cuda 4.0 + cufft launch failure

with CUDA 4.0 both driver and toolkit
I get a kernel launch failure with - cufftExecC2C(plan, d_idata, d_idata, CUFFT_FORWARD);
consistently on the C2050.
I use centOS 5.5 kernel- 2.6.18-194.32.1.el5 #1 SMP x86_64

Any help with this, will be much appreciated !!
Thanks