I just noticed that a cudaMemGetInfo() call takes serveral seconds (~5-10 s) on some setups and while it is very quick on other setups. All systems are running Windows 7 x64 and CUDA 5.5.
- Dual GeForce GTX 680
- Single GeForce GTX 680
- Dual GeForce GTX 285
- Sinlge GeForce GTX 580
Any ideas what the problem mighty be? Is this a known problem?