CUDA version is insufficient for CUDART version.

Accelerated Computing CUDA CUDA Programming and Performance
1

I’ve installed the CUDA Driver 2.3.1a for Snow Leopard without any problems, and successfully built the sample programs.

However, whenever I try to run a program I get some variation on this error, for example:

./clock
cudaSafeCall() Runtime API error in <clock.cu>, line 72 : CUDA version is insufficient for CUDART version.

Anyway, idea what this is all about, and how I can fix it?

2

I have the same problem, when i send calculations to be performed in the GPU, i receive the message:
CUDA version is insufficientfor CUDART version.
someone knows how can i solve it?

ps: i use VS2008 windows xp, CUDA 2.3

3

I’m using WinXp wisth VS8
i’m compling with the emulator flag
i dont have nvidia card
i’m getting the same error after compling is succsefull.

any idea ?

can i even run it without an nvidia card ?
Thanks ahead.

Ps
really need help i’m doing a project for uni and need it

4

Hi, I had the same problem in Ubuntu 9.10 and I solved it by downgrading my toolkit version. I’m also using emulation and it works fine :) it seems like your SDK and Toolkit should have compatible versions. I’ve installed version 2.3 for both and works ok :)

Hope it helps

5

I am using visual studio 2008 with CUDA 3.0 in windows 7. When I tried to run sample code in emu mode, I encountered the same problem.

I have solved it now, with a different approach. That is, I deleted all those “CUDA_SAFE_CALL” function, and replace the with the naive function calling.

For instance, there is a “CUDA_SAFE_CALL( cudaFree(device_result));” statement in sample code which leads to the error.

I replaced it with “cudaFree(device_result)”, and it works fine now!

I hope that would work for you as well

:rolleyes:

Wenxiang Chen

6

I am using visual studio 2008 with CUDA 3.0 in windows 7. When I tried to run sample code in emu mode, I encountered the same problem.

I have solved it now, with a different approach. That is, I deleted all those “CUDA_SAFE_CALL” function, and replace the with the naive function calling.

For instance, there is a “CUDA_SAFE_CALL( cudaFree(device_result));” statement in sample code which leads to the error.

I replaced it with “cudaFree(device_result)”, and it works fine now!

I hope that would work for you as well

:rolleyes:

Wenxiang Chen