Diamond Monster 3D II - A Voodoo 2 rises from the ashes? (Part 2/3)

[@bitsundbolts]

While working on part 3, I noticed that the discoloration of Tomb Raider shown in this video was caused by the wrong 3Dfx patch! Yes, there are two different patches for Tomb Raider. One is only for Voodoo 1 cards (that is the patch I had installed while making this video). The second patch is for Voodoo Rush and Voodoo 2 cards. Of course, I downloaded the patch and added it to Tomb Raider. Unfortunately, with the correct patch, the game starts, but the screen remains black after the intros. The menu screen is never rendered. So, we still have an issue with the card! The problem? Only 1MB of memory is detected by the FBI chip! I hope you can help me figure out why. Is the FBI chip damaged? Could the DAC or the small chip next to it be the issue (both connect to 5V that damaged the memory chips. Any help and hints are appreciated! Thanks everyone!

[@tcscomment]

maybe it's an addressing issue? like, what if the FBI chip has some troubles with connecting to the memory chips? (eg. too much resistance on the trace?)

[@tcscomment]

to me this looks like the FBI chip has not enough memory to work properly more than a RAMDAC issue

[@bitsundbolts]

Good point. I can check the resistance and see if there's something odd. I checked the chip that replaced the one with the hole in it. All connections seem to be fine, but it might just be the one that's part of the 1MB. I'll check all the other chips on that side.

[@petertorda5487]

I would check these vias under exploded chips (on first video they looks quite burned), maybe there is no good connection to rest of memory chips (especially on other side of board). This cloud explain discoloration and only 1MB detected.

[@vladislavknyazkov1962]

@@bitsundbolts try measurements of address/data/all other signals in diode mode (multimeter pos terminal to gnd and negative to signal lines.) All lines should be in more or less same range, if one is too off - may be issue with 3dfx chip. This way you can check if blown up RAM chip damaged 3dfx chip IO.

[@necro_ware]

I had the same color issue recently, had also a V2 to repair. That color problem is normal and is due to V1 support in TR in DOS. It costed me couple of hours of ghosts hunting till it turned out, that in DOS you have to patch TR for V2 to get the colors right. In windows the card would work right and there is nothing wrong with the DAC or something else. The missing FB memory is a problem though. I would suggest to go through the pins of the memory chips and measure the resistance on every one of those. Compare then that all data and address pins have the same resistance to the ground. May be one of the traces in the PCB has melted.

[@bitsundbolts]

Hey there! Thanks for the tip! And yes, hehe, I figured out that there was another 3Dfx patch for tomb raider for voodoo rush and voodoo 2 cards. Unfortunately, with that patch applied, the Voodoo 2 can't render any 3D scenes anymore. Just a black screen where the main menu screen should appear. I'll go back and check the connection of every memory chip connected to the FBI chip. Thanks for your help!

[@voodler]

@@bitsundbolts Try out Unreal as well, since the card is working in Windows :) The flyby certainly looks interesting when the 3d card is broken. My 3Dfx Voodoo 1 that had the RAMDAC fault, once that was fixed the TMU memory CAS lines broke and the castle went kind of 'disco' mode

[@ShamblerDK]

Ohhh, Necro_ware! Fancy seeing you here :-)

[@DefenderOfBoston-yo2tl]

Hi m8! Great to see you over here on Alex' channel, thanks for chiming in, and with such helpful pointers too! Let's hope it indeed is an issue with a trace/minor SMD component along the way rather than with the comparatively irreplaceable (in the sense that they aren't made anymore) FBI chip itself. This is the card I emailed you about a while back btw. 😊

[@quittessa1409]

I was also going to suggest checking the address and data lines from each chip to the fbu, when the old chips exploded it might have taken out one or more of the lines, likely an address line.

[@adriansdigitalbasement]

It freakin' works!! (Kind of) I don't think the DAC is the reason for your lack of frame buffer RAM, but I have seen a bad DAC cause the visual issues you are having with the colors.. (Not on a 3DFX but VGA cards) Might be worth swapping that PLCC chip if you have a spare.

[@bitsundbolts]

Hello Adrian, thanks for your input! If I understand correctly, the DAC is connected to the FBI as well as the memory chips. It could be as you say and the DAC might cause the loss of memory. However, I might get access to something that most probably will eliminate all the guesswork!

[@adriansdigitalbasement]

@@bitsundbolts I don't think so -- from what I understand the DAC simply contains look up tables to convert the digital framebuffer output to analog video and in the process apply the correct color palette. Of course the lack of memory showing up really could just point to the framebuffer IC having been fried too. The damage was so expensive on that board perhaps some of the RAM chips backed too much power into the framebuffer taking it out? It's so hard to know when everything is so highly integrated and undocumented like on those "modern" ICs

[@bitsundbolts]

I agree, it is very unfortunate that there is limited information for those chips available. It turns into guesswork. That might be fine for people who are just tinkering around, but it makes it so much harder for people in the repair business. I stopped this Voodoo 2 project for now because I am waiting for information from a viewer. It might change how we debug faulty Voodoo cards in the future.

[@Madrrrrrrrrrrr]

I think he didn't soldered the 3 pads he replaced. Or that part wasn't in the video or i missed something. And if that was with framebuffer memory because if they are seen in pairs than he would still get 3 mb instead of one. Don't think it's a multiple layered board so the damages couldn't be it and should be searched next in line to the frame buffer chip.

[@tony359]

It's cool that the regulator protected the main chips - and I guess the regulator won't care if 5V or 12V are coming in! I am not familiar with that card - the only thing that comes to mind is to rule out the simple ones first: swap the FBI RAM. Then for sure see if you can also swap all the ICs which are directly connected to 5V - but not the regulator. Also I would investigate more on that hole at the back! One thing you are probably already aware of - but just in case - bear in mind that a multimeter in continuity mode will "beep" from - say - 0 to 30 Ohm. So if you hear a beep it doesn't mean you have a dead short - it could be something like 20 Ohms which might be normal. I'm sure you can revive this poor card! Well done!

[@bitsundbolts]

Hey Tony. I will first try and probe around with the multimeter to see if there are "dead" data lines. Maybe I'll be able to spot which chips are used and which aren't. I use the multimeter to test the connectivity from memory to the FBI chip. Most of those lines go through a 22 ohm resistor - you need to wait a bit before the multimeter beeps, but it does. So, yes, I'm aware of the behavior. Thanks for watching.

