Try a different ISA slot. But, yeah, it does sound like the card is going bad.

Tried both ISA slots. It is having the same issue in either slot as far as I have noticed.
I also can't see any visible signs of damage on the card. Caps look alright as far as I can tell without studying them with a magnifying glass.
The thing I find most weird, is that the SIMMCONN seems to contribute nothing. Makes me wonder if it's the chip that communicates with the memory that is faulty or something.

Have you tried cleaning the cards edge connector with a pencil eraser? Might also want to try cleaning the connection on the ISA slots themselves.

For the ISA slots, get a piece of printer paper and fold it so it is thick enough to have a bit of resistance when inserted into the slot. Then move it back and forth in the slot. If you are not wanting to try the whole move back and forth thing, you can always just insert/remove a whole bunch of times. The move back and forth method is a lot quicker. When you take it out it will most likely have a lot of black residue on it from the paper cleaning off the oxidation on the connectors.

You may have to do this multiple times with different pieces of paper before the paper stops picking up oxidation/dirt from the slot connectors.

Bad connections can do really funny things.

It could also be a bad solder connection on the card itself.

It could also be your power supply not giving the correct voltage. What power supply are you using and how old is it?

I will have to try to clean the ISA slots. I've cleaned the card contacts with Akasa TIM-Clean, and I've done the pencil eraser trick before.
The power supply is an AOpen AO350-12AHNF 350W. Not sure exactly when it's from, as I got it from my boyfriend. He'd had it sitting around for years.
Best info I can find online suggests it is from 2005 or so.

I'd recommend getting out a magnifying glass and looking at the solder connections on the back of the card for any that look like they have cracked, or any that have an obvious separation between the pin and the solder. Also, check the legs of any surface mount components. It's very common for one of the tiny legs on the main chip (or a memory chip) to get bent, and if it touches another it will cause all sorts of issues. Check any jumpers on the board too, in case they are loose... I had a problem with a GUS a couple days ago that was caused by a few loose jumpers that had been damaged.

Also, look for any visible marks or scratches on the board and follow them closely to see if they cross any traces or lead to any damaged components.

Erratic problems may be some kind mechanical issue... by that I mean it will be noticeably affected by applying pressure to the card or moving it. Broken solder joints, damaged traces, bent pins, or bent over (damaged) capacitors cause issues like this.

I think I have finally found the problem.
Some pins on the corner of the large CT8903 chip are bent. They are probably just touching, and hopefully not broken.
Added a link to picture of the bent pins. Are they likely to break if I try to bend them back into position?

https://drive.google.com/open?id=1NP-lSBVhQ_y … ApgIpCbCeuvSy73

Take a fine sewing needle and carefully bend them back (going from bottom to top in small steps) - chances are good that might fix your card. I've had a similar issue on a Hercules Dynamite 128 in the past which is working nicely to this day.

Okay. Now I am even more confused.
I appear to have managed to unbend the pins as well as I'm likely to manage. A macro shot with my Panasonic camera actually makes the "repair" job look surprisingly effective.
I still get the same error in Windows, but when running the diagnose util, it now does start as it should, it appears to check out perfectly and it plays "AWE Synthesized music" perfectly.
Could this be a driver related issue rather than a hardware related issue?

So the DOS test program is using the built in ROM for the test I am pretty sure.

Windows is still showing 0 or 2048KB for the RAM?

Currently, it is reporting 2048KB ram in both the diagnose program and in Windows. I've had about a good dozen reboots now, and it reports 2048KB every single time.
Midi playback with the 2MB creative GMGSMT soundfont works perfectly. So, currently I see this a 1/3 victory. At least the card is usable for gaming.
However, it still will not detect additional RAM installed on the SIMMCONN, and I've tried a good few memory modules of different sizes now.

Can you post some good close up pics of the card?

Right. Pictures of front and back of card, and a closeup of the pins after unbending.
There is a scratch on the traces going from some of the pins near the bent ones, but I think it is only superficial, and that it doesn't actually cut through the traces.
Perhaps I am wrong, and that is the source of the problem? Anyone know the pinout of the chip? That might make it easier to figure out.

https://interruptrequest.files.wordpress.com/ … awe64_front.jpg
https://interruptrequest.files.wordpress.com/ … /awe64_back.jpg
https://interruptrequest.files.wordpress.com/ … 10/p1010608.jpg

Well, not seeing anything obvious. Those pins you unbent do go to the ROM, so it makes sense that those being unbent fixed it being able to play from the ROM.

I was hoping to see a missing capacitor or resistor going to the RAM chips but it looks like everything is present. I'll try to take out my 4390 soon to compare.
It could be that one of the RAM chips has failed and is not being detected.

The scratch definitely looks superficial.

Would a dead ram chip prevent it from accessing memory on the SIMMCONN though? I suppose it might not look for more memory if it detects less than the full 4MB?

I'm not sure... Is your Simmconn the original or one from the revival project?

It is from the revival. I guess it could be that none of my simms work with it.