The example app fails with an assertion because the allocation fails and return a null pointer. I can think of two reasons why that may happen:
(1) The app simply requires more memory than is available on the GPU. I am not familiar with the GT240, and I do not know how much memory is available on the card and how this compares to the app requirements.
(2) CUDA failed to initialize, therefore no memory can be allocated on the GPU. The most common reason for this is a mismatch between the CUDA runtime and the CUDA driver, in particular running a newer runtime on an older driver. If the problem occurs with CUDA 4.1, please make sure to run it with a matching driver (available from the CUDA 4.1 download page).
[later:]
Actually, the message “GPU with -1 SM(s) with Unknown Compute Capabilities” is very much indicative of CUDA not being initialized. Please check your software installation.
I run into the same thing last night. nppGetGpuComputeCapability() was returning -1, nppGetGpuName() was returning garbage, like in your case “(+T♥░·↑” and then any nppiMalloc*() was returning 0. The only nvidia driver I had installed was the one provided by Windows 7. After installing the driver package for my card from nvidia’s website (I have 8600M GT) everything start working. Install the driver and see if it helps.