My customer makes frame grabber pci cards using bt878. My customer's customers buy the frame grabbers and typically build systems where they hook the frame grabber up to a camera and (for example) have software that watches a production line and separates good from bad product as it goes by in front of the camera.
A few of my customer's customers want to switch to an Intel dh55hc motherboard.
The frame grabbers have been selling successfully for years, all the way back to 386. But with this particular model motherboard, we get dma overrun errors nearly every frame, which means the motherboard is useless.
The bt878 is quite sensitive to latency from when the grabber initiates a pci write cycle to when it's granted, and it's possible that the dh55hc is slower to respond.
I'm running clean XP, my customer's drivers and test app, although the failure is also seen with the de facto standard btwincap.sourceforge.net drivers and standard apcap or whatever the DirectShow thingie is.
I stumbled on a baffling factoid. I start a capture session, using my customer's driver and test app, grabbing from a DVD player as a test vehicle. It gets lots and lots of dma overrun errors. If I start a separate console window and run a program that does nothing more than
the dma overrun problem goes away. The capture in the capture app stops getting dma overrun errors.
Obviously it's not a fix, but perhaps it's a hint as to what the problem might be.