Folding Forum

I have a Radeon card

06:32:57:W :CUDA not supported: Failed to open dynamic library 'libcuda.so': libcuda.so: cannot open shared object file: No such file or directory06:32:57:I3:gpus = {
06:32:57:I3: "gpu:03:00:00": {
06:32:57:I3: "vendor": 4098,
06:32:57:I3: "type": "amd",
06:32:57:I3: "description": "gfx1034",
06:32:57:I3: "opencl": {"platform": 0, "device": 0, "compute": "2.0", "driver": "3614.0"},
06:32:57:I3: "device": 29759,
06:32:57:I3: "supported": true
06:32:57:I3: },
06:32:57:I3: "gpu:13:00:00": {
06:32:57:I3: "vendor": 4098,
06:32:57:I3: "type": "amd",
06:32:57:I3: "description": "gfx1036",
06:32:57:I3: "opencl": {"platform": 0, "device": 1, "compute": "2.0", "driver": "3614.0"},
06:32:57:I3: "device": 5710,
06:32:57:I3: "supported": true
06:32:57:I3: }
06:32:57:I3:}

I thought I was sunk, but... it appears to be working !
(and for reference I simply followed the AMD driver installation instructions , i.e. that uses 'amdgpu-install' under linux mint 21.1)

I'm a little fuzzy on all this cuda and opencl, is this message showing up simply because I'm not using an nvidia card ?

The other thing I was wondering - is there any way for me to gauge how much faster, or slower, the GPU card is than the CPU ?

Thanks!

That message is expected if you don’t have CUDA installed. Nothing to worry about, unless you’re trying to fold on nvidia.

GPUs are much faster than CPUs. You will see much higher estimated points per day.

So that's what PPD stands for ! lol

I've noticed the PPD numbers move around. The CPU was 700k and the GPU 200k about an hour ago, but now the CPU is 200k.

I'm reading through the DB trying to figure out what the expected performance is.

It will vary wildly by project. It also won’t be accurate until a few percent done.

Yes, AMD is not using CUDA (yet ). Developer likes this error message, so it is here to stay for us AMD people

You might want to either pause your gfx1036 (Radeon integrated graphics inside of your CPU), or disable it completely in BIOS (if you are not using it for something). That iGPU will not get any work, until HIP arrives, but even then, quite slow for FAH