GPU not supported (CUDA successful)

Hello experts -

i just installed Deep Learning Studio today, with fresh install of nvidia cuda 9.2 (older 9.0 not supported for my gtx 1070).

cuda devicequery shows this:
deviceQuery.exe Starting…
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 1 CUDA Capable device(s)
Device 0: “GeForce GTX 1070”
CUDA Driver Version / Runtime Version 9.2 / 9.2
CUDA Capability Major/Minor version number: 6.1
Total amount of global memory: 8192 MBytes (8589934592 bytes)
(16) Multiprocessors, (128) CUDA Cores/MP: 2048 CUDA Cores

however, when i restart deep learning studio - it shows that it cannot support GPU:
Starting Deep Learning Studio…
Checking GPU support…
GPU not supported
skipping cifar dataset download!

i already downloaded the 2 DLL and put in /mxnet folder - doesnt seem to work.

any other ideas i can try?

i have to say this is a great product btw, especially if i can get gpu to run like everyone else.

best regards,

Jason, Deep Learning Studio is compiled with Cuda 9.0, so if that is missing from the system, then GPU support will not work. With latest version you dont need to use 2 DLLs mentioned in earlier threads.

Cuda 9.0 should be supporting your GTX 1070. You can try uninstalling CUDA 9.2, install CUDA 9.0. Skip the driver installation if possible or just update the driver separately after 9.0 installation.

I think latest Nvidia-driver and CUDA 9.0 combination should work.

thanks Rajendra - absolutely glorious now. interesting that it starts of CIFAR example with very low GPU/CPU usage… and then it fires up later.

just a few points to add -

  • i used DDU (display driver uninstaller) to remove nvidia drivers cleanly before cuda/ driver can be installed successfully.
  • i installed cuda9.2 and then 9.0 successfully but it couldnt recognize.
  • this last run, i removed all cuda (leaving old driver that came with 9.2), restarted, installed cuda 9.0, ignore incompatible driver warning - and DLS recognized GPU!