CUDA 9.0 on macOS 10.13 (High Sierra)

Hi,

I just upgraded to macOS 10.13 (High Sierra) on a Macbook Pro with an NVIDIA GeForce GT 750M GPU. Using the CUDA Toolkit version 9.0 and XCode 8.3.3 I am able to compile the sample code, but running them (for example “deviceQuery”) fails with the message “CUDA driver version is insufficient for CUDA runtime version (35)”. Some more details:

Cuda compilation tools, release 9.0, V9.0.17
Apple LLVM version 8.1.0 (clang-802.0.42)
CUDA Driver Version is 9.0.214
GPU Driver Version 10.24.28 355.11.10.10.10.160

Any hint on how to solve this issue? Thanks!

reproduced +1 : see http://devtalk.nvidia.com/default/topic/1026022

This exact thing is happening to me. It would be nice to get a solution to this.

CUDA driver 9.0.222 (which I have) says it supports OS 10.13.x but I’m getting the same error as above still.

CUDA driver: 9.0.222
GPU driver: 10.26.6 355.11.10.10.15.102
OS: OS 10.13.1

Same Here.

NVIDIA GeForce GTX 780M 4 GB

The same fuckin’ problem

Hi All,

I’m having some trouble getting the CUDA samples to run, getting the following error for “deviceQuery”
Any advice would be greatly appreciated!

cudaGetDeviceCount returned 35
-> CUDA driver version is insufficient for CUDA runtime version
Result = FAIL

My Environment:
MacOS 10.13.6 (MBP 2017)
NVIDIA Web-Drivers 387.10.10.10.40.105
CUDA-Drivers 387.178
CUDA 9.1 Toolkit
XCode 8.3.2
GPU Card: MSI RTX 2080 Ti - Connected through eGPU RAZER X

CUDA preferences window shows:

CUDA driver Version: 387.128 (no GPU Detected)
GPU Driver Version: No version found
pwd: /Developer/NVIDIA/CUDA-9.1/samples

$ nvcc -V
nvcc: NVIDIA ® Cuda compiler driver
Copyright © 2005-2017 NVIDIA Corporation
Built on Tue_Dec_19_21:36:29_CST_2017
Cuda compilation tools, release 9.1, V9.1.128

$ kextstat | grep -i cuda
185 0 0xffffff7f86eab000 0x2000 0x2000 com.nvidia.CUDA (1.1.0) 4329B052-6C8A-3900-8E83-744487AEDEF1 <4 1>
here is the output of deviceQuery

$ ./bin/x86_64/darwin/release/deviceQuery
./bin/x86_64/darwin/release/deviceQuery Starting…

CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 35
-> CUDA driver version is insufficient for CUDA runtime version
Result = FAIL