Well as I said before, I only know about PIO mode turning off UDMA - not about Multiword DMA. To quote: "for dvd rom - says - transfer mode - dma if available and current transfer mode - PIO????"
Otheriwse, there is a Windows function turning off DMA on a device if it produces too many CRC errors. You can find the info in the M$ knowledge base. There is no freely available patch for that. The reason mostly turns out to be a) misbehaving hardware on the IDE port b) dying devices or c) wrong/bad cables.
Your system may work fine, but some do not ... and sometimes the reason are simple incompatibilities. Everyone who had hdds before which worked with one of them always master other one slave, but not the other way round will agree:
From an EE point of view, some people say it is a miracle that IDE works at all, especially if you look at the restrictions the SCSI makers impose on their bus system ("3m with 4 devices but only 1.5 with up to 7" etc etc) and what stuff is sold for IDE really not conforming to the specs, like 90cm UDMA133 cables and so on.
I just dont trust IDE any more than I would eat it and really hope SATA will give less problems than this POC.