[@CMDRSweeper]

I am not into electronics repair other than watching, but I do a bit of troubleshooting when it comes to hardware and computers in general, and this problem to me smells of a dead memory channel as my first instinct. With the tools and stuff you have, I would socket the memory chips on the Voodoo as you mentioned in your first video, but ESPECIALLY around the framebuffer chip. Earlier you mentioned that mojo was a 3DFX tool for diagnosing the cards, so I would trust it's detection from now. My next move after that is to move a memory chip around and see where it reports having memory and where it reports nothing. From that point on, you can either test if the signal lines are intact to the FBI chip or there is hidden board damage, if there isn't, you have an internal fault inside the FBI chip itself. Or if there is a break, well you got a hail mary and you can hunt down the flaw in the PCB and fix it.

[@komplikatorful]

Probaby the PSU had a faulty capacitor on a voltage line that is also a feedback line - old PSU did that, they used one of the lines as a feedback and if a faulty cap pulls this line down, the PSU compensates by rising *all* the voltages. So the voltage on 5V probably was slowly rising until something gave up. That's why you see the burn marks on the PCB - the chips were probably "fine" for a while, just dissipating that overvoltage as a heat...

[@pavelfara9333]

I woud expect the owner of the card had been using some early ATX PSUs. But who knows. Anyway thanks for your insight and explanation.

[@komplikatorful]

​@@pavelfara9333 It would had to be "new enough" to have 12V feedback, the oldest PSUs used to use 5V for feedback - I phrased it badly, sorry 🙂 These PSUs are old from todays POV, but not the oldest. Most of the current draw nowdays is from 12V line and PSU manufacturers switched feedback from 5V to 12V to accomodate this trend. It even did not have to be really "faulty" since these PSUs usually do have a minimal load requirement. It means that you *have* to draw at least some current in order for the feedback (and its output voltages) to be stable. And if someone did not know this and started blindly pulling cables, it could very well cause big voltage spikes and in turn destroy the weakest link (memory chips in this case - maybe more, we will have to wait for part 3 😁). New-ish PSUs usually don't have this limitation.

[@DefenderOfBoston-yo2tl]

@@pavelfara9333 Hi there! (Previous/Intermediate) owner here. It certainly was a crappy early ATX PSU from a cheap pre-built meant for office use. The PC used to be my friend's and I upgraded it for him with this card. It failed some time back then and we (I) unsuccessfully tried to revive it with another PSU at the time. So he discarded the blown up PSU, put the PC aside and we completely forgot about the whole thing, until he dug it out again a few years ago and gave it to me as a b'day present. Knowing about my nostalgia and not remembering that it was actually broken, haha. Still a great present! 😄

[@DefenderOfBoston-yo2tl]

This is the most plausible explanation I have seen so far! Sadly, we discarded the bad PSU when it failed all those years ago and can't perform a post-mortem of it anymore (which would certainly have been interesting too). Only the aftermath, in the sense of the rest of the system, with at least the mobo being a casualty as well. IIRC, the 2D card has some discoloration on it too though!

[@AnnaVannieuwenhuyse]

This mechanism of regulation is called Group Regulation. All rails share the same transformer, and they're all windings on the same secondary side. Thus, they all respond equally to any increase in primary side drive power. These are the same PSUs that need some level of load on all rails to be stable.

[@GadgetUK164]

Great progress! The colour issue I suspect is DAC related! And clearly there's still some issue with the FBI chip - if not a trace or resistor / cap related, it must have died too =/ I would be tempted to check what else on that board is connected to 5v (assuming the DAC), but maybe something else is too (eg. a pull up resistor maybe). Check those 3 ferrites, transistors and resistors on the left of the board near the VGA connector (not sure if those are the 3D out or the VGA in), maybe a ferrite has gone open on one colour channel, or maybe the transistor related to one colour is not working correctly. You could swap them around after comparing them with each other on resistance / diode test. That might resolve the colour issue, if the DAC isnt the issue.

[@bitsundbolts]

I wonder if a faulty DAC could cause the memory to be limited to 1 MB on the FBI chip. That would solve the mystery without me messing around with the memory chips again. I could maybe remove the DAC and try mojo again to see if it makes a difference.

[@GadgetUK164]

@@bitsundbolts I am more inclined to think the 1MB issue is an FBI chip problem (either a trace / passive / fault in the chip). I've never seen a DAC have any relation to how much frame buffer memory is detected. Double check that every VCC pin on every RAM chip measures 5v, it may have a burned out internal trace somewhere meaning 5v isn't getting somewhere.

[@bitsundbolts]

Good point! I'll try to measure the power pins. I also replaced the SMD components around the exploded chip. Unfortunately, unchanged - 1MB on the frame buffer. I could try to replace the frame buffer chip.

[@voodler]

@@bitsundbolts Don't remove the DAC, that causes even worse errors with mojo, my personal experience being that it fails before even giving the mojo readout. The DAC I think is working since it can display white, the RAMDAC being down R / G / B would mean white would look wrong.

[@kingofl337]

It is highly suspicious that the components with the most damage are having an issue. It could be when those memory chips went out it damaged a few IO pins on the FBI. This is also a multilayer board, a via could be damaged. An oscilloscope may point out the issue.

[@BleughBleugh]

You terror! Leaving us dangling like that :-p Can’t wait for part 3

[@bitsundbolts]

Haha, I am sorry! But really, I haven't continued with the card yet. I need suggestions and hints from all of you for part 3 🙂

[@olivedrabcorp3026]

I don't know how far you could go in the Voodoo 2 phoenix special edition resurrection process, but in any case it's damn interesting to follow this technical adventure 😃

[@JensChristianRestemeier]

BTW, 57005 is 0xDEAD.

[@bitsundbolts]

Correct! 😁 Engineering at its best. I discovered this little fact while researching.

[@DefenderOfBoston-yo2tl]

that's just hilarious! 😂 i already had a hunch. one of our university profs loved to play around with some words you can form with a-f in his examples too, affe being his favourite (austrian here).

[@AnnaVannieuwenhuyse]

​Programmers have historically loved playing around with hexadecimal values that spell words. The classic 0xDEADBEEF is one of the well known ones!

[@TheDemocrab]

@@AnnaVannieuwenhuyse Funnily enough, DeaDBeeF is a pretty dang good audio player for Linux.

[@johnmay4803]

fantastic vid thank you for the upload. ive been watching your vids for the past week and i really enjoy them

[@bitsundbolts]

Thank you!

