshua.Fuhs@pm.me wrote:
Hello all, Regarding amdgpu, I've been using some Radeon 5700XTs for compute work with kernels through 5.8.14. I recently tried kernel 5.9.0, and found that the following is no longer allowed: echo "m 1 200" | sudo tee /sys/class/drm/card0/device/pp_od_clk_voltage Is this an expected change? If so, where should I look for documentation regarding how to manually configure these cards with this kernel. Note, I've had this working with 5.8 kernels for months.
It was a bug. Fixed in 5.9.1.
Alex
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Wednesday, October 21, 2020 2:44 PM, Alex Deucher alexdeucher@gmail.com wrote:
On Mon, Oct 19, 2020 at 8:53 PM Josh Fuhs Joshua.Fuhs@pm.me wrote:
Thanks. I tried 5.9.1 and I think there's still a problem, or at least something different. Using the same configuration script, I noticed that my cards are running a lot hotter. For example, here's total power consumption of a two-card system with two different kernels:
5.8.14: 460W 5.9.1: 560+WMemory and system clocks are initially set the same on all cards in all cases.
Can you bisect?
I assume this means using git bisect to narrow down the commit that introduced the effect. I'm not set up for kernel builds. Is there a guide?
Josh
On Thu, Oct 22, 2020 at 2:50 PM Josh Fuhs Joshua.Fuhs@pm.me wrote:
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Wednesday, October 21, 2020 2:44 PM, Alex Deucher alexdeucher@gmail.com wrote:
On Mon, Oct 19, 2020 at 8:53 PM Josh Fuhs Joshua.Fuhs@pm.me wrote:
Thanks. I tried 5.9.1 and I think there's still a problem, or at least something different. Using the same configuration script, I noticed that my cards are running a lot hotter. For example, here's total power consumption of a two-card system with two different kernels:
5.8.14: 460W 5.9.1: 560+WMemory and system clocks are initially set the same on all cards in all cases.
Can you bisect?
I assume this means using git bisect to narrow down the commit that introduced the effect. I'm not set up for kernel builds. Is there a guide?
Yes: https://www.kernel.org/doc/html/latest/admin-guide/bug-bisect.html
Thanks!
Alex
dri-devel@lists.freedesktop.org
-
Alex Deucher -
Josh Fuhs