I got this error message too. But actually I got this only when I’m using 2 different threads for the GPU computation and the rendering. The rendering thread is at 30MHz and the GPU thread (where I’m doing all the CUDA calls) is at 1000MHz. The only item used by the rendering thread is a float array (located on the host memory) which is copied by the GPU thread (no cuda call and no direct interaction with OpenGL via a buffer).
And when I’m using one single thread at 1000MHz (I’m doing the rendering every 30 iterations or so then) everything’s fine.
Any clue? Thanks.
edit: I’m using CUDA 1.0 with a 8800GTX on a Dell Precision 690