The VM with GPU passthrough with 2 L40s. The simpleP2P testing failed in the end.
[./cuda-samples/Samples/0_Introduction/simpleP2P/simpleP2P] - Starting…
Checking for multiple GPUs…
CUDA-capable device count: 2
Checking GPU(s) for support of peer to peer memory access…
Peer access from NVIDIA L40S (GPU0) → NVIDIA L40S (GPU1) : Yes
Peer access from NVIDIA L40S (GPU1) → NVIDIA L40S (GPU0) : Yes
Enabling peer access between GPU0 and GPU1…
Allocating buffers (64MB on GPU0, GPU1 and CPU Host)…
Creating event handles…
cudaMemcpyPeer / cudaMemcpy between GPU0 and GPU1: 21.07GB/s
Preparing host buffer and memcpy to GPU0…
Run kernel on GPU1, taking source data from GPU0 and writing to GPU1…
Run kernel on GPU0, taking source data from GPU1 and writing to GPU0…
Copy data back to host from GPU0 and verify results…
Verification error @ element 96: val = 5760.000000, ref = 384.000000
Verification error @ element 97: val = 5764.000000, ref = 388.000000
Verification error @ element 98: val = 5768.000000, ref = 392.000000
Verification error @ element 99: val = 5772.000000, ref = 396.000000
Verification error @ element 100: val = 5776.000000, ref = 400.000000
Verification error @ element 101: val = 5780.000000, ref = 404.000000
Verification error @ element 102: val = 5784.000000, ref = 408.000000
Verification error @ element 103: val = 5788.000000, ref = 412.000000
Verification error @ element 104: val = 5792.000000, ref = 416.000000
Verification error @ element 105: val = 5796.000000, ref = 420.000000
Verification error @ element 106: val = 5800.000000, ref = 424.000000
Verification error @ element 107: val = 5804.000000, ref = 428.000000
Disabling peer access…
Shutting down…
Test failed!
Is there a way for me to get more info about the failure?
Has anyone seen this problem ? If so, how did you fixed it?