cudaMalloc hang when building x64 version binary

I have seen the same problem with an NVIDIA Titan X (Pascal 2016).

Switching it to TCC, any x64 CUDA app would hang on the first call to cudaMalloc() no matter the app. Switching the card back to WDDM or x86 would work around the problem.

Definitely not a great experience.

Hi txbob,

Do you know is there any progress on this issue? I only have a k20c and it does not support WDDM mode, that means I can not use it any more on win10. I got no money to buy a new one, and it’s not broken indeed…

I believe a fixed driver is in the works but has not been posted yet.

When it becomes available (keep checking at www.nvidia.com/drivers ), try a driver that is 377.48 or newer. Right now the newest driver is 377.35, but when a new one gets posted that is 377.48 or newer, it should have the fix, I believe. It should be posted in the next couple weeks.

Thanks a lot, looking forward build 377.48…