To follow up, I saw another post from somebody doing high speed microframe interrupts and getting 8MB/s throughput, so that sounds positive.
But is the issue i'm suffering from w/bulk transfers (i.e. occasionally they don't get serviced quick enough) going to be resolved by moving to interrupt endpoints? (Again, assuming my app has a separate high priority thread using multiply queued overlapped transactions to read the data out).
Does the kernel level driver (i.e. more real time) in windows handle the turn-around of multiple transactions? Or is that done in the user space (i.e. less real time)?
BULK endpoints dooesn't have dedicated bandwidth.So when you are using your PC the data rate may get decrease.
If you use ISOC high bandwidth endpoint you will get 24MBps dedicated bandwidth.Even if you use your PC and run some other applications also there won;t be any decrease in data rate.