CUDA 10.1 not reaching GPU on RHEL 7.9, Quadro K2200

Accelerated Computing CUDA CUDA Setup and Installation
1

Installed CUDA 10.1 successfully on RHEL 7.9 with Quadro K2200 GPU and compiled/ran the following program using CLion/CMake, which only output result from CPU, not GPU. Does anyone know why CUDA is unable to reach the GPU? Thanks.

#include
#include “stdio.h”

global void cuda_hello(){
printf(“Hello World from GPU!\n”);
}

int main() {
std::cout << “CPU hello world!” << std::endl;
cuda_hello<<<1,1>>>();
cudaDeviceSynchronize();
system(“read -p ‘Press Enter to continue…’ var”);
return 0;
}

//Output:
//CPU hello world!
//Press Enter to continue…
//Process finished with exit code 0

1 Like
2

I face the same issue as you when I run similar code.
Have you tried running with sudo?
For some reason that works for me.

If that doesn’t work, tack on the following code after your call to cuda_hello:

cudaError_t cudaStatus = cudaGetLastError();
printf("%s\n", cudaGetErrorString(cudaStatus);

It may be that your device is not visible to CUDA. The above code should tell you if that’s the case.