Direct3d slowing down other threads

There are a few possibilities that spring to mind:

• You're using a single core CPU (doubtful).
• The program's affinity mask restricts it to using one core (doubtful, can check with Task Manager).
• Your physics thread has some form of synchronisation that causes it to wait for the rendering/main thread.
• One of the two threads is doing some heavy lifting that makes use of multiple cores, perhaps through libraries or system calls, that bottlenecks the other.

Beyond the above reasons and assuming I haven't overlooked anything obvious, it's difficult to say without seeing any code or the program run.

As a note, you ought to double check to make sure that the time wasting white loop is actually indeed wasting CPU time. The common compilers are generally smart enough to optimise out code that has no side-effects.

• You're using a single core CPU (doubtful).

Also

• You're using a dual-core CPU with a cache shared by both cores.

Thanks for the replies. None of these things seem to be the case as far as I can tell. Back to banging my head against the wall.

You haven't posted much info about the issue, e.g.
What amount of slowdown are you measuring?
Is there normally communication between the two, but you disabled it for this test?
What kind of hardware are you using?
Are these your only two threads?

Drivers from AMD and Nvidia (probably Intel too) are heavily multithreaded in order to reduce the overhead of D3D command submission that's visible to your rendering thread. It's very possible that they're fighting with your threads and causing contention. At work we always leave one or two cores available for the driver to use, and we try to make sure that we Sleep on idle threads to give some time back to the OS.