. Advertisement .
..3..
. Advertisement .
..4..
I get the “cuda driver version is insufficient for cuda runtime version” issue when trying to run an example source code in my project. Here is the detail of the error I got the error message:
"cutilCheckMsg() CUTIL CUDA error :
kernel launch failure : CUDA driver
version is insufficient for CUDA
runtime version."
Please give me some advice to solve this problem.
The cause:
When you choose (using the nvidia-settings utility) the built-in Intel GPU and run the deviceQuery script, you encounter this problem, which is associated with the choice GPU mode (Performance/Power Saving Mode).
Additionally, libcuda also causes this error. Even when the versions given by nvidia-smi precisely match, is not discovered. This library is a component of the nvidia-drivers package (x11-drivers/nvidia-drivers on Gentoo, nvidia-driver-latest-cuda-libs on CentOS).
Solution:
You can do as the following suggestion to solve this error:
Let’s access to this developer site to install the driver again and it will run well.
http://developer.nvidia.com/object/cuda_3_1_downloads.html#MacOS
Or you can set up the most recent driver or utilize an outdated toolkit in order to compile your program.
It is important to make sure that your driver version matches, or exceeds, your CUDA Toolkit Version.
For 2.3, you will need a driver with 190.x, 3.0 requires 195.x, and 3.1 requires 256.x. However, anything beyond the fifth number of five is fine, e.g. For 3.1, 258.x is required.
You can verify your driver version either by running the deviceQueryDrv sample or going into the NVIDIA Control panel and choosing System Information.
Download an updated driver from http://www.nvidia.com/drivers.