GPU Owl suddenly not running

[Xebecer]

I've been running GPU Owl for many months now on six RX 5700 XT cards. Suddenly today, none will run, and I get this:


2020-09-22 00:21:41 config: -device 0 -user Xebecer -cpu Birlinn_Oban_GPU0 -cleanup
2020-09-22 00:21:41 config: -d 0
2020-09-22 00:21:41 device 0, unique id ''
2020-09-22 00:21:41 Birlinn_Oban_GPU0 108996493 FFT: 6M 1K:12:256 (17.32 bpw)
2020-09-22 00:21:41 Birlinn_Oban_GPU0 Expected maximum carry32: 2E490000
2020-09-22 00:21:41 Birlinn_Oban_GPU0 Exception gpu_error: clGetPlatformIDs(16, platforms, (unsigned *) &nPlatforms) at clwrap.cpp:71 getDeviceIDs
2020-09-22 00:21:41 Birlinn_Oban_GPU0 Bye

[paulunderwood]

Are you using Linux or Windows? Has the OS upgraded recently?

[Mark Rose]

Have you tried rebooting? Or power cycling your machine?

[preda]

Originally Posted by Xebecer

2020-09-22 00:21:41 Birlinn_Oban_GPU0 Exception gpu_error: clGetPlatformIDs(16, platforms, (unsigned *) &nPlatforms) at clwrap.cpp:71 getDeviceIDs

Does "clinfo" work? Did your system update recently? upgraded something?

Are you running ROCm? did ROCm update?

[kriesel]

Any OpenCL devices listed in gpuowl -h output, after the options, before the fft lengths?
Any other OpenCL utility able to detect your gpus? (Gpu-z, rocm-smi, etc, depending on OS)

[paulunderwood]

I ran apt-get update (Debian) and the system froze while installing ROCm-3.8.0 with the message "building initial module for 4.19-0-9-amd64". Rebooted. I had to do dpkg --configure -a (??) and it then updated the module for 3.19-0-10-am64. Rebooted.

I then recompiled gpuOwl against ROCm-3.8.0. I got my 2 instance of gpuOwl running.

Now preda's "pp.sh" script will not work and so I can't overclock the RAM.

[preda]

Originally Posted by paulunderwood

I ran apt-get update (Debian) and the system froze while installing ROCm-3.8.0 with the message "building initial module for 4.19-0-9-amd64". Rebooted. I had to do dpkg --configure -a (??) and it then updated the module for 3.19-0-10-am64. Rebooted.

I then recompiled gpuOwl against ROCm-3.8.0. I got my 2 instance of gpuOwl running.

Now preda's "pp.sh" script will not work and so I can't overclock the RAM.

How does the ROCm 3.8 performance look like? I understand that you can't compare directly because powerplay ain't working anymore..

I opened an issue about powerplay: https://github.com/RadeonOpenCompute/ROCm/issues/1228

[paulunderwood]

Originally Posted by preda

How does the ROCm 3.8 performance look like? I understand that you can't compare directly because powerplay ain't working anymore..

I opened an issue about powerplay: https://github.com/RadeonOpenCompute/ROCm/issues/1228

Here are parts of my /etc/default/grub

GRUB_CMDLINE_LINUX_DEFAULT="quiet amdgpu.ppfeaturemask=0xffffffff"

and /boot/grub/grub.cfg

linux	/boot/vmlinuz-4.19.0-10-amd64 root=UUID=f93eeec4-4134-4e79-b5c7-019d1dbc1ab2 ro  quiet amdgpu.ppfeaturemask=0xffffffff

and pp.sh:

rocm=/opt/rocm-3.8.0/bin/rocm-smi

pp() {
echo $*

cd /sys/class/drm/card$1/device
echo "m 1 $2" > pp_od_clk_voltage
echo "vc 1 1304 $3" > pp_od_clk_voltage
echo "vc 2 1801 $4" > pp_od_clk_voltage
echo c > pp_od_clk_voltage
$rocm -d$1 --setsclk $5
}

pp 0 1175 820 1050 3

I tried running the commands one by one and it just hangs at echo "m 1 1175" > pp_od_clk_voltage.

Here is the content of that file pp_od_clk_voltage:

OD_SCLK:
0:        808Mhz
1:       1801Mhz
OD_MCLK:
1:       1175Mhz
OD_VDDC_CURVE:
0:        808Mhz        715mV
1:       1304Mhz        826mV
2:       1801Mhz       1138mV
OD_RANGE:
SCLK:     808Mhz       2200Mhz
MCLK:     800Mhz       1200Mhz
VDDC_CURVE_SCLK[0]:     808Mhz       2200Mhz
VDDC_CURVE_VOLT[0]:     738mV        1218mV
VDDC_CURVE_SCLK[1]:     808Mhz       2200Mhz
VDDC_CURVE_VOLT[1]:     738mV        1218mV
VDDC_CURVE_SCLK[2]:     808Mhz       2200Mhz
VDDC_CURVE_VOLT[2]:     738mV        1218mV

I just ran echo c > pp_od_clk_voltage and got the overclock but not the voltage drop (I think).

[preda]

Originally Posted by paulunderwood

I tried running the commands one by one and it just hangs at echo "m 1 1175" > pp_od_clk_voltage.

Try to use the rocm-smi script to set the RAM frequency, I'm curious whether that works. Something along the lines of:

rocm-smi --setmclk 2
rocm-smi --autorespond y --setmemoverdrive 10

If it hangs on memoverdrive, you should check "dmesg" sometimes somethig informative can be there at the end.

[moebius]

Originally Posted by Xebecer

2020-09-22 00:21:41 Birlinn_Oban_GPU0 Exception gpu_error: clGetPlatformIDs(16, platforms, (unsigned *) &nPlatforms) at clwrap.cpp:71 getDeviceIDs
2020-09-22 00:21:41 Birlinn_Oban_GPU0 Bye

This error message is exactly the same as the error message that you receive when you no longer have a GPU runtime available in Google.Colab. So no GPU was just found. Maybe it helps to reinstall the gpu/ROCm drivers?

[Xebecer]

I forced the 'Insider Ring' update to Windows, and all is well. But, that was weird.