Hi there,
I googled this quite a lot, but I could not find anything related, so here I am.
I’m trying to install CUDA 7.5 on a win7-64bit platform with 2 GPUs: a Quadro2000 and a NVS300.
I have been able to install CUDA 7.5 and have it running on the Quadro with the monitor connected to it, but my code has some long calculations and it creates problem with the monitor refresh (my matlab calling the cuda code via mex crashes), so the idea is to have the monitor connected to the NVS and leave the Quadro for CUDA.
Problem is if I install CUDA 7.5 with the latest drivers (353.90) the Quadro works fine but my NVS is not recognized. In device manager I get this mesg: “windows has stopped this device because it has reported problems”.
If I try to install the drivers for the NVS (341.90), then both cards are recognized and able to drive the monitor, but my CUDA does not work and I get the following error from deviceQuery: “CUDA driver version is insufficient for CUDA runtime version”
I also tried to extract the driver for the NVS without running the installer, and update the driver manually from device manager only for the NVS card, but the outcome is the same.
Is there a way to install different drivers for each cards without messing the CUDA setup?
Your NVS 300 is a compute capability 1.2 GPU. Such GPUs are no longer supported by current drivers or CUDA toolkits.
You cannot install two different drivers in this case.
If you want to use both GPUs, you will need to install a driver that supports both, e.g. the 341.90 that you mentioned, and you will not be able to use CUDA 7.5 You could install CUDA 6.5 and use that. CUDA 7.0 and 7.5 do not support cc 1.2 GPUs.
Hi Tech,
I am trying to install CUDA 8 for NVS300 GPU. It installs but when graphics driver installation stage arrives, it says that detected card is not supported however you can continue installation. I complete installation and when I start running tensorflow I get following error:
[b]I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stream_executor\dso_loader.cc:119] Couldn’t open CUDA library cudnn64_5.dll
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stream_executor\cuda\cuda_dnn.cc:3459] Unable to load cuDNN DSO
[/b]
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stream_executor\dso_loader.cc:128] successfully opened CUDA library cufft64_80.dll locally
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stream_executor\dso_loader.cc:128] success
fully opened CUDA library nvcuda.dll locally
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stream_executor\dso_loader.cc:128] successfully opened CUDA library curand64_80.dll locally
For NVS300, you must use CUDA 6.5
You CANNOT use CUDA 8.0 with the NVS300.
However you probably cannot get Tensorflow to work with CUDA 6.5, so your best bet is to get a newer (cc3.0) GPU since Tensorflow requires cc3.0 or higher GPU.