Yes could be something with the bottom #3 or #4 but is hard to tell.
Being it has issues with all those OS am suspecting there is a IRQ/DMA conflict that only occurs when accessing the floppy drive which would be hard to detect when running a test as they do not check with certain items being accessed.
Curious if you have had this Gold AWE for a while or just purchased it off Ebay/etc ? If so could be #2
Added: pull the Tv Tuner, 3Com NIC and Voodoos out and see what happens.....

Horun wrote on 2021-03-17, 00:09:

Yes could be something with the bottom #3 or #4 but is hard to tell. Being it has issues with all those OS am suspecting there […]
Show full quote

Yes could be something with the bottom #3 or #4 but is hard to tell.
Being it has issues with all those OS am suspecting there is a IRQ/DMA conflict that only occurs when accessing the floppy drive which would be hard to detect when running a test as they do not check with certain items being accessed.
Curious if you have had this Gold AWE for a while or just purchased it off Ebay/etc ? If so could be #2
Added: pull the Tv Tuner, 3Com NIC and Voodoos out and see what happens.....

Yes, I purchased my AWE64 Gold in the local market flea market, and I haven't faced such problems for a while.
As I mentioned before, I have tested the configuration without any PCI cards (even the SCSI adapter, booting from IDE disk), and replaced the VGA adapter with simple S3 Trio 3D - issue still persists.

And one more thing I forgot to mention: EEPROM on CT4390 was reflashed with the correct dump - same result, issue still persists.

out of curiosity...

when you take a good floppy and read a known good textfile with the sb gold in the computer, visually inspect the text, can you see errors in the text file ?

weedeewee wrote on 2021-03-17, 10:32:

out of curiosity...

when you take a good floppy and read a known good textfile with the sb gold in the computer, visually inspect the text, can you see errors in the text file ?

Sorry for the photo taken from my phone...

With the AWE64 Value card installed I successfully copied the README.TXT from Win98 boot diskette.
With the AWE64 Gold card installed when I tried to copy the same text file named README2.TXT I received this:

Of course, I can't open these "files".
And the issue will appear with any working FDD drive, tested with NEC FD1231, EPSON SMD-300 and EPSON SD-800.

Also I have swapped the IRQs for COM ports and have started the serial tests in AMIDIAG again - received FIFO error on both COM1 and COM2 ports, like this:

reddrake wrote on 2021-03-17, 12:14:

With the AWE64 Value card installed I successfully copied the README.TXT from Win98 boot diskette.
With the AWE64 Gold card installed when I tried to copy the same text file named README2.TXT I received this:

It seems like the sb gold is somehow picking up on accesses to the io ports for the com port and fdd controller.
Wouldn't be surprised if it's a PnP issue.

Try using Unisound to initialize the AWE64 Gold instead of Creative's drivers.

Joseph_Joestar wrote on 2021-03-17, 12:36:

Try using Unisound to initialize the AWE64 Gold instead of Creative's drivers.

UNISOUND didn't help me. AWE64 Gold was initialized correctly, and playback works fine, but i still lose data while writing to diskette. And even if I disable the sound card by using /D key, the floppy issue still persists.

It seems that the I/O decoder on the AWE64 Gold is broken, and responds to port addresses it is not suppose to respond to. The affected components all reside at a very similar address range: COM1 is either at 3E8-3EF or 3F8-3FF, and the floppy controller at 3F0-3F7 is exactly between them. The AWE64 Gold is not supposed to affect transfers on those addresses if it is properly configured. I don't understand yet how this failure mode can affect floppy writing, but doesn't affect floppy reading. This would sound more like a DMA problem, but the serial ports don't use DMA for sure.

Can you test the AWE64 Gold in a different computer, preferably a system running the chipset (and thus the ISA signals) at 5V? As far as I remember, the 430HX Pentium chipset still used "proper" 5V levels on the ISA address and data lines, but the 430TX chipset uses 3.3V CMOS levels that happen to correctly interact with 5V TTL levels (what ISA is specified for), but fails on some cards that expect 5V CMOS levels (some revisions of the AVM Fritz!Card are known to have problems in this regard). if that AWE64 Gold disturbs floppy/serial access on a 486-class computer the same way, it is very likely that the card is broken. If it works fine in a 486-class computer, it might be that either the AWE64 or the mainboard is marginal on ISA signalling.

mkarcher wrote on 2021-03-17, 19:08:

It seems that the I/O decoder on the AWE64 Gold is broken, and responds to port addresses it is not suppose to respond to. The affected components all reside at a very similar address range: COM1 is either at 3E8-3EF or 3F8-3FF, and the floppy controller at 3F0-3F7 is exactly between them. The AWE64 Gold is not supposed to affect transfers on those addresses if it is properly configured. I don't understand yet how this failure mode can affect floppy writing, but doesn't affect floppy reading. This would sound more like a DMA problem, but the serial ports don't use DMA for sure.

