kneiver wrote on 2026-07-02, 10:46:

Hello. I am receiving a lot of help from Vogons. I would like to thank the various members. I have two questions. […]
Show full quote

Hello.
I am receiving a lot of help from Vogons. I would like to thank the various members.
I have two questions.

1. I recently purchased a Voodoo 3 3000 AGP from a used goods market.

However, upon receiving the item, I noticed that two of the gold finger pins appear to be broken.
Looking up the AGP pin map online, it seems they are SBA0 and the VCC 3.3V pin(B15, B16 pin).
I understand that Side Band Addressing is not very meaningful on the Voodoo 3, but the 3.3V power supply pins are a concern.

Will there be any major issues if I use it as is?
(I asked AI, and they said that the remaining voltage supply pins might be overloaded by 8.4%, causing the card to overheat. Is this true?)

For the voltage supply pins and overheat thing, well I think that LLM hallucinated that, even if somehow the other supply pins were "overloaded" by 8.4%(How does that LLM calculated that, does it knows the exact PCB layout? The Voodoo 3 3000 generally use around 15 watts(https://www.techpowerup.com/gpu-specs/voodoo3-3000-agp.c3555) and the AGP standard says that the slot should be able to provide about 48 watts and almost 20 watts on the 3.3 Volt pins alone? 😁), that would not make the card itself overheat.
About the Side Band Addressing thing, while I'm not experienced with the Voodoo 3 itself as I stopped using Voodoo cards after the Voodoo 2 and went to NVIDIA cards, so I cannot be 100% sure, but few graphics cards actually use the Side Band, and most of those that use it only gain a small extra performance from it.

Also these pins according to the pictures seem to be far away enough from components to be easy to jump if they are required, my guess would be no, but someone more experienced with Voodoo 3 cards can confirm that.

kneiver wrote on 2026-07-02, 10:46:

2. I am currently using a Tualatin Celeron 1.4GHz on Tyan S1846(FSB 100mhz is max). […]
Show full quote

2. I am currently using a Tualatin Celeron 1.4GHz on Tyan S1846(FSB 100mhz is max).

Which Tualatin-S (512k) model would have similar performance to this CPU?
My goal is a gaming machine running Windows 98 and DOS; do I really need to switch to the 512k Tualatin model?
(Of course, to do that, I would need to purchase a later BX motherboard that supports FSB 133.)

The difference between the different Tualatin variants are mostly the FSB and the cache:

• Tualatin Celeron - 100Mhz FSB and 256Kb cache, higher latency than the normal, slightly worse prefetching method
• "Normal" non "S" Tualatin - 133Mhz FSB and 256Kb, normal latency
• Tualatin-S - 133Mhz FSB and 512Kb cache

The core performance is the same, the size of the data the program works with decides the performance difference. For DOS games I don't think there will be any difference in felt performance by upgrading to the non Celeron variants, for Win98 I would give a "maybe" for some of the later games.

Hello kneiver,

It seems the damage is not just mechanical but was due to current arcing between B15 and B16. If so then the Avenger GPU might have been damaged too.
Also, since trace to B15 is broken side band addressing with the chipset is not going to work correctly or work at all (If the gpu is not dead, of course).
If you have tools like a multimeter/ohm meter you should measure resistance between gpu vcc and ground.
The 3dfx gurus here can give you more tips and info hopefully.

Sorry for the delayed reply.
I apologize to the two people who left comments.

Here is an update on what happened next.

I installed the Voodoo3 into a system featuring an Aopen AX6B motherboard and a Pentium II 400MHz (Deschutes, unlocked) processor.

The Voodoo3 worked perfectly and passed the 3DMark 99, 00, 01, and Final Reality tests without any issues.

Consequently, I’ve concluded that the Voodoo3 card is in good working order.

P.S.
However, the AX6B motherboard died during the testing process. One of the four MOSFETs (which have black heatsinks) near the CPU slot became too hot to touch, causing the PCB to turn slightly dark/scorched. It won't boot up anymore.
I plan to attempt a repair by replacing the MOSFET.
I’m not sure if the Voodoo3 was the cause of the failure.

I usually use a Tyan S1846 for testing retro components, but I made an exception this time.
My confidence in Tyan products has grown even more 😀

Oh, and regarding the 8.3% increase in load—I believe the answer was likely derived this way:

Since there are thirteen 3.3V pins but one is damaged, leaving only twelve,
calculating 13 divided by 12 yields approximately 1.083333.

kneiver wrote on 2026-07-08, 11:03:

Sorry for the delayed reply. I apologize to the two people who left comments. […]
Show full quote

Sorry for the delayed reply.
I apologize to the two people who left comments.

