Razer357 wrote on 2023-01-08, 19:36:

(3dfx mode only, DirectX mode is fine).
...
Return to Castle Wolfenstein / Star Trek Elite Force / Quake 3 do not seem to exhibit any banding issues, but I think it is because they use OpenGL.

Someone here might have seen this before but based on these statements, I think it's software. If it were me, I'd try different drivers first, and finally a clean reinstall of the OS before any hardware operations.

Funny thing is, I have a Voodoo 4 4500 as well. If I use that in the same exact system there is no banding.

I do have an alternate system I think I could test the Voodoo 5 in, I might try testing it one more time in that one.

Ok I just did more testing, changed drivers back to the last official beta ones, and the banding is gone from every glide game except Unreal Tournament 99. At this point I think your are correct Maestro, it appears to be some issue with the glide implementation in Unreal Tournament 99, as I can no longer reproduce any issues in any other glide application. Not sure why UT99 has this issue, but I can switch to the D3D driver for UT99 if I have to.

Thanks!

Since the Voodoo4 doesn't show the same bug, it has to do something with the dual GPU configuration of the V5.

Yep, that's a SLI issue.

Hmm, maybe try underclocking to 143 MHz or lower and see if you can tease out another clue. You have to use the VSA100 Overclocker by KoolSmoky with the Voodoo5 to ensure that both chips receive the clock change.

https://www.3dfxzone.it/enboard/topic.asp?TOPIC_ID=9203

Ok I underclocked it, the bands gradually go away the lower I go. At 137 they are completely gone.

It seems that UT99 with the 3dfx driver stresses the card more / uses the card differently than other titles. I have no issues in Quake III arena for example, or if I run UT99 in D3D.

Does this indicate that my card has a defected component?

Theres def a problem.

Try this driver, does something differently.

Ok I used the new widescreen drivers and I am getting a few more fps in UT99 (maybe 5-10 more), so it appears to be a faster driver.

Unfortunately the stripes are still there in UT99, and are now appearing in other glide apps, so it appears the new driver is making it worse.

I rolled back to the last official 1.04.01 beta drivers and the banding is gone from every glide app except UT99.

My theory is that something is out of whack on the board, but it is small enough that most apps don’t have a problem with it.

I am leaning toward recapping the board, as that is the easiest hardware fix to begin with. Here are more system details:

CPU: AMD Athlon XP 3000+
RAM: 512mb SDRAM
Motherboard: Elitegroup 7VTA3 v6.0
Video card: Voodoo 5 5500 AGP
Sound card: Audigy 2ZS

Well, if it's coming back with different drivers and respondent to underclocking then it sounds like a hardware issue. Like you said, it's probably small enough that it only manifests under certain conditions. The only thing left to try on the software side, and I'm not even really sure this is a thing, is to make sure that UT99 doesn't have one of those glide dll files in the root of the game directory, which might explain why it remained buggy while others weren't with the beta drivers.

It really does sound like hardware though, so what about this, and I don't even know if this is right, but what if when you had lowered the MHz you created a lower electrical current condition which might suggest something like a connection somewhere is barely hanging on, loose, or dirty, and the lower current is able to be fed through the weakened connection unimpeded. Maybe carefully inspect the board with a magnifying glass first and hopefully the repair is as easy as touching up the solder at some location. Sorry, it's all just guess, and I really don't know much about how the subsystem makes SLI work. Good luck.

For sure its a hardware issue.

Most glide apps run on glide2, ut99 is one of them.

Opengl goes through glide3 and since you said you have no bands there.
This driver uses a wrapper for glide2, so ut99 runs on glide3, that was my "worth a shot".

It is interesting though. You literally can see how VSA100 renders the image in SLI mode. Might be worth to document it. Can you test if 1600x1200 looks any different?

Yea its worth a try.

It will disable digital sli and use analog, maybe smthing changes.

Thanks all, I plan on doing some recapping this weekend and we will see what happens.

Here are some screenshots at various resolutions, interesting to see the internal sli at work.

Interesting, so it seems different approach to SLI is not exclusive to 1600x1200 resolution and relies on how VSA100 can efficiently divide vertical resolution.

Might be completely unrelated and nothing to do with your current issue but your board has some suspect spots.

Thanks, I saw that, cleaned with 70% IPA, seemed to clean up ok.
I was able to decap the whole board, was actually easier than I I thought. Next is recapping.

I already have the parts, thanks to the thread posted by AkBKukU. He even made a mouser cart with all of the caps.
Voodoo 5 5500 AGP Compatibility Problem

Will see what happens after the recap.

Good luck! Hope it fixes it.

Well, unfortunately, that did not fix the issue, even though I feel better knowing it has new caps.

I've narrowed my testing after the recap and found these observations, using the newest Voodoo5 Widescreen Driver:

• Downclocking to 150mhz makes the issue go away entirely, something can't keep up, maybe a memory chip can't keep up with the clocks?
• Only the color red seems to be affected in these bands
• The bands do not go away when switching to single chip mode, probably meaning either the first VSA100 chip is bad or both are good and a memory chip is bad
• I ran Video Memory stress Test v1.7, it can only scan the first 32mb of ram due to how the Voodoo 5 is made, but it found no errors.

Right now I'm thinking of swapping memory chips, but I don't have the magnification tool for that kind of work yet.