Thermalwrong wrote on 2023-11-11, 02:22:

Wah, I've been buying broken 3DFX cards for a while to see if I can fix them. Today I was looking over a Voodoo 1 that I'd bough […]
Show full quote

Wah, I've been buying broken 3DFX cards for a while to see if I can fix them. Today I was looking over a Voodoo 1 that I'd bought last year and I could never get it to do anything. It would only report 0xDEAD for the TMU status with 57005 TMUs. I'd thought it was down to a broken pin since there were a few on this card - but now I know how to actually use mojo.exe to generate the SST.LOG file which gives actual information:

This 0xDEAD result in Mojo

Info for Voodoo board # 0:
=====================================================
Virtual Base Address: 0xe6000000
Physical Base Address: 0xe6000008
PCI Device Number: 0x9
Vendor ID: 0x121a
Device ID: 0x1
FBI Revision: 2
FBI Memory: 0 MB
FBI PowerOn Sense: 0x0
TMU PowerOn Sense: 0xdead
FBI DAC Output Color Format: 24BPP
Scan-Line Interleaved? No
TMU Revision: 57005
Number TMUs: 57005
WARNING: Board 0: Bogus number of TMUs (57005)!

Was caused by

sst1Init Routines: InitCode $Revision: 8 $
sst1InitMapBoard(): BoardsInSystem = 1
sst1InitMapBoard(): vAddr:0xc8dc0000 pAddr:0xdd000008 Dev:0x9 Board:0
sst1InitRegisters(): Setting TREX-to-FBI FIFO THRESHOLD to 0x8...
sst1InitRegisters(): Setting PRELIM FT-CLK delay to 0x8...
sst1InitRegisters(): Could not detect DAC...

This reference design Voodoo 1 card did have a very bad hit to the empty PCB area and I'd already had to rebuild the VGA pass-through circuitry, but it looks like the DAC took a hit too:
3dfx voodoo1 TVP3409 DAC traces damaged.JPG
Those are all damaged traces where you see the thin metal legs, three of them were the red green and blue signals. I found that after I got the DAC to communicate with the FBI chip by resoldering legs at the upper part, which are the digital connections, the MOJO result changed into a good one:

Mojo and SST log of working Voodoo 1

SST:
sst1Init Routines: InitCode $Revision: 8 $
sst1InitMapBoard(): BoardsInSystem = 1
sst1InitMapBoard(): vAddr:0xc8dc0000 pAddr:0xdd000008 Dev:0x9 Board:0
sst1InitRegisters(): Setting TREX-to-FBI FIFO THRESHOLD to 0x8...
sst1InitRegisters(): Setting PRELIM FT-CLK delay to 0x8...
sst1InitRegisters(): Storing TREX0INIT0=0x5441
sst1InitRegisters(): Storing TREX0INIT1=0xf420
sst1InitRegisters(): Storing TREX1INIT0=0x5441
sst1InitRegisters(): Storing TREX1INIT1=0xf420
sst1InitRegisters(): Storing TREX2INIT0=0x5441
sst1InitRegisters(): Storing TREX2INIT1=0xf420
sst1InitRegisters(): Setting up FAST DRAM Configuration
sst1DeviceInfo: Board ID: Obsidian GE
sst1DeviceInfo: FbiConfig:0x2, TmuConfig:0x11
sst1DeviceInfo: FBI Revision:2, TMU Revison:1, Num TMUs:1
sst1DeviceInfo: FBI Memory:2, TMU[0] Memory:2
sst1DeviceInfo: Dac Type: TI TVP3409
sst1DeviceInfo: SliDetect:0
sst1InitRegisters(): LFB Writes go through memory FIFO...
sst1InitRegisters(): TEXTURE Writes go through memory FIFO...
sst1InitRegisters(): exiting with status 1...
sst1InitShutdown(): Shutting down SST-1 #0...
sst1InitShutdown(): Returning with status 1...
Mojo result:
Info for Voodoo board # 0:
=====================================================
Virtual Base Address: 0xe6000000
Physical Base Address: 0xe6000008
PCI Device Number: 0x9
Vendor ID: 0x121a
Device ID: 0x1
FBI Revision: 2
FBI Memory: 2 MB
FBI PowerOn Sense: 0x2
TMU PowerOn Sense: 0x11
FBI DAC Output Color Format: 24BPP
Scan-Line Interleaved? No
TMU Revision: 1
Number TMUs: 1
TMU 0 RAM: 2 MB

