UDMA cables vs speeds

[p***@pluto.rain.com]

Should an 80-wire Ultra-ATA data cable intended for use at UDMA66
or UDMA100 also work at UDMA133, or does the higher speed require
different cabling?

Background: I've been experiencing occasional UDMA ICRC errors on
the IDE port of a PCI SATA/IDE controller card using the VIA VT6421
chipset, and I've noticed that "atacontrol mode" reports the port
speed as UDMA133. The data cable is brand new, but it came with
a UDMA100 card. (I got much worse results with a different 80-pin
cable, which looked similar but may have been intended only for
UDMA66.)

I've found several similar reports in the forum and mailing-list
archives, with no definitive solution but quite a few "it works
for me" responses, and one observation that the VT6421 seems to
work well with some drives and not so well with others. It occurs
to me that the difference might, perhaps, involve the DMA speeds
(which aren't mentioned).

[p***@pluto.rain.com]

This is for the archives. The original posting is here:

http://lists.freebsd.org/pipermail/freebsd-hardware/2011-June/006704.html

Post by p***@pluto.rain.com
Should an 80-wire Ultra-ATA data cable intended for use at UDMA66
or UDMA100 also work at UDMA133, or does the higher speed require
different cabling?

Answer: as nearly as I've been able to determine, there are only
two kinds of PATA data cables: the original 40-wire variety that
work for PIO and for UDMA33, and the newer 80-wire variety that
should work for all modes.

Post by p***@pluto.rain.com
... I've been experiencing occasional UDMA ICRC errors on the
IDE port of a PCI SATA/IDE controller card using the VIA VT6421
chipset, and I've noticed that "atacontrol mode" reports the
port speed as UDMA133. The data cable is brand new ...
I've found several similar reports in the forum and mailing-list
archives, with no definitive solution but quite a few "it works
for me" responses, and one observation that the VT6421 seems to
work well with some drives and not so well with others. It occurs
to me that the difference might, perhaps, involve the DMA speeds

DMA speed definitely seems to be a factor. The errors seem to have
been completely eliminated after I used

atacontrol mode ad8 UDMA100

to limit the maximum transfer rate.

Throttling the DMA speed is not an ideal solution, and I still have
no clue _why_ it doesn't work reliably at UDMA133, but this does seem
to have gotten around the problem.