I'd normally suspect bad/dirty drives (are the drives know good/working, how
exhaustively did you test them on another system)? If that were the case I
would recomment giving then a good cleaning - my ImageDisk has a pretty good
"clean heads" function where it will "scrub" the heads back and forth on a
cleaning disk.

But... The fact it got a bit further in booting when the drive was connected
to the untwisted part of the cable raises some other questions:

Are you sure the drives are jumpered correcly. The standard PC configuration
only supports two floppy drives, and the BOTH have to be jumpered as the
second drive (1 if 0,1,2,5) or 2 if (1,2,3,4). The actual select driving
it is determined by the position on the cable. (0-1)A: if at the end past
the twist, or (1-2)B: if at the more central untwisted connector.

Does the motherboard you are using support swapping drives/booting from B:
I've seem some that can swap the drives, maybe yours is doing that... but
it would think it would try the normal config first, and the drive at the end
of the cable would be tried first. The fact that yours got further as B:
seems very unlikely.

Any chance your mainboard doesn't use a "normal" cable?

You mention "controller" - is this a separate floppy controller? Is there
an FDC on the mainboard? (if yes, make sure it's disabled when using the
ISA one, and/or try using just it instead).

Sounds like you have another system that can accept floppy drives... try
booting these drives and cable on that system - that would at least let you
know the are working and configured correctly.

Is that CPU running as write-through or write-back?
I have one of these boards and it needed minor modifications for L1 WB to work.

That's a good point .. if might not even be floppy related, perhaps the system is dying for some other reason.

Although.. if it's passing all the POST tests...

I'd suggest setting everything tp be as simple as possible, lowest speed, no (or as simple as possible) cache,
anything not absolutely needed to boot from floppy disabled etc.

If you can get it to work in a simpler configuration, then you can work at figuring out what is causing problem.

If L1 WB isn't working right, then L1 invalidation/flushing will not work right, and one of the first places where this shows up is ISA DMA including floppy.

The situation is described in Jan Steunebrink's upgrade page here: http://www.steunebrink.info/amd5x86.htm where it talks about needing a DOS 5 or 6 floppy.

Symptoms like this can also be caused by a missing / interrupted or damaged "DRQ2" line:
Re: Repairing a "Yoke 401-386SX-08" mainboard / 386 Split BIOS Question

Hi all,

Many thanks for the input here. I've been lurking in this forum for 15 years and knew if I posted, the community would help!
So I have finally fixed this issue. It appears a couple different things were causing the erratic behavior.

It would appear most of my floppy cables are beginning to go bad. They worked when installed in their respective systems, but then when removed to check on the build in question, they either wouldn't work at all, or would display erratic behavior. My guess is one of the leads is going bad, and reorienting the cable is enough to get it to malfunction. I did manage to get one of my cables to test good across all of my systems and in different orientations.

The other issue was mentioned here as well. After disabling only the internal L1 cache (I didn't see a setting in the BIOS to set to WB), the system had no issues reading or writing to the drive. Now I believe I read somewhere this can be fixed using a bodge wire, if anyone here has a diagram or pictures if this repair, please send it my way. Otherwise, I am satisfied to resolve the floppy issue. The headaches were real with this one.