[@harukaapl]

Awesome job! You can try check memory connections to FBI , maybe resistors/capacitors are dead near memory and need to replaced. Also you can try to check using osciloscope if there are signals on memory pins, especially control pins. Worst case scenario will be that Voodoo chip is damaged and no matter what you do it will not detect more memory. When I got my Voodoo 2 I was also been supriced by blue Tomb Raider artifacts, then I've read info about patch for VooDoo 2

[@bitsundbolts]

Yes, that Voodoo 2 patch got me 😄. I tested all the connections on the front with the multimeter already. However, I'll try the oscilloscope! Thanks

[@johnvanwinkle4351]

Great job repairing it! You have a lot of patience and skill for that level of work.....grin

[@flyguille]

If the frame buffer only sees 1mb but is writting like it has 2mb, would have sense that the GREEN part of RGB INFORMATION got missing, if only 1mb it is. Page addressing Error, so, some of the higher adress lines are faulty, so for addressing for 1nb up to 2mb it is used the bit 20 must be set to 1, but, as dynamic ram multiplex addressing in rows and columns, it will the address line A10 at fault.

[@flyguille]

Now, this comment doesn' take in account how the memory chips are banked, if you has the card schematics , that is easy to figure out.

[@flyguille]

BY example , If the memory chips are to say up to 1MB addressing, so it is normal to use the /CE (chip enable) line to activate lower bank / upper bank, so, upper bank will mean ACCESS TO 1mb up to 2mb addresses, and tricks like that are common in computing.

[@bitsundbolts]

Ah! Great details! Let me check those data lines and probe a bit with the oscilloscope! Thanks for the info!

[@ApostolCV]

As electricians say, there are two types of failure: 1. There is a short circuit where there should not be one. 2. There is no contact where it should be. Looks like the second variant is your. First , try to figure out which pair of memory chips do not see FBU. Second , check resistance at this traces to FBU and in work chanel. 3. There is a chance that the "one memory chanel in FBU has gone". 4. Maybe some issues in layouts.

[@hopper7755]

Most likely something went wrong in the PSU feedback loop, AT/ATX PSUs can output up to 28Volts on 12Volt line and 12-16V on 5V line (the worst case, no feedback at all) and a lot of current. You can make really nice adjustable PSUs from them by hacking the feedback loop.

[@michvod]

With really bad caps that can happen yeah as there would be no smoothing. If there are still good caps, it will greatly exceed their ratings and they will start clamping the voltage until they heat and vent

[@TzOk]

The voltage regulator is probably fine with input voltages up to at least 15V, so I wouldn't worry about it. A faulty DAC could result in missing some color information, and look like a faulty memory in the framebuffer, but it should not affect the memory size FBI sees. However, the FBI is directly wired to both FB memory and a DAC, so a fault in either can affect or even damage the FBI itself. Finally, a charcoaled PCB is conductive...

[@bitsundbolts]

I learned this conductive PCB just recently! Very good point. I hope it was just the surface and nothing is shorted (like a data line). Maybe I have to do some PCB digging below that one memory chip that burnt a hole in the PCB.

[@mal2ksc]

@@bitsundbolts Is this a typical 4 layer board where data lines are on the visible layers, and the two planes in between are Vcc and ground? The hole in the board might be an exploded via that used to connect to either Vcc or ground, and now some chip isn't getting power. (Or even half a chip since you showed the Vcc pins aren't necessarily shared in those RAM chips). I'm hoping this is true because this would be an easy problem to bodge wire around.

[@bitsundbolts]

@mal2ksc I'll check all power and ground pins again. I think there are even more of them somewhere in the center. Thanks for the hint.

[@Imperious685]

All I can add to others is check the resistor arrays, they are know to go out of spec, especially after what happened to that card. Also I mentioned previously You can test the memory chips with a program called VMTCE in a 2d card with EDO expansion sockets. Good Luck! Looking forward to Part 3.

[@Ale.K7]

Can't wait for part 3!

[@bluethunder8383]

Wow, so close, awesome to see you get it this far, crossing my fingers for the final touch to make part 3 a reality

[@bitsundbolts]

Fingers crossed!

[@LeeMc007]

Absolutely amazing that it works at all, well done getting this far, I think my first port of call would be to check all the resistors and caps in what was the burnt area bottom right of the card and also check the VIA's that are close by, that was some serious burning around there, I was amazed no traces were melted and like everyone else said, the connections from the mem chips to the FBI, if you have a scope, check the data lines to see which chip is the one working and go from there. I don't have a huge amount of experience with Voodoo's but I have 2 to repair myself so very interested in the outcome, best of luck.

[@bitsundbolts]

Thanks! Yes, I'll focus on the memory chips and the connection to the FBI chip. Something is up, and we will figure it out!

[@LeeMc007]

@@bitsundbolts Well, you inspired me to have a look at my faulty Diamond Monster 3d2, it would black screen crash playing a game/benchmark and occasionally I'd get some random lines after it crashed, messed around with Mojo and didn't find any faults at all, gave the exact same results as both my other V2's, I have an Orchid Righteous 3d that's also faulty (displays fine but freezes) as a last resort I swapped the DAC(which is the same ICS 5342) and it works perfectly now, need a DAC for the Orchid card now but can't seem to find them anywhere? Just wanted to drop by and says thanks for the inspiration, good luck with yours. 👍😁

[@mal2ksc]

As I mentioned in another comment, the mysterious burned hole might be a blown via between the Vcc plane and a chip, so that chip is no longer getting power.

[@bitsundbolts]

@LeeMc007 I did get a few DAC chips some time ago - AliExpress again. Not sure if they work or if they're original, but one of them might come in handy for this card. We will see. I hope you'll be able to fix the Diamond Voodoo 2. I might get access to something for part 3. I have a feeling it will revolutionize debugging voodoo and voodoo 2 cards... Interesting times ahead!

[@JathraDH]

I had 2 of these in SLI. Good memories!

[@petr56321313]

Most of the digital chip pins have a protection diodes inside. Two, one to groud and one to Vdd, In closed polarity normaly. But you can measure it.And, in most cases, that is the good way how to detect internal damage of I/O pin (shorted to gnd or power rail) or test if the line is really connected to the chip.

[@fft2020]

You are the best of the best ! You make it look soooooo easy ! great inspiration for me ! thank you

[@bitsundbolts]

Thank you 😊!

