Try the floppy with a controller card.
Still nothing?

Reflash BIOS.

Clean pins, maybe oxide.

Quick caps check, smds around the fdd connector.

So ..trying to get this straight in my head ..
The Bios is the first running code ...it searches for any bootable medium..
Floppy or Hard drive ..
The search is for a 2 byte reference so 16 bit operating at this point..
Which the bios finds in both cases..be it floppy or hard drive.
The boot loader code is then loaded into memory and executed ...which in turn boots the kernal.
The bootloader must be the same from both sources.
In the case of from HD all goes correctly and the process completes..
If the source is floppy it goes pear shaped after the boot process starts.
In both cases the same bit of memory is used ..so this can be discounted as correct...so not a RAM issue..
Also can discount Bios problematic.
The source of the bootloader storage medium can also be discounted be it floppy or HD.
We are left with a faulty floppy drive or cable...
Both have been tested in that problems ocurr with 5.25 inch floppy ...or 3.5 inch floppy and each uses its own cable..
The process almost completes with the 5.25 inch floppy ...but totally fails with 3.5 inch floppy.
and both have been tested by hooking up to another PC again with their respective cables/ribbons.
I think that eliminates the drives to my satisfaction.
Curious that the larger format floppy gets further but eventually corrupts.
But it runs slower so the data is maybe less easy to corrupt.
It does sound like the data gets corrupted in transit...but less so at lower speeds.
2 different power supplies have been tried ...so I guess PSU is not the problem.
I'm thinking out loud here..
caps on the motherboard ..leading to noise spikes ...thus corruption....or damage to whatever buffers the floppy drives to the system.
I could change any and all electrolytic caps on the motherboard...but that wont fix the problem if it's dodgy silicon.
noisy power lines or damaged chipset/buffering.
The first option may be fixable ...the second definately isn't.
Or I could just cope with it as is.

All that said...the problem is only apparent when trying to boot ..
booted to the Hard Drive ..I can clean read the floppy contents...and that is what I find confusing and is frying my brain ???
Hmmm
rgds
VS

Just saw your reply Nexxen ...
good ideas and a lot of similarities with my ideas

Have you tried a different IDE + floppy I/O card?
I had to go through several until I found one, a Goldstar Prime 2C, that didn't corrupt data on a rather strange 486 board I was using earlier.
In my experience UMC controller cards tend to be more problematic on the whole. I normally trust Winbond controller cards, but I've had to chuck a few that weren't working right lately so I don't know what to recommend apart from Prime 2Cs.

Eep386 wrote on 2021-09-27, 23:49:

Have you tried a different IDE + floppy I/O card?
I had to go through several until I found one, a Goldstar Prime 2C, that didn't corrupt data on a rather strange 486 board I was using earlier.
In my experience UMC controller cards tend to be more problematic on the whole. I normally trust Winbond controller cards, but I've had to chuck a few that weren't working right lately so I don't know what to recommend apart from Prime 2Cs.

Not tried a controller card as yet Eep (but is on my to-do list) as this motherboard has on board FDC amd IDE.
I would have to disable these in the bios.

So ..I think it safe to assume...This is not a controller issue..
Disabled on board FDC in bios ..
Fitted a SCSI card with on board floppy FDC enabled.
connected my 1.44 floppy to this ..booted into DOS on the hard drive ..popped in a floppy and read the DIR listing no problem.
Rebooted with the floppy inserted..
It was recognised as bootable...but the exact same problems as with the on board FDC controller.
It hangs the computer....thus boot does not complete.
???

Incorrect/improper RAM timings or AT bus clock divider in CMOS Setup?
Bad RAM?

Oh, a common one: Video BIOS set to cacheable? This causes all kinds of weird errors. Not sure if it would help in your case but given how much trouble that has caused me in the past, I wouldn't discount anything.

Maybe a clue here ..who knows ?
I removed all the cache chips and tag ram..
No difference
I tried various wait states and DRam configurations in the bios ...
..not that there are many options to choose from.
No change
What did make a slight difference ..
I pulled the CPU ....I had no idea what it was as it was under a clamped on heatsink that came with the mobo.
It was another AMD DX4-100
not only does this give me chance to test the one from the gigabyte mobo....but it might prove ...something.
Well it sort of does.
OK it still wont boot from floppy but it does get a little further...in that instead of locking up with the rapidly darting cursor.
I can ctrl alt delete ...and a message now appears after .."Starting MS DOS..."
I get " Bad or Missing Config.SYS"
Error in CONFIG.SYS Line 0"

I assumed the motherboard was set up for the fitted cpu ...as it came with it...but maybe not ..
I need to find some jumper settings for this motherboard....and check them.
rgds
VS

Recap:

if used to boot, FDD doesn't work
if used after starting from HD floppy drive works: reads/writes/formats 100% ok

Right?

