I have been measuring the performance of a USB 2.0 device we have developed
and a driver based on the code provided in the Walter Oney book and
discovered some strange things.
1. On dual processor machines, the throughput is 50% greater when one of the
processors is removed. This is under Windows 2000 SP4.
2. On some machines, the throughput is more than twice as much under USB1
than under USB2. This is generally with USB1 ports integrated on the
motherboard and with a PCI USB2 card. This has only been observed under
Windows 2000, never under XP.
I wonder if anyone else has observed similar behaviour, or has ideas as to
how I could investigate further.
Regards,
John Hall