It even has full RAM showing for both the texture and frame buffer - so I started a game and only saw a totally black screen, but the computer was still running. I was probing at the DAC while it was running to see if the VCCA (DAC analog side power) and ground were connected when my probe on ground slipped and maybe completed the circuit since the ground trace there was completely broken away. But when my multimeter probe slipped something good happened instead of something bad - I could see the unreal flyby in blue!
Pressing the DAC area more I think I managed to get red as well and turned it off to work on the DAC. I thought the DAC might be damaged initially but if it's working well enough to generate 2 colours of 3 and make an image it should be fine. Instead of using hot air I used the low melt solder stuff and that got the DAC off pretty cleanly except as you can see in the picture a bunch of traces were broken probably before the chip was removed.

Took an hour or two to get all the traces in place and then I marked the DAC legs where there were wire legs instead of copper traces, so I could squash some of the "J" shaped legs into "L" legs - allowing the remaining good traces to still make contact with the DAC's PLCC legs, while giving clearance for the wire leg 'traces'.
That went great though, I put the card back together again and could actually test it - it went straight into tomb raider demo with the 3DFX splash screen showing all colours 😁

This card has been with me for something like 18 months and I've now and then taken another go at getting it working with no result each time because I didn't know what the diagnostics could show, didn't know that no DAC would upset the whole card. I've taken chips off, replaced RAM, broken traces in the process of replacing RAM, resoldered the QFP legs probably multiple times... It's had a rough time and I've never been able to test it.
I found that although it can now run games with proper display, it's glitching sometimes - I resoldered the RAM again and added some more capacitors to no avail, but there were some 0-ohm resistors on the back that had somehow had the markings rubbed off and replacing those seemed to stabilise the display. To test it I ran the unreal flyby for a while and it looked fine.
I was celebrating a job well done when...
IMG_2328 (Large).JPG
Party mode activated... I may still have a bit more work to do...

That is some seriously nice detective work figuring out what was wrong. I really love seeing old hardware fixed and working again. Good luck figuring out party mode.

Wash/recap dead Epox EP-7KXA w/Athlon 650 Pluto Slot A (~$15 flea market)
ovreclock (115x6.5 - max available) mem @153 sdram
memtest 12hrs test 8 (random numbers) - ok
another milestone (SlotA combo) - done
upd: 17 of 18 2200mkF capacitors was ok by multimeter (even leaky ones), but not for ESR tester: Vloss 20-35%, ESR - 1.5-2.7Ohm and around half of capacitance.
New one Sanyo are almost identical w/~3% tolerance on capacitance, 0.07Ohm ESR and 2.5% Vloss.

Ati Mach64 vram testing, it...takes time!

PcBytes wrote on 2023-11-10, 19:51:

EDIT: And just as I finished this, a G735 popped up over the classifieds. AMD K6-2 @ 400MHz.

Green735 is actually an AlphaTop laptop, ECS bought them out in 2001. If it has a K6-2, it's a G735+ model. They seem to be pretty neat machines.

I've been playing around with a Sapphire AGP Radeon HD 2600 XT, but it runs really hot. At first, it idled at 70° C with the fan at 76%. I cleaned the GPU and repasted it, and now it idles at 60° C with the fan at 74%. It has good airflow, so I don't understand the temps. Is this typical for this GPU?

