Program crashes when using Unified Memory

Hi! I am new to CUDA programming.

Whenever I try running a program that uses the Unified Memory, it stops responding and after 10 secs, the whole system freezes, forcing a hard reboot.

Event the CUDA example for ‘UnifiedMemoryStreams’ crashes the same way. No Error Code is returned when calling ‘cudaMallocManaged()’, the system just freezes.

‘cudaMalloc()’ and ‘cudaMemcpy()’ work as expected.

Specs :

  • Nvidia GeForce 920M (GK 208) [Compute Capability 3.5]
  • Cuda Toolkit 8.0
  • Driver : NVIDIA 390.65 (Windows 10 / 64-bit)

The Program was compiled in Visual Studio 2015, Release/Debug 64 config.