CUDA SDK 4.0 Release DLL VS 2008 Heap access violation

Just upgraded to CUDA SDK 4.0 Release from RC2 and my program stopped working. It starts to complain about the heap access violation on random CPU code that uses pointers, including c++ std library. Further checking the Modules in VS debugger, I found both nvcuda.ll and cutil.dll have warning sign with “The module didn’t load at the default load address”. Can someone help?

Following is the system configuration:

OS: W7 64-bit
CUDA SDK: 4.0 Release
CUDA Diver: 270.81

Thanks a lot!