It never seemed to me that they designed for anything less than 100% fan on load before about 2014-2015 and idle power consumption was pretty high 10s of watts rather than 10W so do not apply last decade's standards to the preceding decades hardware.

I don't mind 100% fan speed under load, but 70-something percent at idle is driving me nuts! It's a tiny, load fan.

fosterwj03 wrote on 2023-11-11, 17:37:

I've been playing around with a Sapphire AGP Radeon HD 2600 XT, but it runs really hot. At first, it idled at 70° C with the fan at 76%. I cleaned the GPU and repasted it, and now it idles at 60° C with the fan at 74%. It has good airflow, so I don't understand the temps. Is this typical for this GPU?

I have been testing a Quadro K600, K620, and K1200 this week. They all push mid 80C under heavy FurMark load, even after getting some fresh thermal paste.

What is your temperature under load?

I installed and tested this pretty period correct combo of Thunderbird 1400, Abit KG7 RAID, Radeon 8500 and Audigy1 on my bench and set up Win98SE with drivers on a 120GB SSD. Everything works just fine. System is currently running stock with zero OC and I ended up installing two 256MB HyperX KHX2700/256 sticks with Samsung TCB3 chips and those things fly 2-2-2-4 on this.

I probably end up swapping these inside my Abit BH-6, P3 800MHz box just for the fun. I probably take the Voodoo2 from that set up to have even better backwards compatibility with late 90s glide games.

vsharun wrote on 2023-11-11, 09:56:

Wash/recap dead Epox EP-7KXA w/Athlon 650 Pluto Slot A (~$15 flea market) ovreclock (115x6.5 - max available) mem @153 sdram m […]
Show full quote

Wash/recap dead Epox EP-7KXA w/Athlon 650 Pluto Slot A (~$15 flea market)
ovreclock (115x6.5 - max available) mem @153 sdram
memtest 12hrs test 8 (random numbers) - ok
another milestone (SlotA combo) - done
upd: 17 of 18 2200mkF capacitors was ok by multimeter (even leaky ones), but not for ESR tester: Vloss 20-35%, ESR - 1.5-2.7Ohm and around half of capacitance.
New one Sanyo are almost identical w/~3% tolerance on capacitance, 0.07Ohm ESR and 2.5% Vloss.

Very cool! Slot A is def on my list of platforms to check out eventually.

I need to get into recapping. Does anyone know if there's a newbs guide to figuring out what caps I need and the best way to get them? And what's the prevailing wisdom in terms of replacing all caps vs just visibly bad ones on say... a ca. 2004 ASUS Althlon XP mobo?

CharlieFoxtrot wrote on 2023-11-11, 21:02:

I installed and tested this pretty period correct combo of Thunderbird 1400, Abit KG7 RAID, Radeon 8500 and Audigy1 on my bench and set up Win98SE with drivers on a 120GB SSD. Everything works just fine. System is currently running stock with zero OC and I ended up installing two 256MB HyperX KHX2700/256 sticks with Samsung TCB3 chips and those things fly 2-2-2-4 on this.

I probably end up swapping these inside my Abit BH-6, P3 800MHz box just for the fun. I probably take the Voodoo2 from that set up to have even better backwards compatibility with late 90s glide games.

IMG_1984.JPEG

Nice!

I think this period is when I decided onboard sound was good enough and stopped using sound cards. (As soon as AC97 mobos were out I guess) Pretty sure the last sound card I bought was a Hercules Game Theater XP. Never bothered with Audigys and whatever - is there really any substantial benefit or is it more in the realm of subjective audiophile qualities?

chrismeyer6 wrote on 2023-11-11, 07:46:

Thermalwrong wrote on 2023-11-11, 02:22:

Wah, I've been buying broken 3DFX cards for a while to see if I can fix them. Today I was looking over a Voodoo 1 that I'd bough […]
Show full quote

