Maybe it was hardware acceleration in the particular machine with the particular video driver you use?

Can you try the test on an entirely different video setup just to prove it can get through, even if the numbers cannot be compared?