I have similar issues with a Voodoo 3 PCI.
The AGP version does a lot better, but I remember similar problem with it under the right circumstances.
Clocking down the Voodoo 3 with a BIOS mod makes it worse.
The FSB frequency also affected the problem, IIRC it got worse with lower FSB, like 50 or 66 MHz.
And like you said, on the PCI card it is noticable right after a cold boot. It takes a few minutes to disappear.
Depends on the video mode too.

I am using the AGP version, actually...
The "snow" has been there ever since I got it, but it got much worse only recently. Most of it is on the right side of the screen. And I've certainly never noticed the screen "shaking" up to now. Both seem to be most noticable in high-res text modes.
It also won't boot at 133MHz anymore, while I didn't have any problems at that speed earlier. The Voodoo3 init screen comes up as usual but the POST screen doesn't, and the only way out is to reset the BIOS the hard way. 100 and 112 MHz are still good.

I suspect the mainboard caps are in need of replacement, and I was planning to do that anyway, but if that's the reason I would have more instability problems, right?

Just swapped my mainboard, leaving all other parts the same... but the problem remains. 🙁

Tried different RAM, RAM timings, FSB speeds, but nothing changed... then I replaced the graphics card by a Matrox G400 and that fixed it. So the Voodoo3 is actually degrading somehow, but how did that happen? Could this be caused by running it at 75MHz AGP clock all the time?

That's sad to hear. Does the Voodoo have caps?

Did you try changing the PSU?

Mau1wurf1977 wrote:

That's sad to hear. Does the Voodoo have caps?

There are five small SMD caps on the card, all 10µ 16V. I guess there's no harm in replacing them.
On a closer look the screen "shaking" happens with the Matrox card, too, but it's so small it might be my imagination. With the Voodoo it's very noticable on 132x60 text mode, but apparently the Matrox does not support this resolution so I can't check that.

Logistics wrote:

Did you try changing the PSU?

I recently installed a new Seasonic SII so that should be good. I just had a look at the ripple voltages however, and it doesn't look right. On the 4-pin molex plugs the ripple levels are practically zero, while on the ATX connector there's 200mVpp noise on +3.3V, and 260mVpp on +12V! Maybe the caps on this board are busted, too... Or the PSU is defective somehow.

jwt27 wrote:

Mau1wurf1977 wrote:

That's sad to hear. Does the Voodoo have caps?

There are five small SMD caps on the card, all 10µ 16V. I guess there's no harm in replacing them.
On a closer look the screen "shaking" happens with the Matrox card, too, but it's so small it might be my imagination. With the Voodoo it's very noticable on 132x60 text mode, but apparently the Matrox does not support this resolution so I can't check that.

Logistics wrote:

When the decline was gradual then it is most likely the caps and since the card is less stable then it is almost certain the caps are bad. Those small caps are electrolytic and known to leak over time. Go for any high grade Jap caps of the same voltage with a higher rating for better overclocking or the same just to fix.

you can probably check for bad caps with the hair blowdrier.
when heated up bad caps could regain some function.
you say after cold boot there is problems. maybe you can find the bad ones with the hair drier.

I don't think I have a hair drier... but if the caps would improve with heat, does that mean that cold temperatures make it worse? I had a go with freeze spray but it doesn't seem to change anything.
Think I'll just replace the caps on the mainboard and Voodoo3, just to rule that out. If the high ripple on the ATX connector remains after that, there might be something wrong with the PSU.

Also here's a short video to show what this "snow" looks like. It seems to randomly misplace entire rows of pixels, and these rows are shorter (1-2 pixels) on the left side of the screen.
https://app.box.com/s/7wqmsr14m10n0b9my8cv

Note that you can see the screen shaking here a bit, but it's much worse on 132x60 resolution.

My voodoo 3 PCI issue is a little different.
I tried to capture it with my phone, you can see pixels blinking left of the column of three zero's, and a few other places.
This chip is clocked 95MHz. The problem is in text modes, and disappears after a minute or two.

Yes cold makes the caps go worse.
Seems like caps and or bad psu filtering, should also take apart the psu and check for bulging caps on the end filter or replace.
most pci and agp cards take the 3,3v for powering the memory chips straight from the bus 3.3v power.
If it is noisy / filthy it can cause memory corruption or the ram to turn bad in the end.
I hope you voodoo ram is still ok after replacing the caps.

good luck

Alright, I made a list of all caps and will order them ASAP.

@gerwin, seeing your video I'm not even sure if these artifacts are the same thing. On mine it seems much more random, but consistently worse on the right side of the screen. Also it does not disappear after a few minutes, it's just always there.

Something more unusual: I accidentally booted with 1/1 AGP divider on 120MHz and it seems completely stable. Now without fastvid, the snow in adtrack2 is completely gone! Enabling fastvid makes it much worse than before though, and now it also appears in games/programs which were snow-free earlier on 2/3 divider. But for some reason, it is possible to boot with 120MHz AGP clock! With 133MHz and AGP divider on 2/3 it does not boot at all, and the BIOS will beep long-short-short, implying a graphics card failure.

