Question about nbody.exe vs oclnbody.exe speed


I’ve Just installed recent gpu computing sdk, on my Vista x64 and tried to run some of the examples, I’ve noticed that some of the OpenCL runs much slower, for example CUDA nbody.exe runs at approximately 27fps and 190 GFLOPs and the OpenCL oclNbody.exe runs much slower.
What causes that, are OpenCL slower on nvidia or is it just my PC or those example differs so much that I shouldn’t compare them?

Ups I just found similar thread here , sorry about that I don’t how I did missed that for a first time :/

I’m using a 2008 vintage Macbook Pro and I find the same is true. It seems to apply no matter what demo you run. I can the Black-Scholes demo and with the same number of options found that CUDA was 5x faster than the OpenCL version. Nbody seems like a MUCH bigger difference but it’s also one of the only demos with a OpenGL and OpenCL component.