[@bzuidgeest]

My guess is to start by checking all connections, starting with the chip enable lines on the memory. Considering it can use variable amounts of memory it likely checks what is there by seeing what it can read write to successfully. So ras cas etc are next. You might also use a scope to check for weird signaling on the memory lines. The color.... I would ignore that until the memory is fixed, I don't think it's likely to be related to the ramdac. But one thing at the time.

[@bitsundbolts]

Agreed! Focus is on fixing the memory amount on the FBI chip. It is the most obvious issue for now.

[@ricargoncalves]

Really impressed this card is alive!! After such damage I had a hard time to believe to be possible. For the FBI issue I would try to see which memory chips are working. From there, check if there is any connection problem for the chips that supposedly should add more memory but don't. Could have melted something in the internal planes, maybe around the area where the PCB was damaged. For the discoloration, my hunch in on the DAC as well, maybe just went bad. What about a voltage issue? Measuring with multimeter is ok, but the voltage regulator could be producing ripple after being exposed to such deadly voltages.

[@bitsundbolts]

Well, at least I know the area I need to look at. Both TMUs seem to be fine. The issue must be in the area below the memory chips in the front since I never got more than 1MB for the FBI chip. I'm curious what others have to say - especially regarding the DAC. But the focus should probably be to first fix the memory issue.

[@argoneum]

I'd think that the FBI chip might have an address line broken (internally fried), probably also some data line(s) connecting to RAM. Or there is some pull-up or pull-down action happening 'cause of charred PCB. DAC might have gotten it too, it shouldn't cause the RAM size issues though, unless it is using the same bus as RAM. Just a hunch, and lots of guessing… And indeed, *very* impressive fix!

[@mal2ksc]

12V isn't going to kill that voltage regulator, at least not directly. Obviously it won't be heatsinked for operating at 12V, but I doubt it was in that state long enough to take heat damage. The machine would have burnt to a halt in seconds, not minutes or hours. They'll typically handle 24V or even 30V input, as long as you cool them sufficiently.

[@krizator]

Hi. Check the resistors and capacitors around the burned memory chip under which the textolite burned out, they were under the influence of high temperature from the burning chip and could change their characteristics or fail.

[@bitsundbolts]

I checked those SMD components - even replaced most of them with components from a donor card. The values of the original SMDs seem fine. After I replaced them, the card still only sees 1MB.

[@DefenderOfBoston-yo2tl]

Hell yeah, it's alive! 🤘 Well, somewhat. We all know that Voodoo magic sometimes comes with unintended side effects. 😄 Amazing though to see that at least some of the irreplaceable chips should still be good. Ballsy move to plug it in at that stage, I would have disconnected the output from the VRM and checked its voltage first. And I would have powered it up while watching it like a hawk with that fancy InfiRay IR camera, ready to power it off at the drop of a hat. I was super relieved to see you successfully helped it get rid of its smoking habit! ✌ That color issue looked very reminiscent of VGA connector issues, my guess would have been the DAC, especially since it is alsodirectly connected to 5V. Glad to see it was just an issue with the TR patch and is already "gone". Still kind of a happy accident since we wouldn't have gotten to see Lara with the correct patch yet, right? As for the lack of FBI memory, I concur with what some have already suggested, checking all the data lines between the FBI and memory chips should be the next step. Hopefully it's just a damaged line between the two, causing it to stop looking past the first MB, rather than the FBI chip itself. If it is, maybe you can fix it by manually adding a wire or by doubling up memory chips like you did on the Voodoo 1 cards? Hopefully, the observed slowdown was just a consequence of the inappropriate TR patch or of it having just 1 MB of frame buffer memory to work with. I couldn't think of anything else causing such an issue on a card of that generation really. As for (hopefully useful) suggestions that have not already been made: Hit it with the thermal camera, comparing it to the identical model you have. Any observable differences may be a pointer as to where (physically) things still go wrong. Since the 1 MB issue already showed up when installing the first 2 MB, it would make sense to be located around there? Seeing a Mega Monster (that's what Diamond called a pair of them IIRC) in action once again would be awesome, I'd love to see some videos about that! I actually built one of these (from two retired cards) for a neighbor back then, it was a great option to extend their useful lifespan for a bit! SLI was way better when it still stood for scanline interleave! 😉

[@mal2ksc]

One known effect of Voodoo is the creation of zombies, which may be happening right now.

[@andrewb6]

I imagine that the FBI issue could result from any of the following: A bad memory chip, or two, or three; faulty trace, either disconnected, or a bad connection is adding resistance/capacitance/etc.; damaged FBI circuitry. The possibility exists that one of the old faulty memory chips failed in a way that send too much voltage along a data or address line to the FBI chip, and it 'popped' an input protection circuit on the die. If a sort of memtest existed for each memory bank on this card, that might make the issue very clear. I am uncertain if that functionality can even be tested in a user-visible way, but since the output is user visible, I imagine a creative solution exists to test the memory. Swapping the memory chips around could also help test the bad chip thesis. Have you not received any sockets yet, or do the 100MHz chips malfunction in those fixtures? Thank you for the quick 2nd video on this! In your own time, for the next one.

[@bitsundbolts]

I didn't put sockets because there is no space on the board. The memory chips are too close to each other or SMDs are in the way. I hope to gain access to something really cool that might eliminate all the guesswork!

[@tarajoe07]

Those old PSUs were all notorious for taking the whole system with it

[@SianaGearz]

Only the bad old PSUs. I have actually worked in computer repair back then at the tail end of 90s early 2000s and we have indeed often seen wide ranging system damage alongside the dead PSU. But not a single FSP (Fortron/Source) unit ever did that, we recorded s bunch of failed FSP since they were actually quite widespread in major-brand prebuilt PCs but all of them failed silently without fire and the computer was completely in order as well. Delta PSUs were at least as good but they weren't quite as common.

[@DefenderOfBoston-yo2tl]

@@SianaGearz yeah, in my experience, FSP and delta are definitely good PSU brands, still making solid products today!

[@Stefan_Payne]

...which is why the Retro Community should promote much safer modern PSU that also have MUCH MUCH lower Ripple than older ones (even when new)... Because stuff like this can happen and the older components get rare and pricy and expensive... A modern PSU is just much cheaper than a dead Retro System...

[@SianaGearz]

@@Stefan_Payne Unfortunately a modern PSU is tuned against modern PC's demands, which are 12V-heavy and 5V-light. Group regulated units are outright out, as they exhibit crossloading issues, and DC-DC topology ones have a wimpy 5V rail, while in that era, the 5V rail demand was kind of high.

