Cuda and DirectCompute interop bug

I believe you guys created a bug in your Cuda-DirectCompute interop (the last driver version that worked for me was 314.22). I’m currently using 331.93, which is broken, along with with all the previous versions I tried after 314.22.

Details on my app:
Single threaded (the main thread, Cuda and Direct3d are all using the same thread).
I’m using a Cuda B40C radix sort to sort some particles into the appropriate screen space tiles, but after that, when I write them to a UAV in a compute shader, they turn into a flickering mess. (This works 100% perfect on version 314.22 and versions before that).

I tried adding cudaDeviceSynchronize() after the cuda sort (although I never did this before), and it was still broken.

If there’s something I’m not understanding, I’m all ears!