I’ve been trying to install cuda 6.5 for a couple days without success on what I consider to be
a pretty plain vanilla system. I am running;
- Linux Mint 17.1 (Rebecca ~ Ubuntu 14.10)
- Cinnamon 64-bit
- gcc 4.8.2 (originally had 4.9 but this led to various error messages so I changed)
To install cuda I run cuda_6.5.14_linux_64.run which leads to
You are attempting to install on an unsupported configuration.
Do you wish to continue? ((y)es/(n)o) [ default is no ]:
I’ve learned to say yes even though I don’t understand why it objects to my configuration (running
nvidia-smi gives):
Fri Jan 16 21:08:20 2015
+------------------------------------------------------+
| NVIDIA-SMI 340.65 Driver Version: 340.65 |
|-------------------------------+----------------------+----------------------+
| 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 750 Ti Off | 0000:01:00.0 N/A | N/A |
| 40% 31C P8 N/A / N/A | 469MiB / 2047MiB | N/A Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Compute processes: GPU Memory |
| GPU PID Process name Usage |
|=============================================================================|
| 0 Not Supported |
+-----------------------------------------------------------------------------+
The first time I tried the installation I accepted its offer to install a video driver which was a BIG mistake. Cinnamon got blown away somehow and I had to uninstall my video drivers and cuda 6.5 and start over. The driver I was using originally was 343.36 and I thought maybe that was causing the problem so I went to Nvidia 340.65 as shown above.
When the cuda installation is completed I run deviceQuery and get
CUDA Device Query (Runtime API) version (CUDART static linking)
cudaGetDeviceCount returned 35
-> CUDA driver version is insufficient for CUDA runtime version
Result = FAIL
I’ve tried running the install in a terminal while in Cinnamon and I’ve tried it by first stopping mdm but the result is always the same.
At this point I don’t what to try next…