Can you test the AWE64 Gold in a different computer, preferably a system running the chipset (and thus the ISA signals) at 5V? As far as I remember, the 430HX Pentium chipset still used "proper" 5V levels on the ISA address and data lines, but the 430TX chipset uses 3.3V CMOS levels that happen to correctly interact with 5V TTL levels (what ISA is specified for), but fails on some cards that expect 5V CMOS levels (some revisions of the AVM Fritz!Card are known to have problems in this regard). if that AWE64 Gold disturbs floppy/serial access on a 486-class computer the same way, it is very likely that the card is broken. If it works fine in a 486-class computer, it might be that either the AWE64 or the mainboard is marginal on ISA signalling.

Ok, now OP has me confused about what works with the sbawegold and floppy combo and what not, but no matter.
OP is using sbawegold on a VIA mainboard so the i430Tx idea is off the table. too bad, I kinda liked it.
Also from the first comment op tried the sbawegold in other computers without any glitch.
So... Beats me. 😀

weedeewee wrote on 2021-03-17, 19:25:

Ok, now OP has me confused about what works with the sbawegold and floppy combo and what not, but no matter. OP is using sbaweg […]
Show full quote

Ok, now OP has me confused about what works with the sbawegold and floppy combo and what not, but no matter.
OP is using sbawegold on a VIA mainboard so the i430Tx idea is off the table. too bad, I kinda liked it.
Also from the first comment op tried the sbawegold in other computers without any glitch.
So... Beats me. 😀

The VIA mainboard is newer than the i430TX, so VIA might have tried to pull the same 3.3V-only stunt on the ISA bus that Intel started with the PIIX4. It worked quite well for Intel, so I don't see why competitors that usually work under higher market pressure wouldn't also try to skip the costs for true 5V ISA signalling. So that idea doesn't seem completely of the table. I didn't read the OP well enough, so I didn't notice that the card was already tested in a different computer, which makes my question obsolete. The card obviously isn't completely broken.

I suppose the OP doesn't have a scope, but if a scope a present, scoping the ISA address lines A0..A9, especially checking the "high" level of those lines, might be a good hardware trouble-shooting step.

mkarcher wrote on 2021-03-17, 22:09:

The VIA mainboard is newer than the i430TX, so VIA might have tried to pull the same 3.3V-only stunt on the ISA bus that Intel started with the PIIX4. It worked quite well for Intel, so I don't see why competitors that usually work under higher market pressure wouldn't also try to skip the costs for true 5V ISA signalling. So that idea doesn't seem completely of the table. I didn't read the OP well enough, so I didn't notice that the card was already tested in a different computer, which makes my question obsolete. The card obviously isn't completely broken.

I suppose the OP doesn't have a scope, but if a scope a present, scoping the ISA address lines A0..A9, especially checking the "high" level of those lines, might be a good hardware trouble-shooting step.

True, though from looking at the datasheet, page 24, seems a lot of multiplexing going on, and looking at a photo online, I'm guessing those address lines are buffered with a 74HCT245.

reddrake, while you're at it, could you run this http://www.users.on.net/~fzabkar/BIOSutil/Pho … ix/nvram120.exe util and post or attach the output of it, probably not gonna show anything but maybe we get lucky.
run it in plain dos. like F5 boot dos.

weedeewee wrote on 2021-03-17, 23:24:

reddrake, while you're at it, could you run this http://www.users.on.net/~fzabkar/BIOSutil/Pho … ix/nvram120.exe util and post or attach the output of it, probably not gonna show anything but maybe we get lucky.
run it in plain dos. like F5 boot dos.

Here is a dump of NVRAM:

mkarcher wrote on 2021-03-17, 22:09:

I suppose the OP doesn't have a scope, but if a scope a present, scoping the ISA address lines A0..A9, especially checking the "high" level of those lines, might be a good hardware trouble-shooting step.

Fortunately, I have an old analog scope. Not the best choice, of course, but that's all I have. Also I can use my DSLogic U3Basic logic analyzer, if necessary.