Hit another wall with this one...

I soldered in 4 known good RAM chips into the TMU positions. I still get a black screen on the Voodoo output when starting 3d applications... MOJO detects 2MB and 2MB and the output seems good.

I have a suspicion that one or more of the 4 RAM chips soldered to the FBI may be faulty...

Time to put it down again for a bit and clear my head.

current MOJO log (this is identical to my known good Orchid Voodoo 1 card)

1sst1Init Routines: InitCode $Revision: 8 $
2sst1InitMapBoard(): BoardsInSystem = 1
3sst1InitMapBoard(): vAddr:0xe6000000 pAddr:0xe6000008 Dev:0xf Board:0
4sst1InitRegisters(): Setting TREX-to-FBI FIFO THRESHOLD to 0x8...
5sst1InitRegisters(): Setting PRELIM FT-CLK delay to 0x8...
6sst1InitDacDetect(): Entered...
7sst1InitDacDetectICS(): Entered...
8dacWr(0x7,0xb)
9dacRd(0x5,0x79)
10dacRd(0x5,0x2e)
11dacWr(0x7,0x1)
12dacRd(0x5,0x55)
13dacRd(0x5,0x49)
14dacWr(0x7,0x7)
15dacRd(0x5,0x71)
16dacRd(0x5,0x29)
17sst1InitDacDetectICS(): Exiting...
18sst1InitRegisters(): Storing TREX0INIT0=0x5441
19sst1InitRegisters(): Storing TREX0INIT1=0xf420
20sst1InitRegisters(): Storing TREX1INIT0=0x5441
21sst1InitRegisters(): Storing TREX1INIT1=0xf420
22sst1InitRegisters(): Storing TREX2INIT0=0x5441
23sst1InitRegisters(): Storing TREX2INIT1=0xf420
24sst1InitSetGrxClk(): Entered...
25dacWr(0x7,0xe)
26dacRd(0x5,0x0)
27dacWr(0x4,0xa)
28dacWr(0x5,0x7b)
29dacWr(0x5,0x6c)
30dacWr(0x4,0xe)
31dacWr(0x5,0x0)
32sst1InitRegisters(): Setting up FAST DRAM Configuration
33sst1DeviceInfo: Board ID: Obsidian GE
34sst1DeviceInfo: FbiConfig:0x0, TmuConfig:0x11
35sst1DeviceInfo: FBI Revision:2, TMU Revison:1, Num TMUs:1
36sst1DeviceInfo: FBI Memory:2, TMU[0] Memory:2
37sst1DeviceInfo: Dac Type: ICS ICS5342
38sst1DeviceInfo: SliDetect:0
39sst1InitRegisters(): LFB Writes go through memory FIFO...
40sst1InitRegisters(): TEXTURE Writes go through memory FIFO...
41sst1InitRegisters(): exiting with status 1...
42sst1InitShutdown(): Shutting down SST-1 #0...
43sst1InitSetGrxClk(): Entered...
44dacWr(0x7,0xe)
45dacRd(0x5,0x0)
46dacWr(0x4,0xa)
47dacWr(0x5,0x41)
48dacWr(0x5,0x62)
49dacWr(0x4,0xe)
50dacWr(0x5,0x0)
51sst1InitShutdown(): Returning with status 1...

Ok, I've actually managed to get something on screen... "something".

3D Mark99 spits out some messed up stuff over the Voodoo output. I tested my known good Orchid and this is also spitting out garbage - so I must have some messed up drivers at the moment.

That aside, in DOS - the Tomb Raider demo only outputs a black screen 640x480 signal (I can see that the 640x480 signal is there on my scaler). I have tested the same demo in the same system with my known good Orchid V1 and it works as expected.

I feel like I'm so close on this one now...

It's alive... well sort of...

Finally got it running 3D apps but still have some texture corruption.

Time to put the card to bed again and pick it up again another day...

Lostdotfish wrote on 2025-11-07, 22:19:

It's alive... well sort of... […]
Show full quote

It's alive... well sort of...

The attachment PXL_20251107_221610287.MP.jpg is no longer available

Finally got it running 3D apps but still have some texture corruption.

Time to put the card to bed again and pick it up again another day...

Well that's some great progress, what changed?

