Ok, further but still very confused, I checked all connections that witchery said I should check between u20 and u1 on main board, none show continuity. Witchery also said I should check voltages on u20 which I will do next. I also checked all FBs and Ls in the areas of u20, u23. All had continuity. Not sure if continuity between u1 and u20 is good or bad? Witchery also mentioned u65 might be an issue as well as xtal. It didn't mention u23.

Board 2 ran and passed all memory checks, save one. u38. Witchery had me check a pin on that ram chip and it did have continuity with the corresponding pin on u1. Not sure what to try next. Board looks clean. I can't find any visible flaws with magnifying glass.

Just checked voltages on u20, ramdac chip. On main board. I am getting 5v on pins 9 and 61. No real voltage on pins 27, 41, 43, 50....
. hmm probably a bad u20?

quattro64 wrote on 2026-06-02, 22:32:

Just checked voltages on u20, ramdac chip. On main board. I am getting 5v on pins 9 and 61. No real voltage on pins 27, 41, 43, 50....

No voltage on 27, 41, 43 and 50 indicated interrupted power to the RAMDAC. There must be FB, L or low-value R nearby that is open and not conducting.

quattro64 wrote on 2026-06-02, 21:41:

Witchery also mentioned u65 might be an issue as well as xtal. It didn't mention u23.

As I said, Witchery doesn't know X-24 board, its chips and markings.

Ok, I will keep checking.

Ok, I was confused about u20 orientation and pinout, rechecked all connections between u1, u20 everything fine. rechecked all voltages to u2o pins, all fine. Looked at every FB, L ,C, and R on the main board. all measured as expected. How hard is it to find a ramdac chip for these cards? I think I could learn to use a hot air solder station.

thanks

quattro64 wrote on 2026-06-06, 14:14:

Ok, I was confused about u20 orientation and pinout, rechecked all connections between u1, u20 everything fine. rechecked all voltages to u2o pins, all fine. Looked at every FB, L ,C, and R on the main board. all measured as expected. How hard is it to find a ramdac chip for these cards? I think I could learn to use a hot air solder station.

Do you have an oscilloscope? Also, I'd like to see the log file.

No, no oscilloscope. I will attempt a repair, starting with xtal, then the ramdac. It should be one of those. I've checked all fb, l, c, r, and resistor arrays on the mainboard. It never gets past the errors to test the ram on that board. I will leave any ram repairs for later. It should run ok with minor ram issues, right? I did view the board with a flir, no ram chips are running hot.

Hi!
After months of research juggling work, family, and everything else, My Maxi Gamer 3D² is fixed.

I removed U13 with hotair gun, but it was hard for me to solder new 100Mhz DRAM chip with this tool.
First attempt to resolder was a mess 😒
Happily, pads were safe and I used solder iron to put this chip.

I tested pin to pin continuity between chip to each pad, soldering task was not perfect but all tests succeded.

I also noticed C113 and C158 were... missing. I don't know why or when but maybe when I used hotair gun for preheating they fell, maybe they already was in bad shape.
It's a real mystery—those capacitors were still there just a week ago!
Anyway, I didn't have 0603 in stock, I used 1206, not ideal but they did the job.
They aren't positioned perfectly straight, but they'll do the trick for connecting them to the pad underneath.

Now everything works perfectly, and I'm so grateful to the creator of Witchery for offering this program that lets us fix our beloved Voodoo cards, which sometimes haven't aged well.

This adventure is now over.

Hey, I'm back with quick question. Witchery doesn't say or suggest that I have anything wrong with my fbi chip, only possible connections to the chip. But is that also a possibility? It mentions the ramdac error u20 , it talks about the x-tal it mentions u65, which is the clock buffer chip. I've swapped out the u65 ,ramdac and verified that all its connections are working between it and the fbi, and also that it is getting the xtal xin signal from the clock. I've tested without any fb memory chips in place as I know sometimes that can get in the way of the ramdac working. I've visually inspected, poked, prodded, reflowed, measured everything with a dmm. All checks out. The only thing I haven't done is swap in a new fbi chip. I'm hesitant because it is a lot of work, and because witchery doesn't specifically say it could be broken.

below is from witchery:

Awesome software, helped tremendously to fix my card!

I had a spare parts voodoo 2 that I bought a long time ago already broken, with significant damage to some of the FBI chip pins. The whole chip was also twisted to the side in a way that lots of the pins were shorted together. But after discovering this program, I finally thought it was worth the effort to try to fix the card.

So after straightening the chip, some "delicate" repair for replacing the completely missing legs with magnet wire, and with the help of Witchery, the card is now working perfectly. Below is the picture of the card while I was soldering on and testing the frame buffer memory.

Are there any plans to ever look into doing TMU tests, or is the lack of documentation a total show stopper? One of my cards present symptoms of bad TMU and/or TMU memory or interconnects. It presents in MOJO with a full working 12MiB and launches Need For Speed 2SE in GLide but doesn't display the car's dashboard (which was why I upgraded to a 12 MiB back in the day). And afterwards MOJO says no texture memory and Bogus number of TMUs

First, reflow your card.