[@Stefan_Payne]

​@@SianaGearz 20A is PLENTY of power for 5V heavy systems. That's easily 100W, which most older Systems don't even come close to... As for the Group Regulation issue: Most older PSU are Group Regulated. Hardly any 20 year old PSU is independantly regulated... And if you need more than 20A on 5V, you can go with a higher end one that has 30A on 5V, but the PSU is really expensive. Or you could try to get a movement for "Retro Power" Supplies...

[@JensChristianRestemeier]

I'd try a test program in Glide that writes to the front buffer directly with the usual bit patterns to detect address or data errors. Not sure how much mojo does, it may just read/write until it find the first error, even if it is only in one data or address bit.

[@TrolleyMC]

Incorrect memory sizing is almost certainly addressing related. The FBI isn't able to address anything past the 1 MB range (0x100000). I would try checking address lines from the FBI and seeing if they connect to each memory chip. After that you might just be SOL with that FBI.

[@Vermilicious]

I would suspect a bad trace. Also, perhaps check out that burnt spot you think you saw. I mean, it should be an indicator of something bad.

[@mal2ksc]

I think the black spot is a melted via, which means it was probably connected to the Vcc plane on one end and a part that needs power on the other -- or even half a chip, since the RAM chips have two Vcc pins each and they don't seem to be connected internally. I kinda hope this is true because it's easy to bodge around and there's no worry about signal timing or anything.

[@Sh1neful]

Not only memory chips can be affected by overvoltage. There also can be problems on data/address lines on 3dfx chips, wich were connecred to the burnt memory chips. No one knows what was happened with data/address pins when the "magic smoke" escaped from the memory chips.

[@bitsundbolts]

I agree. I hope this was not the case, but it could be a possibility. That would mean, I have to replace the FBI chip. I have a card that could serve as a donor, but the chip has bent legs 😔

[@danielktdoranie]

Great work, would be great if you used sockets for the memory 🙂

[@bitsundbolts]

There is, unfortunately, not enough space between the memory chips 😕. And if there is, SMD components are in the way.

[@danielktdoranie]

@@bitsundbolts good point!

[@SobieRobie]

Almost there...

[@therealjammit]

I remember there were some mainboards from certain manufacturers where they had what looked like a standard 24 pin ATX power connector but they wired the connector different than from standard. I think some old Dell machines that had the mini style ATX power supply did that. If you think they're using a standard ATX pinout and try to test a machine with a spare ATX power supply, it's possible you could send the 12V to the 5V on the mainboard. If this is what happened I'm quite certain the original machine this came from was totally borked.

[@bitsundbolts]

Sigh, that reminds me of my recent adventure of EISA and OPTi local bus. Same connector, electronically different! Great!

[@attel2091]

I would check that every data, chip select and address line has any activity in the fbi memory with oscilloscope as sanity check before replacing anything. My guess is that something is stuck high or low and is the reason why it sees only 1MB and in bad case it's the FBI chip internal diodes

[@bitsundbolts]

Agreed! Better to get as much information as possible before replacing chips randomly. I might get access to something that will hopefully eliminate all the guesswork!

[@GTXDash]

Yeah, it really does look like the Frame Buffer Processor is the main culprit.

[@bitsundbolts]

Hope I still can find another reason. Replacing that chip would be a lot of work and spares are difficult to get.

[@AlexC-z7e]

in the analog technic ( whoch VGA certainly is) a pinky discoloration is usually caused by a weak or conpletly missing green color. so maybe chech the connection from VGA-green to the digital analog converter in comparision to the red and blue one. fantastig work so far. i enjoy your videos a lot. i hope the card will come back from the dead. fingers crossed. regards alex

[@bitsundbolts]

Thanks! I hope the FBI chip is not at fault. I have spare DACs, more memory, and a donor card for the other chips (I just don't want to replace the 3Dfx chip if it's not necessary). Part 3 will be interesting - can't wait to get it ready, but it may take a bit more time...

[@rogert151]

if you're going to check the address lines for good connections also check for short to PWR GND some lines might be stuck because of PCB damage

[@bitsundbolts]

I checked all those lines multiple times. Nothing suspicious and no shorts. Part 3 is ready with the resolution of what was the issue. I'll release it soon!

[@lovedeathhatepoison8737]

I just wish that I can find a voodoo in the wild in my country these cards are as rare as diamonds in a coal mine

[@memadmax69]

One of the chips for the framebuffer is bad, or bad trace. Could also had high power come out of one of the melted chips onto most likely one of the address or data lines and fried one of the TMUs or FBIs. I'm not too familiar with the architecture of this card, but you could easily find the culprit with an oscilloscope and start probing the TMUs/FBIs for weird signals. Once you find the weird signals, trace back to the ram chip and remove it. If it clears up it was a bad ram chip or chips. If it doesn't clear up, TMU/FBI bad.

[@sebastian19745]

I first would blame the PCB. I would check those vias under the blown memory chips. In my opinion the memory seen is really 1M, maybe some adress/data/control lines to the rest of memory chips have no continuity. If there is continuity from the memory chips (all of them) to the FBI, then I would blame the FBI. For sure, the DAC is not affected (in my opinion) because it outputs all 3 colors well, (I saw R, G and B). Also, that is what mojo also tells you: the memory for the FBI is the culprit. So, test all the lines from EACH memory to the FBI chip.

[@bitsundbolts]

Thanks for the details! Ok, I'll check all connections to the FBI chip. So far, I focused on the chip that sits on the pads I repaired, those connections seem to be fine. But of course, it could be another chip that is the problem.

[@sebastian19745]

@@bitsundbolts I do not know for sure the memory configuration for Voodoo cards, but it seems to me that you have there two memory chips for each megabyte of memory. So, I would guess that there are two banks of 4 memory chips that are wired in parallel (data and adress lines) and the control lines are connected directly to the FBI chip. Maybe at boot, the FBI chip checks the available memory and if not found the remaining 3 memory chips, it assumes that there is only 1M available. This can be even if one data/adress line is faulty. That is why, I would check this first, also because those vias under the blowed chips may be affected by the failure (that first chip had melted metal inside so the extreme heat might caused bad contact in these vias)

[@MrQuist125]

Once Again Nice work sir. :)

[@bitsundbolts]

Thank you

[@VladoT]

I can't believe I was NOT Subscribed. Absurd 😂

[@bitsundbolts]

Oh well, I guess you did now! So, thank you! 🙏

[@charlesdorval394]

Regarding the memory issue, the way it scans for the size might make it stop at 1mb if as mentioned it's a trace issue, but who knows, maybe you're the unlucky guy with a new faulty IC Grasping at straws a bit

[@germanskena6537]

Hello BuB. I love all your videos and as always, give you my like. I want to ask you. Why do you not socketed the memory chips? you said in the previous video you want to do it. And added to that, what happens if one of the memory chips is bad? one of the first chips. Even i see all the video, I get the idea of the first mb of memory have a kind of a problem while you add more memory, the issue persist. That's the strange feeling of the first memory is the faulty. But obviously could be other thing... (but in the video sequence you like... that's the problem, from the watcher perspective i mean).

