We installed a server with 4 TESLA C1060.
The sample code seems to pass the tests but they are very slow. 2s on my geForce 8600GTS against 6s on the server for the oclScan. Actually a simple code that only creates a context, takes 3s on the server. (no time on my card).
Even worst, my code now works randomly on the server : I’ve just added and cout<<nbGPUDevices<<endl and it worked.
So I don’t know what should I do to check my installation or what shoul I try to reinstall. The current config is
Ubuntu 64 9.10 (g++ 4.3) OpenCL 1.0 CUDA 3.0.1 driver 195.17 (gpucomputingsdk_3.0-beta1_linux.run)