Hi guys, I am working on a Dell EMC with 8xA30 cluster, with
Driver Version: 535.86.10 CUDA Version: 12.2
The driver version and CUDA version should be the most updated as far as I post this issue.
However, one of the A30 (CUDA device id:7) failed to initialize CUDA no matter what version of driver and CUDA I installed (tried CUDA 11.7, 12.0, with their supported NVIDIA driver). While the other 7 A30 devices works fine, when I set CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6 (which excludes device id 7), CUDA initialized successfully. Here are some example run on cuda-samples:
export CUDA_VISIBLE_DEVICES=0,7
./bin/x86_64/linux/release/deviceQuery
./bin/x86_64/linux/release/deviceQuery Starting…
CUDA Device Query (Runtime API) version (CUDART static linking)
cudaGetDeviceCount returned 3
→ initialization error
Result = FAIL
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6
./bin/x86_64/linux/release/deviceQuery
./bin/x86_64/linux/release/deviceQuery Starting…
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 7 CUDA Capable device(s)
……
……
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
Peer access from NVIDIA A30 (GPU0) → NVIDIA A30 (GPU1) : Yes
Peer access from NVIDIA A30 (GPU0) → NVIDIA A30 (GPU2) : Yes
Peer access from NVIDIA A30 (GPU0) → NVIDIA A30 (GPU3) : Yes
Peer access from NVIDIA A30 (GPU0) → NVIDIA A30 (GPU4) : Yes
Peer access from NVIDIA A30 (GPU0) → NVIDIA A30 (GPU5) : Yes
Peer access from NVIDIA A30 (GPU0) → NVIDIA A30 (GPU6) : Yes
Peer access from NVIDIA A30 (GPU1) → NVIDIA A30 (GPU0) : Yes
Peer access from NVIDIA A30 (GPU1) → NVIDIA A30 (GPU2) : Yes
Peer access from NVIDIA A30 (GPU1) → NVIDIA A30 (GPU3) : Yes
Peer access from NVIDIA A30 (GPU1) → NVIDIA A30 (GPU4) : Yes
Peer access from NVIDIA A30 (GPU1) → NVIDIA A30 (GPU5) : Yes
Peer access from NVIDIA A30 (GPU1) → NVIDIA A30 (GPU6) : Yes
Peer access from NVIDIA A30 (GPU2) → NVIDIA A30 (GPU0) : Yes
Peer access from NVIDIA A30 (GPU2) → NVIDIA A30 (GPU1) : Yes
Peer access from NVIDIA A30 (GPU2) → NVIDIA A30 (GPU3) : Yes
Peer access from NVIDIA A30 (GPU2) → NVIDIA A30 (GPU4) : Yes
Peer access from NVIDIA A30 (GPU2) → NVIDIA A30 (GPU5) : Yes
Peer access from NVIDIA A30 (GPU2) → NVIDIA A30 (GPU6) : Yes
Peer access from NVIDIA A30 (GPU3) → NVIDIA A30 (GPU0) : Yes
Peer access from NVIDIA A30 (GPU3) → NVIDIA A30 (GPU1) : Yes
Peer access from NVIDIA A30 (GPU3) → NVIDIA A30 (GPU2) : Yes
Peer access from NVIDIA A30 (GPU3) → NVIDIA A30 (GPU4) : Yes
Peer access from NVIDIA A30 (GPU3) → NVIDIA A30 (GPU5) : Yes
Peer access from NVIDIA A30 (GPU3) → NVIDIA A30 (GPU6) : Yes
Peer access from NVIDIA A30 (GPU4) → NVIDIA A30 (GPU0) : Yes
Peer access from NVIDIA A30 (GPU4) → NVIDIA A30 (GPU1) : Yes
Peer access from NVIDIA A30 (GPU4) → NVIDIA A30 (GPU2) : Yes
Peer access from NVIDIA A30 (GPU4) → NVIDIA A30 (GPU3) : Yes
Peer access from NVIDIA A30 (GPU4) → NVIDIA A30 (GPU5) : Yes
Peer access from NVIDIA A30 (GPU4) → NVIDIA A30 (GPU6) : Yes
Peer access from NVIDIA A30 (GPU5) → NVIDIA A30 (GPU0) : Yes
Peer access from NVIDIA A30 (GPU5) → NVIDIA A30 (GPU1) : Yes
Peer access from NVIDIA A30 (GPU5) → NVIDIA A30 (GPU2) : Yes
Peer access from NVIDIA A30 (GPU5) → NVIDIA A30 (GPU3) : Yes
Peer access from NVIDIA A30 (GPU5) → NVIDIA A30 (GPU4) : Yes
Peer access from NVIDIA A30 (GPU5) → NVIDIA A30 (GPU6) : Yes
Peer access from NVIDIA A30 (GPU6) → NVIDIA A30 (GPU0) : Yes
Peer access from NVIDIA A30 (GPU6) → NVIDIA A30 (GPU1) : Yes
Peer access from NVIDIA A30 (GPU6) → NVIDIA A30 (GPU2) : Yes
Peer access from NVIDIA A30 (GPU6) → NVIDIA A30 (GPU3) : Yes
Peer access from NVIDIA A30 (GPU6) → NVIDIA A30 (GPU4) : Yes
Peer access from NVIDIA A30 (GPU6) → NVIDIA A30 (GPU5) : Yes
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 12.2, CUDA Runtime Version = 12.2, NumDevs = 7
Result = PASS
$ nvidia-smi
Fri Aug 11 10:42:06 2023
±--------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.86.10 Driver Version: 535.86.10 CUDA Version: 12.2 |
|-----------------------------------------±---------------------±---------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA A30 Off | 00000000:12:00.0 Off | 0 |
| N/A 30C P0 32W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
| 1 NVIDIA A30 Off | 00000000:13:00.0 Off | 0 |
| N/A 30C P0 30W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
| 2 NVIDIA A30 Off | 00000000:48:00.0 Off | 0 |
| N/A 31C P0 32W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
| 3 NVIDIA A30 Off | 00000000:49:00.0 Off | 0 |
| N/A 29C P0 31W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
| 4 NVIDIA A30 Off | 00000000:89:00.0 Off | 0 |
| N/A 31C P0 32W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
| 5 NVIDIA A30 Off | 00000000:8A:00.0 Off | 0 |
| N/A 30C P0 31W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
| 6 NVIDIA A30 Off | 00000000:C1:00.0 Off | 0 |
| N/A 30C P0 30W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
| 7 NVIDIA A30 Off | 00000000:C2:00.0 Off | 0 |
| N/A 30C P0 31W / 165W | 18MiB / 24576MiB | 0% Default |
| | | Disabled |
±----------------------------------------±---------------------±---------------------+
±--------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
| 1 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
| 2 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
| 3 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
| 4 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
| 5 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
| 6 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
| 7 N/A N/A 2305 G /usr/lib/xorg/Xorg 4MiB |
±--------------------------------------------------------------------------------------+
WARNING: infoROM is corrupted at gpu 0000:13:00.0
$ nvidia-smi nvlink -s -i 7
GPU 7: NVIDIA A30 (UUID: XXXXXXXXXXXXXXXXXXXXXXXXXXXX)
Link 0: 25 GB/s
Link 1:
Link 2: 25 GB/s
Link 3: 25 GB/s
$ nvidia-smi nvlink -s -i 6
GPU 6: NVIDIA A30 (UUID: XXXXXXXXXXXXXXXXXXXXXXXXXXXXX)
Link 0: 25 GB/s
Link 1:
Link 2: 25 GB/s
Link 3: 25 GB/s