[@bitsundbolts]

I wanted to install memory sockets, but unfortunately, there is no space. The memory chips for the TMUs are too close to each other. There's no space. Same for the memory chips on the FBI chip. They are spaced a bit further apart, but there's still not enough space for four sockets. The other issue is that there are capacitors and resistors around the memory chips which are another obstacle for sockets. That is why I did not install sockets.

[@NullReference119]

There are... If I'm reading the block diagram on Vogons correctly... 88 lines from the FBI chip to the DRAM, and termination. I have a hunch that some of them may have gotten vaporized when that chip exploded. Likely very fixable with bodge wires. As others have said it's time to start toning things out. That said I'd also check that all the chips are getting vcc and vss, it's not guaranteed those lines didn't vaporize too.

[@bitsundbolts]

I checked all connections to the memory on the front. There could still be an issue with the memory on the back of the card, but the issue with 1MB on the FBI chip emerged when installing the front memory chips. I'll continue to look for the cause, but I'm confident that it can be solved.

[@1kreature]

I'd recheck all the connectivity for the FB memory after you repaired those pads. There is a chance you lost a connection there and since each chip appears to be half a MB I'd say it fails on chip 3 which could be the one you repaired. It most likely detects memory by cycling through the first address of each chip and sees if data can be stored there during init. Common way to check memory.

[@bitsundbolts]

The chip in place of the burnt one on the front is the only chip I have checked thoroughly. I need to check all the other memory chips on the FBI chip and see if I can find the problem. Worst case, I would need to replace the FBI chip.

[@Neria2288]

Continuity problems with data lines for memory for FBI chip? Nice work by the way!

[@leonardoliveira]

First learn how the memory is laid out. Likely it is laid out in chunks of 1MB by 32bits. The second bank is likely the one with the chip that welded into the pads and you had the pads repaired. After that is confirmed you can then check each bank for control signals (RAS/CAS/OE/WR) address lines and data lines. An method of viewing is needed (logic analyzer or oscilloscope). It is very likely that MOJO checks for RAM sequentially and gaps of malfunctioning memory will prevent detection of more memory. Motherboard chipsets do map memory dynamically but videocards are not required to do such a thing.

[@kc9nyy]

Am curious if you remove the 2 chips which had board damage under them if the reported memory would stay the same.

[@bitsundbolts]

One of those two chips was for a TMU chip. Those ones seem to be fine. I'll try and remove some chips and see what I get in mojo.

[@MarkusHorbach-j9v]

the tomb raider ingame scenes miss the green colour channel. You can use a six segment colour wheel to detect which colour is missing. The legs of lara should be yellow instead of cyan, so no green is present. Maybe a faulty DAC or a faulty VGA switch on the voodoo card. Check the green VGA output pin if any signal is present with an oszilloscope. This does not explain the missing memory. There are address lines missing. If data lines are missing, the game should have more picture defects.

[@bitsundbolts]

I will first try to solve the memory issue as the discoloration seems to be related to the 3Dfx patch I had installed on my tomb raider installation.

[@UpLateGeek]

I'm surprised that voltage regulator survived, it's rated for a maximum input of 10V, so I guess those memory chips shorted and blew up before it reached that voltage, or it didn't stay high long enough to damage the regulator. In any case, this is likely what saved those 3Dfx chips. The fact that the FBI chip is only seeing 1MB is definitely suspicious. That suggests only two of the memory chips are working, so I'd be checking the RAS, CAS, and address lines of those chips to see if you're seeing activity on all of them.

[@bitsundbolts]

I will try to figure out what is going on with the memory. 🙏 Thanks for your input!

[@juniorbcm5375]

One of the burned memory chips might have sent higher voltage to the FBI and damaged it. Check if it is getting too hot.

[@bitsundbolts]

I checked the FBI chip with a thermal camera already. Seems normal to me - around 40 degrees idle. Those chips heat up to 70+ degrees under load. So far, I haven't noticed anything suspicious.

[@johnk7134]

If you want me to measure something, please let me know. I have the exact same card both 8mb and 12mb versions. Also had creative one. Put the cards into a better system. Also when i searched for the memory chips i saw that there differences in memories, not only speed but timings etc.

[@bitsundbolts]

Thanks for the offer. I also have a few working Voodoo 2 which I am using for reference.

[@PROSTO4Tabal]

While you test your card in this condition run tomb raider, start touching your voodoo everywhere and see if you can see any changes on screen, this way you might detect faulty area

[@bitsundbolts]

Unfortunately, I tried that already :( no change. Maybe there's something going on in the inner layers of the PCB, but that will be tough to find.

[@timwikblad6119]

Only thing i came to think about is if the board delaminated and some trace is bad somewhere

[@bitsundbolts]

That would be sad, but a possibility.

[@unimportant5122]

I'd remove the RAMDAC and the line driver/buffers and try Mojo. They are connected to the framebuffer memory aswell (the framebuffer contents are what as to be converted to analog VGA by the RAMDAC). Any defect in these chips can cripple the memory aswell, making it "invisible" to the FBI. Since those chips are connected to 5V aswell it's not far fetched for them to be damaged.

[@bitsundbolts]

Hm, so far I everyone is telling me to keep the RAMDAC in place as it seems to be working. Interesting and I am with you regarding the connection to the 5v line. Those two chips are definitely connected to 5v. Maybe i will replace them and see if it makes a difference!

[@unimportant5122]

@@bitsundbolts In my opinion there's only 2 reasons why the FBI does not see all the memory (barring bad connections, which you claim to have checked and are OK). Either the FBI chip itself is damaged, or something else connected to those same memory chips is "pulling them down". That other thing can only be the RAMDAC/line drivers. I guess there's only one way to find out...

[@bitsundbolts]

I might get access to something really interesting! Maybe the guesswork is going to end quickly! Part 3 would be super interesting!

