I installed CUDA 1.1 SDK and Toolkit the other day. I didn’t install the updated driver; for one thing, I don’t yet have a card, here.
I often use the Cygwin bash shell, but when I went to try a few simple codes, using the usual nvcc -deviceemu -o simple simple.cu etc. I receive only one message: “Internal Error”. If I give no arguments to nvcc, I get the usual messages, but apart from that it’s always the same.
Curious, I tried with the standard Windows console, and the programs built and ran fine.
I have now reverted to CUDA 1.0. Thanks to NVIDIA for keeping that available.