You have about right Nexxen
Got a brand new floppy from my small stash.
booted C: DOS...Then WIN
Created a System Floppy which appeared to go well ..
Rebooted and left the newly created system floppy in the drive.
No boot ..and the computer just hung (original DX4-100 now re-installed)
So I could create a system disk from win 3.11 but cannot boot from it.
Just to try writing to floppy from DOS I copied over CHECKIT.EXE from C: to A:
That also went well..
So ..I can read and write to floppy but simply cant boot from it.
This is a very odd and confusing problem I think.
rgds
VS

can you boot that floppy on another system? and run checkit from it ?

weedeewee wrote on 2021-09-29, 10:48:

can you boot that floppy on another system? and run checkit from it ?

Just tried it on a win98 build ...and yes boots and runs checkit no problem..

Try cleaning the bios' chip legs and the socket.

If the loading process fails, it must be some data transfer issue. The exact point it happens is difficult to determine.
Can you enable disable WB/WT for L2 cache? Cpu?

Also a bad BIOS string could be.
Maybe try another BIOS, I wouldn't know which but it's an idea...

This one is tough and I guess it's going to be something astonishingly simple.

Cant see an option to actually disable L2 Chache ...though I have tried removing cache chips a Tag physically..last night.
Problem persisted.
These are the options in bios for this mobo in chipset options.

Disabling all ports except FDD?
No lpt, com, ide.

I'd wonder what the difference is, in software, between booting from harddisk and booting from floppy.... ?

and yeah, if you've got another bios to try, or another bios image for your motherboard.. go for it.

Is copying files from the floppy drive on this motherboard as quick as on another one? I'm wondering if maybe once it's booted DOS allows for more retries before failing than the boot loader. I know I had a failing floppy drive where it would still work, but copying files from it was slow as it'd keep re-seeking to track 0 and back again. In this case, if there's something making communication from drive to processor unreliable, then maybe DOS does enough retries to get good data and the bootloader just fails. Drive, cable and controller have all been swapped out, so maybe a DMA problem?

snufkin wrote on 2021-09-29, 13:44:

Is copying files from the floppy drive on this motherboard as quick as on another one? I'm wondering if maybe once it's booted DOS allows for more retries before failing than the boot loader. I know I had a failing floppy drive where it would still work, but copying files from it was slow as it'd keep re-seeking to track 0 and back again. In this case, if there's something making communication from drive to processor unreliable, then maybe DOS does enough retries to get good data and the bootloader just fails. Drive, cable and controller have all been swapped out, so maybe a DMA problem?

Nice!

I was wondering, isn't there some buffer between fdd and data path?
Maybe getting corrupted before startup but as you say enough to reseek and send back to buffer?
Or some other ecc-type operation failing? (guessing)

With HD it'd be some LBA or Large setting wrong in BIOS, usually with CF cards.

Are there density settings on floppy drive? Maybe it needs one that dos doesn't?

Do you have another cpu, dx/2, to test?
Maybe the bus is too fast and corrupting the floppiy"s data transmission?
Maybe data isn't with correct divider in bios. Try tinkering with bus/2 /4 /8 ... or similar

Finally got the system to boot from floppy...
There is an option in bios options to disable CPU internal cache.
set to enable by default.
Disabling this allows to boot from floppy..
Faulty CPU ???
If it is the cpu at fault then I have 2 x DX4-100 cpus with exactly the same fault.
what are the chances ?

Vipersan wrote on 2021-09-29, 17:38:

Finally got the system to boot from floppy... There is an option in bios options to disable CPU internal cache. set to enable by […]
Show full quote

Finally got the system to boot from floppy...
There is an option in bios options to disable CPU internal cache.
set to enable by default.
Disabling this allows to boot from floppy..
Faulty CPU ???
If it is the cpu at fault then I have 2 x DX4-100 cpus with exactly the same fault.
what are the chances ?

You have one of the jumpers on your board relating to a Write-Back vs. Write-Through CPU set wrong (or the board is missing one of said jumpers)