Oh, yeah, I should've mentioned, that I did reflow the card, and checked all TMU-memory connections with a multimeter 😀

H3nrik V! wrote on 2026-07-12, 20:26:

Are there any plans to ever look into doing TMU tests, or is the lack of documentation a total show stopper? One of my cards present symptoms of bad TMU and/or TMU memory or interconnects. It presents in MOJO with a full working 12MiB and launches Need For Speed 2SE in GLide but doesn't display the car's dashboard (which was why I upgraded to a 12 MiB back in the day). And afterwards MOJO says no texture memory and Bogus number of TMUs

You may like to look into the Youtube video called "How Software Helped Me Find A Hardware Fault! New Voodoo2 Debugging Tool!" by creator Bits und Bolts.
https://www.youtube.com/watch?v=BVhEIGM3n0o

Generally-speaking Witchery is the gold standard for Voodoo2 testing, but the purpose tool demonstrated in the above video may assist with some TMU memory tests in the meantime.

Environment variables which allow the disabling of one TMU might also help narrow the fault down.

**If** you are not confident in the robustness of your reflow, then be sure to give all the pins the Gentle-but-firm Push Test. The joint can be visually misleading at times.

Shreddoc wrote on 2026-07-13, 10:24:

You may like to look into the Youtube video called "How Software Helped Me Find A Hardware Fault! New Voodoo2 Debugging Tool!" b […]
Show full quote

H3nrik V! wrote on 2026-07-12, 20:26:

Are there any plans to ever look into doing TMU tests, or is the lack of documentation a total show stopper? One of my cards present symptoms of bad TMU and/or TMU memory or interconnects. It presents in MOJO with a full working 12MiB and launches Need For Speed 2SE in GLide but doesn't display the car's dashboard (which was why I upgraded to a 12 MiB back in the day). And afterwards MOJO says no texture memory and Bogus number of TMUs

You may like to look into the Youtube video called "How Software Helped Me Find A Hardware Fault! New Voodoo2 Debugging Tool!" by creator Bits und Bolts.
https://www.youtube.com/watch?v=BVhEIGM3n0o

Generally-speaking Witchery is the gold standard for Voodoo2 testing, but the purpose tool demonstrated in the above video may assist with some TMU memory tests in the meantime.

Environment variables which allow the disabling of one TMU might also help narrow the fault down.

**If** you are not confident in the robustness of your reflow, then be sure to give all the pins the Gentle-but-firm Push Test. The joint can be visually misleading at times.

Thanks, I'll look into that at some point. I'm pretty confident in my soldering skills - been doing it for 35'ish years - using a scalpel to check for robustness 😀

quattro64 wrote on 2026-07-01, 22:35:

Hey, I'm back with quick question. Witchery doesn't say or suggest that I have anything wrong with my fbi chip, only possible connections to the chip. But is that also a possibility?

It is. Witchery can not detect the type of the RAMDAC, so it assumes some most possible causes for that. If you already replaced (presumably for checked good ones) RAMDAC and clock buffer chip and checked all PCB connections it leaves FBI to be blamed.
Also, I'm pretty sure that Witchery recommends to re-run tests with '-v' option in case of unsuccessfull run. Did you try that? It can show more detailed info. Also, text log would be preferred instead of screenshots.

Also you can try to skip RAMDAC tests (you board better have the ICS DAC for that) by running with '-e3' option. It would move tests further, at least you could see other test for FBI performance.

quattro64 wrote on 2026-07-01, 22:35:

The only thing I haven't done is swap in a new fbi chip. I'm hesitant because it is a lot of work, and because witchery doesn't specifically say it could be broken.

It actually does:

FBI was listed and nothing else suggested helped.

makechu wrote on 2026-07-12, 18:35:

Awesome software, helped tremendously to fix my card!
...
So after straightening the chip, some "delicate" repair for replacing the completely missing legs with magnet wire, and with the help of Witchery, the card is now working perfectly. Below is the picture of the card while I was soldering on and testing the frame buffer memory.

Great to hear that! Another one saved.

H3nrik V! wrote on 2026-07-12, 20:26:

Are there any plans to ever look into doing TMU tests, or is the lack of documentation a total show stopper? One of my cards present symptoms of bad TMU and/or TMU memory or interconnects. It presents in MOJO with a full working 12MiB and launches Need For Speed 2SE in GLide but doesn't display the car's dashboard (which was why I upgraded to a 12 MiB back in the day). And afterwards MOJO says no texture memory and Bogus number of TMUs

There are plans but I've just got to the stage when I can reverse engineer TMU memory mapping. And I've paused it to move other projects forward.
I've made this debugging contraption:

Next step is to write some low-level tests, poke and prod at some pins with oscilloscope, put and remove RAM ICs and make some notes.
There are no documents about TMUs on Voodoo, so it's all uncharted territory and it'll take a lot of work.

quattro64 wrote on 2026-07-01, 22:35:

It mentions the ramdac error u20

I've looked into your screenshots, performed some tests on my side and this "error" could be false positive.
There's maybe an error in RAMDAC type detection logic. I'm working on it now.

May I ask here?
What log filename is better in you opinion:
1) 260717_22-26-28_witchery.log
2) witchery_260717_22-26-28.log