Wah, I've been buying broken 3DFX cards for a while to see if I can fix them. Today I was looking over a Voodoo 1 that I'd bought last year and I could never get it to do anything. It would only report 0xDEAD for the TMU status with 57005 TMUs. I'd thought it was down to a broken pin since there were a few on this card - but now I know how to actually use mojo.exe to generate the SST.LOG file which gives actual information:

This 0xDEAD result in Mojo

Info for Voodoo board # 0:
=====================================================
Virtual Base Address: 0xe6000000
Physical Base Address: 0xe6000008
PCI Device Number: 0x9
Vendor ID: 0x121a
Device ID: 0x1
FBI Revision: 2
FBI Memory: 0 MB
FBI PowerOn Sense: 0x0
TMU PowerOn Sense: 0xdead
FBI DAC Output Color Format: 24BPP
Scan-Line Interleaved? No
TMU Revision: 57005
Number TMUs: 57005
WARNING: Board 0: Bogus number of TMUs (57005)!

Was caused by

sst1Init Routines: InitCode $Revision: 8 $
sst1InitMapBoard(): BoardsInSystem = 1
sst1InitMapBoard(): vAddr:0xc8dc0000 pAddr:0xdd000008 Dev:0x9 Board:0
sst1InitRegisters(): Setting TREX-to-FBI FIFO THRESHOLD to 0x8...
sst1InitRegisters(): Setting PRELIM FT-CLK delay to 0x8...
sst1InitRegisters(): Could not detect DAC...

This reference design Voodoo 1 card did have a very bad hit to the empty PCB area and I'd already had to rebuild the VGA pass-through circuitry, but it looks like the DAC took a hit too:
3dfx voodoo1 TVP3409 DAC traces damaged.JPG
Those are all damaged traces where you see the thin metal legs, three of them were the red green and blue signals. I found that after I got the DAC to communicate with the FBI chip by resoldering legs at the upper part, which are the digital connections, the MOJO result changed into a good one:

Mojo and SST log of working Voodoo 1

SST:
sst1Init Routines: InitCode $Revision: 8 $
sst1InitMapBoard(): BoardsInSystem = 1
sst1InitMapBoard(): vAddr:0xc8dc0000 pAddr:0xdd000008 Dev:0x9 Board:0
sst1InitRegisters(): Setting TREX-to-FBI FIFO THRESHOLD to 0x8...
sst1InitRegisters(): Setting PRELIM FT-CLK delay to 0x8...
sst1InitRegisters(): Storing TREX0INIT0=0x5441
sst1InitRegisters(): Storing TREX0INIT1=0xf420
sst1InitRegisters(): Storing TREX1INIT0=0x5441
sst1InitRegisters(): Storing TREX1INIT1=0xf420
sst1InitRegisters(): Storing TREX2INIT0=0x5441
sst1InitRegisters(): Storing TREX2INIT1=0xf420
sst1InitRegisters(): Setting up FAST DRAM Configuration
sst1DeviceInfo: Board ID: Obsidian GE
sst1DeviceInfo: FbiConfig:0x2, TmuConfig:0x11
sst1DeviceInfo: FBI Revision:2, TMU Revison:1, Num TMUs:1
sst1DeviceInfo: FBI Memory:2, TMU[0] Memory:2
sst1DeviceInfo: Dac Type: TI TVP3409
sst1DeviceInfo: SliDetect:0
sst1InitRegisters(): LFB Writes go through memory FIFO...
sst1InitRegisters(): TEXTURE Writes go through memory FIFO...
sst1InitRegisters(): exiting with status 1...
sst1InitShutdown(): Shutting down SST-1 #0...
sst1InitShutdown(): Returning with status 1...
Mojo result:
Info for Voodoo board # 0:
=====================================================
Virtual Base Address: 0xe6000000
Physical Base Address: 0xe6000008
PCI Device Number: 0x9
Vendor ID: 0x121a
Device ID: 0x1
FBI Revision: 2
FBI Memory: 2 MB
FBI PowerOn Sense: 0x2
TMU PowerOn Sense: 0x11
FBI DAC Output Color Format: 24BPP
Scan-Line Interleaved? No
TMU Revision: 1
Number TMUs: 1
TMU 0 RAM: 2 MB

