T. Ment wrote:
The PCCHIPS M811 motherboard is known to have IDE UMDA errors at ATA-100
speed. Windows XP downgrades it to ATA-66 and that works. But DOS has no
way to downgrade it. BIOS can enable or disable UDMA per drive, but has
no option to limit the speed. BIOS just uses the highest speed the hard
drive reports.

https://forums.tomshardware.com/thre...v266ad.435007/

They advised flashing a BIOS from the similar ECS K7VTA3. I tried that,
but it didn't help.

For other reasons (to save an IRQ) I decided to put both IDE devices, a
hard drive and a cdrom on the same cable. Some people advise against IDE
slaves, but for a cdrom that gets little use, I don't see a problem.

Linux starts with the speed BIOS sets, but will downgrade when it gets
too many UDMA errors. It's hard to miss, it spits rude messages on the
screen. I have seen this many times.

But now, with both cable positions filled, the ATA-100 errors are gone.
I tried to cause errors with some large directory copying, which always
caused a downgrade before, but now, I get no errors at all.

Looks like filling both IDE cable positions somehow changes the cable
signaling to make this motherboard work right at ATA-100. But I have no
idea of the EE involved.

Just wondering if the experts do ...

You want 80 conductor wire for your ATA IDE cable.
This reduces crosstalk at higher speeds due to the interleaved
grounds on the cable.

The signal integrity on IDE looks awful. SATA is such
an improvement, by comparison (SATA being point to point,
not a shared bus, with better impedance control from end
to end as well).

"Proposed Annex D for Ultra DMA Mode 5
(Ultra ATA/100 Implementation guide)"

http://www.t13.org/Documents/Uploade...l/e00126r2.pdf

Having an 80 wire cable, with just one device on the end, should
give better looking signals than with two devices. As the "middle"
device is a stub capacitance and a reflection point. But because
the signals look like wavey gravy, it's really hard in this
case to say why this is happening. The purpose of doing lots
of analog analysis is to bound the conditions to be found
on the cable, and to provide a sufficiently good S/N to make
the damn thing work.

Paul