This may seem banal but have you tried different drivers?

Doornkaat wrote on 2020-05-17, 12:12:

This may seem banal but have you tried different drivers?

I didn't, mainly because the FastVoodoo2 4.6 are recognized as best drivers for windows 98.

That being said I did some more testings and this is what I found:
On advanced properties there are 2 options for Direct3D: force trilinear texture filtering and automatic mipmapping.

With both options disabled, both options enabled and just the first option enabled the card worked fine on 3dmark except sometimes I get the white screen during the load of texture rendering speed.

With automatic mipmapping enabled and the other disabled the card went nuts on 3dmark: second test wrong colors, Fill Rate With Multi-Texturing with graphical glitches (while the one without didn't show issues), I also had wrong colors and glitches during those tests: Point Sample Texture Filtering Speed, Bilinear Texture Filtering Speed, Trilinear Texture Filtering Speed.

Then I went to the option screen and disabled the option, did another 3dmark and everything was right (still white screen).

I'm more convinced that there is something wrong with the card because just after I tried the surely working card and even with the option enabled I didn't have the graphical glitches or the white loading screens.

I also updated the file on gdrive with the last updates.

Yeah, that doesn't sound like too much a driver issue.
I think the chips are still avaliable. With advanced smd soldering skills you should be able to swap them.

Doornkaat wrote on 2020-05-17, 13:33:

Yeah, that doesn't sound like too much a driver issue.
I think the chips are still avaliable. With advanced smd soldering skills you should be able to swap them.

advanced smd soldering skill that I do not own, any advice where I could find them?
Thanks

Having the card reworked professionally is likely going to cost more than sourcing a new one so that's probably not an option.

I would not be comfortable working on somebody else's card with my limited equipment because it takes a lot of improvisation, but if you ask with your locale maybe you'll find a member who'll do it for you.
Otherwise maybe there's a hackspace or a repair cafe in your area?

Doornkaat wrote on 2020-05-17, 17:40:

Having the card reworked professionally is likely going to cost more than sourcing a new one so that's probably not an option.

I would not be comfortable working on somebody else's card with my limited equipment because it takes a lot of improvisation, but if you ask with your locale maybe you'll find a member who'll do it for you.
Otherwise maybe there's a hackspace or a repair cafe in your area?

Noone that I know of unlucky, for repairs I always rely on a user of this forum who brought back to life another v2 already, I was wondering mainly which chip (or passive element?) could be the culprit, because I don't have other 3dfx chips but I have plenty of memory chips and also resistors and stuff like that I suppose could be found easily.

This is what I noticed:

I can use powerstrip only to change the clock of the card, but I'm quite convinced that it doesn't really change the clock since I tested the card at 80mhz and at 100mhz, I got the very same results on 3dmark, not a single difference about fps.
The overclock tab despite being enabled on 3dfx menu doesn't appear.

Most likely the ICs are OK. Gently poke all TMU/FBI pins and see if any of them move. Then check for broken traces, corrosion, missing/chipped passives.

sdz wrote on 2020-05-17, 22:17:

Most likely the ICs are OK. Gently poke all TMU/FBI pins and see if any of them move. Then check for broken traces, corrosion, missing/chipped passives.

The tmu\fbi pins seems ok, I took detailed pictures of both tmu and fbi, if you are willing to take a look (https://drive.google.com/open?id=1dBwhqJyrhoU … hzlwlwVulqgnoQe).
The only weird thing is the FBI chip who has some gunk\corrosion on one side (it is most probably just gunk):

I changed the system to a more robust 440bx, I noticed that the card has apparently the clock of 93 mhz by default, I tried to underclock it to 85mhz and 90, everything works fine (except for that damn white screen during load that can be seen in one of the video), at 93mhz despite having a huge fan blowing air on it it crashes with black screen.

Already looked at the pictures. The pins can be loose and look fine. You can only tell when poking them.

sdz wrote on 2020-05-17, 23:10:

Already looked at the pictures. The pins can be loose and look fine. You can only tell when poking them.

This is what I was thinking. I've had a few cards lately where I've found loose legs\pins that looked fine but moved when poked.

To narrow the problem down, look for any nearby visible damage, scuffs or scrapes. Then check the pins closest to the damage. If there isn't any damage, at least start with the corners. Poke them very gently to see if they move at all (they should not). Be extremely careful not to bend them, because it isn't fun straightening them. Having to solder them again is bad enough. 🤣

I checked the pins with a small insulin needle, they all seem ok and not loose...
Is it a mad idea to desolder both the tmus and fbi chip to see if there is any sign of corrosion under them?

Just to update the situation, on the 440bx configuration, even in SLI, the card works apparently fine except for the damn white screen during the loading.
I did test the card as slave and then as master, at 800x600 no issues while at 1024 it gives many more white screens during the loading (processing textures) it also had a freeze for like 1 second during the benchmark, but no bad colored textures or visual glitches.
On quake 2 also worked fine.
The performance are ok, they are almost doubled from single card to sli (while with apollo 133a they were the very same if not slower).

I still believe something is wrong with this card but since I can't identify where the problem could be I'm going to give up.

Thanks everyone for the advice.