It even has full RAM showing for both the texture and frame buffer - so I started a game and only saw a totally black screen, but the computer was still running. I was probing at the DAC while it was running to see if the VCCA (DAC analog side power) and ground were connected when my probe on ground slipped and maybe completed the circuit since the ground trace there was completely broken away. But when my multimeter probe slipped something good happened instead of something bad - I could see the unreal flyby in blue!
Pressing the DAC area more I think I managed to get red as well and turned it off to work on the DAC. I thought the DAC might be damaged initially but if it's working well enough to generate 2 colours of 3 and make an image it should be fine. Instead of using hot air I used the low melt solder stuff and that got the DAC off pretty cleanly except as you can see in the picture a bunch of traces were broken probably before the chip was removed.

Took an hour or two to get all the traces in place and then I marked the DAC legs where there were wire legs instead of copper traces, so I could squash some of the "J" shaped legs into "L" legs - allowing the remaining good traces to still make contact with the DAC's PLCC legs, while giving clearance for the wire leg 'traces'.
That went great though, I put the card back together again and could actually test it - it went straight into tomb raider demo with the 3DFX splash screen showing all colours 😁

This card has been with me for something like 18 months and I've now and then taken another go at getting it working with no result each time because I didn't know what the diagnostics could show, didn't know that no DAC would upset the whole card. I've taken chips off, replaced RAM, broken traces in the process of replacing RAM, resoldered the QFP legs probably multiple times... It's had a rough time and I've never been able to test it.
I found that although it can now run games with proper display, it's glitching sometimes - I resoldered the RAM again and added some more capacitors to no avail, but there were some 0-ohm resistors on the back that had somehow had the markings rubbed off and replacing those seemed to stabilise the display. To test it I ran the unreal flyby for a while and it looked fine.
I was celebrating a job well done when...
IMG_2328 (Large).JPG
Party mode activated... I may still have a bit more work to do...

That is some seriously nice detective work figuring out what was wrong. I really love seeing old hardware fixed and working again. Good luck figuring out party mode.

It seems like I'm going to need it. Fixed a couple of loose legs by resoldering all the pins on both chips, they're already covered in flux anyway so it doesn't take long. Broke a pin off that was loose at the edge of the TMU and used my one of a kind ceramic pick tool to carve into the chip - this thing's made up of some idealtek ceramic tweezers that I broke through mis-use, the ceramic is a special very hard but somewhat brittle material, it's great in this form for scraping soldermask off or gently carving into PQFP chips 😀
I have now come across enough broken pins on PQFP chips that they're becoming trivial to fix, here's the other reference voodoo1 card I just got and got working that spurred me onto trying to fix the one I'd bought last year:

I've got the broken one into a slightly less broken state, it's now got black <> diamond kinda shapes in tomb raider and no more crazy colours. Kinda like it's not able to write to some of the texture memory.

Does anyone know if there are schematics or pinouts known for the Voodoo 1 or Voodoo 2's chips? I tried beeping them out a couple of years back but gave up 70% of the way through the TMU - today I thought I'd look for schematics again and realised that arcades used the 3dfx chips... arcades cost enough to be worth repairing so they have schematics....
Check out page 42 onwards of this PDF if you ever want to know what the pinout is for the 500-0004-02 or 5000-0003-03 chips 😁 https://arcarc.xmission.com/PDF_Arcade_Manual … _Schematics.pdf
If anyone knows of any arcade systems that integrated the Voodoo2's chips, that could yield interesting results.

Now I don't need to beep anything out and now need to start probing to figure out what's still not working on this bashed voodoo1.

ubiq wrote on 2023-11-11, 21:17:

Very cool! Slot A is def on my list of platforms to check out eventually.

