Forefront Technologies Offers Risk-Free AmigaOne XE USB and DMA Fix

[downix]

Um... I'm not sure if I'm reading this right, but using a PCI-based IDE controller doesn't fix the DMA problems, in fact it would increase the cache-misses.  However, they could be saying that even with the fix, the onboard IDE within the VIA"s chipset is not the most stable, so they recommend an off-chipset controller, and in that case it sounds as if they have done a good patch job.

[downix]

>Cache coherency must be dealt with in software in
>the OS4 driver. I don't know why SiI's PCI chip
>would have more cache misses than Via's chip, can
>you explain this more?

Cache misses aren't the DMA bug, data corruption is.  The VIA's chipset only uses a small segment of the full DMA capability, which limits the data corruption which occurs over it.  Now, the SiI chip uses the full DMA spectrum, which causes more data corruption.

Even today people STILL don't grasp what the MAI problem is.  It isn't a cache miss, it's data corruption.