My System:
OS: Ubuntu 18.04
GPU: GTX 970
CUDA: 9.0
cuDNN: 7
I’ve installed CUDA/cuDNN and when i run ./mnistCUDNN I got these results:
cudnnGetVersion() : 7301 , CUDNN_VERSION from cudnn.h : 7301 (7.3.1)
Host compiler version : GCC 6.4.0
There are 1 CUDA capable devices on your machine :
device 0 : sms 4 Capabilities 2.1, SmClock 1800.0 Mhz, MemSize (Mb) 4536, MemClock 2050.0 Mhz, Ecc=0, boardGroupID=0
Using device 0
Testing single precision
CUDNN failure
Error: CUDNN_STATUS_ARCH_MISMATCH
mnistCUDNN.cpp:394
Aborting...
I found in this link https://blog.pattapongj.com/install-tensorflow-with-gpu-cuda-support-on-ubuntu-16-04-42c90592855b that the minimum supported CUDA architecture is 3.0. However, I found in wikipedia that GC GTX 970 has Maxwell architecture with compute capability of 5.2.
Has anyone an explanation about that?
These are the results of ./nvidia-smi
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 390.87 Driver Version: 390.87 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GTX 970 Off | 00000000:0F:00.0 N/A | N/A |
| 40% 39C P12 N/A / N/A | 217MiB / 4536MiB | N/A Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 Not Supported |
+-----------------------------------------------------------------------------+
Thank you very much.