I need to get into recapping. Does anyone know if there's a newbs guide to figuring out what caps I need and the best way to get them? And what's the prevailing wisdom in terms of replacing all caps vs just visibly bad ones on say... a ca. 2004 ASUS Althlon XP mobo?

For early 2000s boards the main concern is all the ultra low ESR caps used on the VRM. Many of these boards had failed caps after a year or two of normal use. What I’ve done is I’ve searched the original cap datasheets and then bought caps which are at least equivalent in terms of ESR and ripple capability. You may need to use polymer caps as replacements for some of the caps as no one has manufactured those ultra low ESR electrolytics for over a decade or so.

If you recap the board, always change at least all the caps of the same series. Caps don’t always fail so that they look visibly bad and if one has failed, rest of them are most likely on the way out.

If you don’t have much soldering experience, I suggest you practice with some broken board of the same era. Socket A era boards are multilayer PCBs and those are much trickier to solder compared to older two layer ones. And they are almost impossible to repair if you screw something up.

ubiq wrote on 2023-11-11, 21:23:

Nice!

I think this period is when I decided onboard sound was good enough and stopped using sound cards. (As soon as AC97 mobos were out I guess) Pretty sure the last sound card I bought was a Hercules Game Theater XP. Never bothered with Audigys and whatever - is there really any substantial benefit or is it more in the realm of subjective audiophile qualities?

This like many similar 2001 boards doesn’t have sound capabilities. They did start to become more common during the Socket A era, but most of them were rubbish. IMO the first good one was and is nForce2 MCP-T, ie Soundstorm, which had good performance and was relatively noiseless.

Kahenraz wrote on 2023-11-11, 20:16:

fosterwj03 wrote on 2023-11-11, 17:37:

I've been playing around with a Sapphire AGP Radeon HD 2600 XT, but it runs really hot. At first, it idled at 70° C with the fan at 76%. I cleaned the GPU and repasted it, and now it idles at 60° C with the fan at 74%. It has good airflow, so I don't understand the temps. Is this typical for this GPU?

I have been testing a Quadro K600, K620, and K1200 this week. They all push mid 80C under heavy FurMark load, even after getting some fresh thermal paste.

What is your temperature under load?

91° C is the worst I've seen with the card after repaste. The card performs properly under load. About 65 FPS in Crysis at 1024x768 (mostly low settings, medium textures).

CharlieFoxtrot wrote on 2023-11-11, 21:56:

ubiq wrote on 2023-11-11, 21:23:

Nice!

I think this period is when I decided onboard sound was good enough and stopped using sound cards. (As soon as AC97 mobos were out I guess) Pretty sure the last sound card I bought was a Hercules Game Theater XP. Never bothered with Audigys and whatever - is there really any substantial benefit or is it more in the realm of subjective audiophile qualities?

This like many similar 2001 boards doesn’t have sound capabilities. They did start to become more common during the Socket A era, but most of them were rubbish. IMO the first good one was and is nForce2 MCP-T, ie Soundstorm, which had good performance and was relatively noiseless.

Ahh, got it. I remember being really excited about nForce2 - got one of those Shuttle mini systems based on it, paired with a Radeon AIW 9700 Pro IIRC. What an amazing little kick-ass system for the time.

Lol @ the blinding blue LED era. Game Theater XP breakout on top too - must have been hooked up to another system...

My blue power LED is so bright it lights up my whole bedroom. I've given some serious consideration to putting an ugly piece of electrical tape over it. I think they needed to use a different diffuser.

Or clear tape. Or even maybe a dab of diluted acrylic paint.

Several things in my bedroom have green masking tape on them for that reason, can still see enough light through the tape to tell you it's operating but cuts down the retina sear or daylight at 2AM problem.

I use gaffer tape, not electrical tape. Electrical tape turns gross after a while, once the adhesive turns to slime.