sp1D.cu cuFFT kernel error

On a 32-bit Windows XP system:

I replaced a GTX295 with a GTX480 and started getting some very unusual runtime errors on cuFFT calls. All the latest versions of the driver/toolkit3.1/SDK 3.1 were installed when the GTX295 version was working. Since having problems after swapping for the GTX480, I’ve reinstalled the driver/toolkit/SDK and the problem persists.

I am able to run the SDK programs, including ones that use the FFT.

Here is the error I am receiving:

cutilCheckMsg() CUTIL CUDA error: fftSmem1024D_kernel<FFT_FORWARD> execution failed
in file <D:/Bld/rel/gpgpu/toolkit/r3.1/cufft/src/accel/interface/sp1D.cu>, line 265 : invalid device function .

Alternatively, with an inverse FFT, I get:

cutilCheckMsg() CUTIL CUDA error: fftSmem1024D_kernel<FFT_INVERSE> execution failed
in file <D:/Bld/rel/gpgpu/toolkit/r3.1/cufft/src/accel/interface/sp1D.cu>, line 479: invalid device function .

The only thing I’ve found that is close was mentioned in:
http://forums.nvidia.com/index.php?showtopic=176207
but the actual problem seems to be totally different.

I would appreciate any ideas as to the source of this problem and solutions for fixing it. Thanks.

Problem solved.

I discovered that I was using an outdated version of CUDA build rules inside of Visual C++ which I suspect was not setup to properly build for 2.0 compute capability devices.

Thanks.

Problem solved.

I discovered that I was using an outdated version of CUDA build rules inside of Visual C++ which I suspect was not setup to properly build for 2.0 compute capability devices.

Thanks.