Here is an update on what happened next.

I installed the Voodoo3 into a system featuring an Aopen AX6B motherboard and a Pentium II 400MHz (Deschutes, unlocked) processor.

The Voodoo3 worked perfectly and passed the 3DMark 99, 00, 01, and Final Reality tests without any issues.

Consequently, I’ve concluded that the Voodoo3 card is in good working order.

P.S.
However, the AX6B motherboard died during the testing process. One of the four MOSFETs (which have black heatsinks) near the CPU slot became too hot to touch, causing the PCB to turn slightly dark/scorched. It won't boot up anymore.
I plan to attempt a repair by replacing the MOSFET.
I’m not sure if the Voodoo3 was the cause of the failure.

I usually use a Tyan S1846 for testing retro components, but I made an exception this time.
My confidence in Tyan products has grown even more 😀

While I find it unlikely that the Voodoo3 was the cause of the failure, it might be worth it to take another picture of the damaged pins and see if further damage occurred to them, as the components of the card are disconnected from the pins, it is possible that the pins of the card can arc over in such a way that the card doesn't get damaged, but it shorts components of the motherboard. Unlikely, but I would still check to make it sure that this doesn't become a "jinx card" kind of situation.

MagefromAntares wrote on 2026-07-08, 11:11:

kneiver wrote on 2026-07-08, 11:03:

Sorry for the delayed reply. I apologize to the two people who left comments. […]
Show full quote

Sorry for the delayed reply.
I apologize to the two people who left comments.

Here is an update on what happened next.

I installed the Voodoo3 into a system featuring an Aopen AX6B motherboard and a Pentium II 400MHz (Deschutes, unlocked) processor.

The Voodoo3 worked perfectly and passed the 3DMark 99, 00, 01, and Final Reality tests without any issues.

Consequently, I’ve concluded that the Voodoo3 card is in good working order.

P.S.
However, the AX6B motherboard died during the testing process. One of the four MOSFETs (which have black heatsinks) near the CPU slot became too hot to touch, causing the PCB to turn slightly dark/scorched. It won't boot up anymore.
I plan to attempt a repair by replacing the MOSFET.
I’m not sure if the Voodoo3 was the cause of the failure.

I usually use a Tyan S1846 for testing retro components, but I made an exception this time.
My confidence in Tyan products has grown even more 😀

While I find it unlikely that the Voodoo3 was the cause of the failure, it might be worth it to take another picture of the damaged pins and see if further damage occurred to them, as the components of the card are disconnected from the pins, it is possible that the pins of the card can arc over in such a way that the card doesn't get damaged, but it shorts components of the motherboard. Unlikely, but I would still check to make it sure that this doesn't become a "jinx card" kind of situation.

Yeah, I think I'll give that a try. It’s unlikely the AX6B would just die like that, but I do have my suspicions.

I tried installing a Voodoo3 in the S1846 as well, and it ran perfectly fine.

Interestingly, the AX6B struggled to boot with a P2 400 installed, but it managed to boot up when I used a P3 700 (Coppermine).
I strongly suspect the CPU is the culprit. I think I'll stick to testing with Coppermine chips from now on. 😀

P.S.
Back when I had a faulty power supply (I believe it was failing to provide enough 5V power), other motherboards all failed the 3DMark test, but the AX6BC passed without issues.
After swapping in a new power supply, that graphics card (it was a GeForce4 series model) completed 3DMark without errors on all the boards.
Ever since then, I’ve trusted AOpen motherboards—even if I don't have any concrete technical reason for it.

I've got a huge stash of motherboards and my Aopen ones have proven to be very reliable, probably the most reliable brand in my stash. I'm a bit of an ABIT fanboi 🤣 but tend to now use my Aopen boards if I want stability. I do mainly use Coppermine or Tualatin chips though so all pretty low wattage. Hope you get the MOSFET sorted!

To give you a final update: I recently checked the AGP characteristics using utilities like PCILIST and PowerStrip, and it turns out that Sideband Addressing is functioning correctly.

The Voodoo3 appears to be in perfect working order.
Thank you.

kneiver wrote on Yesterday, 11:59:

To give you a final update: I recently checked the AGP characteristics using utilities like PCILIST and PowerStrip, and it turns out that Sideband Addressing is functioning correctly.

afaik Voodoo3 doesnt even support Sideband Addressing 😀

kneiver wrote on Yesterday, 11:59:

To give you a final update: I recently checked the AGP characteristics using utilities like PCILIST and PowerStrip, and it turns out that Sideband Addressing is functioning correctly.

The Voodoo3 appears to be in perfect working order.
Thank you.

Nevertheless, it's fixable...