[@AceMcCrank]

When you were repairing the pads for the FBI memory, you didn't show actually connecting the pads to the traces. Could that cause the issue with the missing FBI memory?

[@bitsundbolts]

I did connect the pads to the traces. Yes, it's not in the video. I didn't capture that part. But connectivity to the FBI is there after the memory chip is in place.

[@modlabs]

First i would check the memory modules that used for FBI chip. Second - check the traces around FBI, they could burned. Then - FBI itself.

[@microtrigger]

It seems that this unit determining amount of installed memmory just writing smt to it. And it have common adreses or common data lines for all memmory set - sharing same lines for all memeiry set and I sugesting that some of them damaged via overvoltaged memmory chips. To catch them up jus use multymeasuring unit in p-n mode to see transaction voltages for intire circuits of IC to GND and to io power domane. Shure you will find smth wrong there.

[@alekseystepanov6099]

Increased voltage on memory chips could lead to breakdown of the FBI chip's address line.

[@bitsundbolts]

That would be unfortunate, but is a possibility.

[@Stefan_Payne]

Easiest explanation: Some Addressing error. So either something is pulling addressing lines down (RAMDAC perhaps?), bad contact or an issue with the FBI that got some addressing lines blown... So time to get out the scope and scope the Address line of the Framebuffer?

[@bitsundbolts]

Yes, I guess that's the best thing to do now. Checking all lines on the memory chips for connectivity.

[@Stefan_Payne]

@@bitsundbolts You might even have to go further and get the scope out and do some scoping of the address lines...

[@NetrunnerAT]

Check all trace from fbu ram. Your pcb is realy hard burned.

[@bitsundbolts]

I'll check again. It's probably also a good idea to check the chips on the back since I don't know which chips belong to which memory bank.

[@NetrunnerAT]

@@bitsundbolts good way!

[@bzuidgeest]

What glue did you use for the replacement pads? Must be something special if it can take the heat of soldering.

[@krizator]

As far as I understood it was an ordinary UV solder mask. After it hardens, the contacts will stick.

[@bitsundbolts]

Correct

[@bitsundbolts]

It is solder mask, which I kept outside in the sun for about 30 minutes. Once it cures, it can withstand the heat.

[@Chriva]

Noticed some pink stripes in static positions which would point towards something being wrong with the framebuffer. Don't know much about these cards at all but DACs of that generation often required negative and positive rails, often at fairly high voltages like +-12v that forces you to connect it directly to the psu rails so it could perhaps be borked?

[@Chriva]

Went and looked up the datasheet for that DAC. It does NOT need negative voltages and runs at a comfortable 5v.

[@bitsundbolts]

Yes, but exactly this 5v line was the cause of the memory blowing up. I am surprised that it seems like the DAC is working. The only hint I have is the limit to 1MB for the frame buffer. Would be sad if this chip is bricked.

[@zuldanfpv4634]

Another brilliant video. I have a Janton Voodoo 2 12MB to repair. Any chance you could provide the Aliexpress link for those memory chips? Also would love the link to those 3D printed CPU trays. They look really good.

[@bitsundbolts]

I can't share those in a comment. I need to put all those things on my website. I'll share the details in the next video, but you can search for the memory chips on AliExpress. Look for "M11B416256A-25J". The store is called "Shenzhen Jinjiasheng Electronics Store". Let me know if you find it.

[@zuldanfpv4634]

@@bitsundboltsthank you I found them! I’ll also keep a lookout for the 3D printed CPU trays on your website 👍

[@yanndiy]

Also had a memory detection issue, but on tmus. It was a solder bridge, well hidden. Not sure this is the case here considering your skills 😂😂😂. Maybe a cracked line? Or a grounded line on adress selection lines?

[@bitsundbolts]

Thanks, well, I already reflowed the solder on all three 3Dfx chips. I doubt there's something, but a trace could still be an issue. Although, I have checked a lot of pins on the memory already - so far I didn't find anything.

[@yanndiy]

I really hope you manage to find out the issue. Your videos are really valuable when it comes to understanding how these cards work. I'm really amazed at the "yolo" design: not a single fuse on the 5v line. In the end it did work out well... But still! Thanks again and crossing fingers

[@thebayandurpoghosyanshow]

I usually start with re-flashing the VBIOS.

[@iamdkk]

The QS5244 bus driver is on the 5V rail right? Worth checking it looks right on the scope.

[@bitsundbolts]

Yes, that is the small chip next to the DAC, right? This chip is indeed also connected to the 5v line. It would be an easy chip to replace - low hanging fruit so to say!

[@markborchers6613]

Hi, have you ever tried to get the feature connector working the way it was designed to? Is there anything that supports (and uses) the feature connector?

[@Lilithe]

Is the text on the memory chips rubbing off? Might the FBI chips you installed be bad? It's a long shot but maybe.

[@bitsundbolts]

The chips look legit and original. I rubbed them with a lot of isopropyl alcohol to get rid of the flux from the card. All markings are still there. I'm still trying to find a clue why we have only 1MB at the FBI chip.

[@DarkZenith]

Would you have any interest in working on a matched pair of 8800 Ultras showing artifacting and such? Original owner and they both failed after several years.

[@bitsundbolts]

Those are already BGA memory chips, aren't they? I'm not experienced with this type of repairs yet.

[@DarkZenith]

@@bitsundbolts :( thanks for your consideration though. Love your videos, been watching them for a few months.

[@JensWeiß-h2r]

Maybe the chip is defective, because it was connected to the burned RAM chips. If a RAM IC burnes, then it melt internally and it could be, that the memory connections to the chip get a higher voltage. Then the FBI chip will die. You can propably measure it. In overvoltage conditions the internal clamping diodes die. This reults in an effect, the the pin didn't get a proper high-level. And you can take pictures with a thermal camera. In some cases you can see, that there are faults in higher temperature. Best wishes

[@DefenderOfBoston-yo2tl]

That would be the worst case scenario I am afraid of. I also think that observing temperatures with IR, and comparing it to the identical good card, would be a good way to proceed and physically locate the origin of the issue. Would be such a shame if it's the FBI chip, losing anything that has 3dfx written on it is always sad. Let's hope it's just one of the lanes that got toasted.🤞🏻

[@bitsundbolts]

Well, we could also see a positive in that we recovered two TMU chips 😂 - glass half full mentality!

[@wewillmakeit3615]

Did you check all the 5V rails to all new chips separately? Maybe one of the lines melted inside the PCB when that PSU failed.

