[Farshad]

Originally Posted by zaphod

12 is the slower method, and is supported in all cuda platforms (it's available for cuda 1.0). For MAC OSX for example we'll support mode 12 (other modes have a crash issue but we are trying to fix it).

13 is faster than 12
14 is faster than 13

If you don't have exclamation points the selected mode is safe and is running.

This is the same for FFTW if you install the DLL downloaded from FFTW.org. In that case mode 8 is faster. In nebula3 pro we'll implement an automatic diagnostic for selecting the best algorithm (it was already implemented, but for TIMED item only).

For CUDA: you should install
1) the proper NVIDIA driver. This is true for Windows, in Mac OSX the step is not applicable.
2) the proper CUDA toolkit (our installer could simply copy the official DLL, but we think it's better if you install the latest version available from their site).
3) Our Cuda bridge, which is a DLL located in your windows/system32 folder.

Thanks for the clarfication. I already have the latest driver, toolkit and bridge. So seting both to 14 seems to be my best bet. y the way where should I see the exclamation marks in case there are any?

Cheers