Looks like the framebuffer memory is all okay now and you just need to track down some bad memory address / data lines on the TMU, not easy. It's something I've been through with a similar very clean looking Orchid Righteous 3D with similar texture memory faults: Re: What retro activity did you get up to today?

Check the pins on the SOJ memory chips since they've been resoldered, bridges can hide underneath as can pads without enough solder. Looks like you've only got a few bad address/data lines since games are looking good and mojo shows 2MB of memory. I don't know what the patterns in the texture could indicate though, but them being repetitive like that and all correct colours suggests maybe it's a bad address line?.
You've checked the soldering on the TMU so that should be okay, but you can see the pinout here: 3Dfx Voodoo 1 - Low level hardware information and diagnostics thread
Definitely your FT-DATA and TT-DATA lines are all good on the TMU since games are now working with most textures. The FBI is 100% since those patterns are just on textures.

Perhaps you could remove one TMU memory chip at a time to see which one gives the worst output, assuming games still work with a whole missing memory segment on the TMU.

Hopefully your fault isn't the same as my Orchid Righteous 3D card where the TMU was indeed the fault the whole time and taking one from a donor card got it working properly again.

Thermalwrong wrote on 2025-11-08, 05:00:

Lostdotfish wrote on 2025-11-07, 22:19:

It's alive... well sort of... […]
Show full quote

It's alive... well sort of...

The attachment PXL_20251107_221610287.MP.jpg is no longer available

Finally got it running 3D apps but still have some texture corruption.

Time to put the card to bed again and pick it up again another day...

Well that's some great progress, what changed?

I'd messed up soldering in the TMU RAM. I tried using solder paste (first time using it) and hot air reflow. it looked like it had gone well but on closer inspection I'd totally failed to flow one side of one RAM chip....

I'd also been using my Pinecil for solder work instead of my trusty Metal.

So I busted out the Metcal and cooked everything good and proper.

In the meantime, I put 2 of the old RAM chips on the Ati Rage Pro card I salvaged the replacement TMU RAM from. First I got major corruption. So I flipped out one RAM IC and put the 3rd one from the Orchid in. Card now tests fine with no errors. So I have 2 more known good RAM and have identified at least one bad RAM IC from the Orchid.

I'll test the last RAM IC on the Rage today then do a rolling replacement on the Orchid to see if I can identify the problem.

I'll reflow the TMU one more time first using my Metcal.

That's great progress.

I should also have mentioned, If You have a card with sockets for an edo memory upgrade, like s3 virge, trio, cirrus logic, You can
put the chips from the voodoo in there (after cleaning solder pads) and test for errors with VMTCE

Imperious wrote on 2025-11-09, 05:50:

That's great progress.

I should also have mentioned, If You have a card with sockets for an edo memory upgrade, like s3 virge, trio, cirrus logic, You can
put the chips from the voodoo in there (after cleaning solder pads) and test for errors with VMTCE

I've been doing the hard version of this... soldering them back into my ATi Rage Pro to test....

Identified 1 faulty module but I think I might still have another. It's tedious work swapping them onto the Orchid as they are very closely spaced making resoldering particularly problematic.

I did it!

Needed to reflow the TMU RAM again. It super hard on these Orchid cards as the modules are so so close to each other, it's hard to get an iron to them and then hard to visually inspect afterwards.

Anyway, it's alive again and soak testing as we speak. I'll solder mask up the trace repair and call this one done.

Great News. Another Voodoo 1 rises from the grave!

Super! How did you solder it – with a soldering iron or a hair dryer?

Postman5 wrote on 2025-11-13, 12:18:

Super! How did you solder it – with a soldering iron or a hair dryer?

I ended up using my knife edge tip on my Pinecil set at 400 degrees C and plenty of Amtech NC-559-v2

This one.

The profile of this tip allowed it to sit in under the RAM pins and just about make contact with the PCB pads. I basically filled the tip with solder on one side, then ran it as tight as I could. along the IC. Then dragged a few times to distribute the excess solder evenly and finally checked for and cleared any bridging.

Several long sessions of soak testing and it's running perfectly. Just waiting on some green solder mask so I can finish up the trace repair. Couldn't bring myself to chuck the blue mask that I have on it.