Have you tried this website?
I had the same problems with my voodoo cards.
http://falconfly.3dfx.pl/voodoo2.htm
You have the mismatched drivers, the falconfly 4.6 and an athlon driver. Eventually i got them to work properly.

No dice, all drivers returned the same Glide error and none of them had dxdiag show the card as a working 3D accelerator. All the drivers installed fine at least, and device manager showed no issues. Just trying to get the card running solo before testing the SLI setup again. Might try one of my Win98 rigs. Seller got back up with me and said they were using Win98 with the latest ref drivers to at least get Glide working.

EDIT: I did notice this card has slightly different chips than my working card - see attachment for the working Diamond card I currently have, compared to first post photos of the currently non-working card

unless the card has been altered by adding memory or replacing it, it seems a bit odd to have two different brands of memory on one card
just my 2c

also, your working card seems to have a resistor added between R148 and R153, though no idea what the significance of that is.

I haven't been able to get the card stable enough to check the settings and read what the memory count is on it. This is supposed to be a 12MB card, but I know at least certain 8MB cards could be user-upgraded to 12MB, correct? Possibly the reason for the different brands?

Not sure on the resistor, but good catch, I didn't notice that on the first pass.

weedeewee wrote on 2021-04-01, 08:56:

unless the card has been altered by adding memory or replacing it, it seems a bit odd to have two different brands of memory on one card
just my 2c

It is not uncommon to find Voodoo 2 cards with RAM by two different manufacturers. Like in OP's case the RAM is always different between the TMUs and the FBI chip. This is most likely stock.

Doornkaat wrote on 2021-04-01, 18:31:

weedeewee wrote on 2021-04-01, 08:56:

unless the card has been altered by adding memory or replacing it, it seems a bit odd to have two different brands of memory on one card
just my 2c

It is not uncommon to find Voodoo 2 cards with RAM by two different manufacturers. Like in OP's case the RAM is always different between the TMUs and the FBI chip. This is most likely stock.

Like in OPs case, his working card has the same ram chips all over the board, yet the non working one has two brands. 😁

weedeewee wrote on 2021-04-01, 18:41:

Doornkaat wrote on 2021-04-01, 18:31:

weedeewee wrote on 2021-04-01, 08:56:

unless the card has been altered by adding memory or replacing it, it seems a bit odd to have two different brands of memory on one card
just my 2c

It is not uncommon to find Voodoo 2 cards with RAM by two different manufacturers. Like in OP's case the RAM is always different between the TMUs and the FBI chip. This is most likely stock.

Like in OPs case, his working card has the same ram chips all over the board, yet the non working one has two brands. 😁

Correct, but what's the point?

Doornkaat wrote on 2021-04-01, 19:03:

weedeewee wrote on 2021-04-01, 18:41:

Doornkaat wrote on 2021-04-01, 18:31:

It is not uncommon to find Voodoo 2 cards with RAM by two different manufacturers. Like in OP's case the RAM is always different between the TMUs and the FBI chip. This is most likely stock.

Like in OPs case, his working card has the same ram chips all over the board, yet the non working one has two brands. 😁

Correct, but what's the point?

Old men like me don't bother making points... There is no point. 😁

I have returned from my quest with the two Win98 machines, and... no luck. Actually went worse: The card was recognized on first boot as a PCI multimedia device, but when directing Windows to automatically pick up the drivers from the folder (used the ref drivers this time), Windows couldn't find them. Did it the manual way, and sure enough, Windows felt they were the wrong drivers for the device. Went ahead and installed them, rebooted, device manager showed the device as working properly, but no settings available in display properties, nothing in dxdiag, and no Glide found in games. Similar to the Win95 rig, Everest only found the device in device manager and not as a physical device on the PCI bus or anywhere else, which makes me think the card's device ID is missing or wrong. I'm sure that's not the problem in and of itself, but it makes me wonder if that doesn't hint at what the root of the problem is?

murrayman wrote on 2021-04-01, 19:43:

I have returned from my quest with the two Win98 machines, and... no luck. Actually went worse: The card was recognized on first boot as a PCI multimedia device, but when directing Windows to automatically pick up the drivers from the folder (used the ref drivers this time), Windows couldn't find them. Did it the manual way, and sure enough, Windows felt they were the wrong drivers for the device. Went ahead and installed them, rebooted, device manager showed the device as working properly, but no settings available in display properties, nothing in dxdiag, and no Glide found in games. Similar to the Win95 rig, Everest only found the device in device manager and not as a physical device on the PCI bus or anywhere else, which makes me think the card's device ID is missing or wrong. I'm sure that's not the problem in and of itself, but it makes me wonder if that doesn't hint at what the root of the problem is?

I recently had a similar problem with a rendition V1000 card which wasn't detected, or partially detected 😒
Turned out to be a cold solder joint.
Working fine now.

Deleted the initial reply to this topic. New info:

weedeewee wrote on 2021-04-01, 19:53:

I recently had a similar problem with a rendition V1000 card which wasn't detected, or partially detected 😒
Turned out to be a cold solder joint.
Working fine now.

Initially, I checked for cold solder joints, didn’t find anything on first pass. Used 2x and 11.5x magnification, swept the whole board, still didn’t see any issues. Tried the oven method (385*F, three ball foil stand, reflective foil up base, 8 min bake), still having the same issue - no change.

But I just finished cooking it a second time, and it's now KIND OF working. Still very broken.

Using the latest ref drivers. First thing is, it's now recognized as a 3dfx Voodoo FX (??? Does that even exist?). Display properties also only recognizes 4MB of the total 12MB of RAM, and Everest only shows 8MB (worth note: my working Diamond card only shows 8MB of total 12MB in display properties, despite Everest showing all 12MB). When testing the card under dxdiag, the spinning cube represents just fine in both software and hardware mode, but 3dMark 99 test has no textures at all, just polygons. But when running Glide API - in this case, Descent: Freespace v1.06 - the textures all show up fine, but the fps bog WAY down once a fair bit of action is going on screen, compared to the performance of my other Diamond card.

So, does this sound like a RAM issue, or a TMU issue? Or something else entirely? Photos of the aforementioned stuff below.

Just compared the display properties info to the working card: the working card shows 2 TMUs, whereas the broken card is only showing 1. Got it running on a Glide game (Descent 3) right now, and sure enough, the PPU and the top-right TMU are hot, but the top-left TMU is room temp.