Been playing some GTA and Tomb Raider in Glide at sky-high AGP clocks and didn't notice any snow at all. So I think I can safely say the RAM and DAC are still good, only the 2D "part" is broken. I could still see the screen shaking a bit (with a magnifying glass!) but it's much less noticable than in text mode. I am worried this might be caused by my monitor (which wouldn't really surprise me, given its age and uptime). Will check with another monitor to be sure.

I hope the recap will fix anything but I don't have high expectations. If anything, a recap would improve the longevity of the mainboard, plus I can check if my PSU is faulty or not. In that case I'll just RMA it, it's still under warranty. Caps will be delivered on tuesday so we'll see soon enough 😀

Sounds like the same issue I had on every single 3dfx card I own. I posted about this multiple times in the past and even made a video. In games it's really most apparent in plain VGA mode, and it can be quite annoying. Thought it was my LCD at first but only 3dfx cards exhibit this.

I don't buy this being a cap issue considering that other vintage cards don't have this...

Alright, the caps arrived yesterday already (shipped from Texas on friday night! 😳) so I immediately recapped the Asus P2B and Voodoo3. I didn't look at the µF rating at all, just noted dimensions and picked the biggest cap that would fit on 😀
For the Asus board that meant:
1500µF -> 3300µF
1000µF -> 1500µF
1000µF -> 1200µF (between the ISA/PCI slots)
100µF -> 220µF
Also filled the empty spots with 1500µF, 6.3V caps.
On the Voodoo3 I replaced the 10µ electrolytics by 22µ solid polymer.

So, did it fix anything? The answer is yes.
- Screen shaking is pretty much gone, it's still there but I need a magnifying glass to see it now in 132x60 resolution.
- I am now able to run Doom with the Katmai 550 at 124 FSB and 1/1 AGP. Previously it would crash consistently in the same spot. Quake still pagefaults at this frequency though.
- With the Coppermine 800, it will now POST at 140 FSB and 1/1 AGP. Graphical artifacts are all over the place and it won't boot into DOS though.
- Booting with PC100 CL3 ram at 150MHz and 2-2-2 timings is now possible. Results in instant parity errors however when booting to DOS, but at least it doesn't start beeping outright. PC133 CL3 seems stable at this frequency.

All of these points are pretty minor, of course, but do show that system stability improved quite a bit.

On the other hand, the snow was still there, so I gave the 3dfx BIOS editor a try, as recommended by gerwin. (First of all, who the hell thought it was a good idea to place the Exit button (without confirmation!) where you'd normally expect the File menu!?)
I tried some different clock speeds and enabled triple buffering by default. At lower clocks it does start to do some really WEIRD things. For example, the text and background colours start changing randomly after every CLS in text mode. Yeah...
Higher clocks seem more stable somehow, and will allow slightly higher AGP clocks too. I managed to brick it a few times but that was easily fixed by booting with an S3 PCI card and reflashing.

I noticed that the BIOS I was using previously (v2.15.11-SD) had different RAM timings from what 3dfx recommended. These are:
tRAS = 7 (4 recommended)
tRC = 10 (7)
tRL = 0 (1)
I changed these to the recommended values. I was unable to set tBWL to 1 for some reason, it kept changing back to 2. I think this is a bug in the BIOS editor.
Clock speed was set to 1,1,68 (K,N,M) which results in 167.045MHz. Default was 1,1,58 (143.182MHz).
With these settings, 100MHz FSB and 2/3 AGP divider... The snow is gone! It comes back at higher FSB/AGP clocks, however, but this shows that it can be reduced and possibly even fixed completely using a modified BIOS. Will experiment some more with that.

edit: screen shaking appears to have returned with this BIOS 🙁... Still, it's not as bad as before.

I don't get this. Was this BIOS with the different memory timings you are talking about already on there or did you flash it yourself? I also don't get the point of trying to overclock the FSB to get rid of artifacts? 😀 You may want to run Descent to check if it's gone in any case, as the issue is particulary apparent in this game.

Considering that I got the exact same behavior on Voodoo3 3000, 3500TV, Voodoo5 5500 cards, this is enough for me to assume a hardware flaw in the 2D core and simply use a different computer/card for VGA games 😀

d1stortion wrote:

I don't get this. Was this BIOS with the different memory timings you are talking about already on there or did you flash it yourself? I also don't get the point of trying to overclock the FSB to get rid of artifacts? 😀 You may want to run Descent to check if it's gone in any case, as the issue is very apparent in this game.

Considering that I got the exact same issue on Voodoo3 3000, 3500TV, Voodoo5 5500 cards, this is enough for me to assume a hardware flaw in the 2D core and simply use a different computer/card for VGA games 😀

There used to be a different BIOS on there, I don't think I saved it however. A few years back I reflashed with the latest BIOS I could find online, which apparently had the different timings. I don't remember suddenly seeing any more artifacts back then, only a short while ago I noticed it got worse.

And of course I'm not overclocking to remove the artifacts 🤣. I am just trying to overclock it to the max, and fixing the artifacts at the same time. I realize posting about both topics in the same thread can get a little confusing. On the other hand, both seem to be closely related. In any case, I had hoped the recap would've helped with both issues.

I don't have Descent, so I'll see if I can find a shareware copy to try.

That is the video I made of it back then. Descent shareware will do fine yeah. If you are getting anything like this it's the same issue for sure 😀