Posted Fri Jan 11, 2013 11:26 AM
A new test to focus on the USB and avoid the video delays. The on/off state of the NUM-, CAP-, & SCROLL-LOCK LEDs on a keyboard are controlled by the OS. I reprogrammed an adaptor to act as a keyboard. Pressing a switch transmits a NUM-LOCK key press. The OS processes that and sends back the NUM-LOCK state (on or off). The adaptor then receives that and sets it's LED acts as the keyboard NUM-LOCK LED. It might have been nice to have a test with a joystick, to keep it apples to apples, but I would contend that to the USB it is just packets of data to move from point A to B.
The measured round time, switch press to LED turn on/off, was usually 2ms and with an occasional 3ms. One thing about USB is it uses 1ms time frames and transmissions are at 1ms intervals, and I take this was the cause of the measured time to be in whole 1ms intervals.
That test was an unloaded system. Tried again with 99-100% CPU usage, and still got mostly 2ms with the occasional 3ms. So not affected by a busy CPU.