[@bitsundbolts]

I checked those lines before I soldered the memory chips. I don't think the 5v lines are the issue, but maybe a broken/missing address line. I will investigate further. Thanks 👍

[@ajdothack]

you could try removing the chip that had the most damage under and se if something changes, if not try to check there are not shorted data lines between them also observe the data lines whit and ociloscope to be sure nothing is missing.

[@kokodin5895]

very interesting problem if we could determine which memory channel is reported working and which ones are not put sockets on the card and play with memory chips untill you see the difference? if a card can work with just 1 mb buffer it should have 4 2-chip pairs "memory channels" it would be great if we could manualy pole memory areas because at this time we have no idea if the momory you installed is actually not the faulty chip in a batch, we have no ideahow badly blown up memory backfeed to the graphic engine, it would be worth checking if blown meory modules have shorts between dara ,adress and power pins because you may be licky and memory contriller is working just missing some pull up or pull down resistors and that could be fix externaly if you unlucky the whole chip is in zombie mode and you will have to replace it you could check with oscilloscope how data lines look on afected memory or just with a logic probe it is very wierd that movies get pass trough in full color and the 3d is wacked because it is not just missing color from a dac colors were there but all wrong good luck

[@bitsundbolts]

Thanks! The movies are coming from the primary VGA card. Only 3D scenes are rendered by the Voodoo 2. That is why we need the loop cable to be able to seamlessly switch between the 2D and the 3D VGA signals.

[@clintcolombin]

Wild guess, but could there be a bad passive related to the memory chips around the area of the damaged chips? The only way to figure out which one would be to trace each memory chip though

[@bitsundbolts]

I replaced the caps around that one chip with the hole and measured all resistors in that area. All seems to be fine.

[@knightsun2920]

I have one voodoo2 card with 12 chips and another one with 24 chips bot are the 24 MB versions. I don't know if the bios is different.

[@bitsundbolts]

I don't think those cards have a BIOS. Simpler times 😉

[@xenoxaos1]

Take a look at the chip select on the 4 chips with a scope?

[@2001pulsar]

I'd suspect the plated through hole under the TMU ram is not connecting all the chips to the ram bus correctly.

[@Aeduo]

Shame it seems to totally prevent initializing in windows. Would be neat to have a wacky rendering mode to make all your games surreal.

[@bitsundbolts]

Haha, true! But even tomb raider doesn't work with the proper patch installed. The moment the game goes into 3D mode, the screen remains blank.

[@Aeduo]

@@bitsundbolts ah I see. So the wacky colors were just unrelated.

[@JuanManuel-h1h]

Hello nice video, do you have de Datashit for de Voodoo 2 from Diamond ?

[@TerroMin]

Try look with oscilloscope address line on memory chips, maybe some lines are corrupted

[@SidneyCritic]

Maybe check the chips are getting 5V, ie, how did the traces survive when the chips were totally destroyed. Maybe scope the data lines for activity. The PS for the onboard SSD on the 19" Mac Pro are feed with 14V, and that PS chip routinely shorts killing every SSD chip.

[@bitsundbolts]

I can try tomb raider with the voodoo 1 patch - at least the game starts and I can probe the data lines - see if there are chips that don't have activity

[@awilliams1701]

I'm wondering if you have a broken trace causing the 1mb. If I recall the chips that exploded were FBI chips. So you should check the connections from the ram to the chip. I know that sucks, but sometimes that's the way it goes.

[@bitsundbolts]

Yes, it is a lot of work, but I guess this is the way forward. Test everything and rule out certain things to narrow down the problem.

[@SianaGearz]

Unfortunately the AMR on the regulator is 10V. I wouldn't leave it in if sourcing a good fresh replacement is possible. Which i know might be a problem, i wouldn't trust Ali parts more than this one. I suggest checking the IO pins that connect DRAM to framebuffer chip for signs of latch-up or latch-down, i.e. that the ESD diode or output circuit on the pin is burned and constantly pulls that pin. It may show as a short to 3V regulator output in case of latch-up. Or some sort of anomaly in resistances to both that and GND in comparison to other pins. Obviously the RAM ICs in the framebuffer circuit have suffered the most. I don't think the DAC is faulty, since it can output both black and white; furthermore gamma ramp commands don't mess it up at all, when the game fades to black and fades in. I think the whole problem is between framebuffer controller and its memory.

[@bitsundbolts]

I have a donor Voodoo 2 that should have a good voltage regulator. I could use that. Then I need to spend more time investigating the memory.

[@Rod_Knee]

Perhaps the traces to the frame buffer RAM on the lower right hand of the board took more damage than was visible? You may already have rechecked the ground and +5v rails to those chips, but if not it may be worth it.

[@bitsundbolts]

I'll measure the connections to power once more. There are more VCC and VSS pins along the chip. Maybe I missed something there.

[@Rod_Knee]

@@bitsundbolts Are address lines in close proximity to the board damage? Edit: You channel it great - long may it continue!

[@cocusar]

when one of the ram chip died, might have shorted an address or data line of the fbi, or tmu, so I'd just replace the fbi and see if it works

[@bitsundbolts]

That would be sad if true, but it is absolutely a possibility. Let's see if I can find out something else before I replace the FBI chip.

[@laurdy]

When you put the new solder pads in I didn't see you solder them to the existing traces, myybe it's not making contact. PS: Those linear regulators are rated for an input voltage of 10v but are characterised up to 35v! I doubt it was damaged.

[@bitsundbolts]

Yes, I didn't record that part, but I did solder the pads to the remaining traces. I also measured the connection to the FBI chip - all is good. Yes, good that the voltage regulator survived and saved the 3Dfx chips.

[@MeganFoxx426]

Frame buffer chip could be partially damaged, maybe replace the frame buffer chip.

[@bitsundbolts]

That would be unfortunate, but it is a possibility. Let's hope there's something else that is the problem.

[@R.Daneel]

Could it be a faulty connector/cable missing a colour channe pinl? I thought maybe the pass-through cable is ok, but... grasping at straws and it's an easy check.

[@bitsundbolts]

I know, it's a lot of guesswork. But I know the pass-through cable is working. It could be a connection issue with the ram chips. I need to investigate further.

[@jaybird57]

Check resistance on all fbi memory lanes. See if any are different.

[@GabrielTibola]

Hi, where can I find the design of thos 3d printed socket trays? I would appreciate if you could share those, thank you in advance!

[@bitsundbolts]

I can't post links here, but go to the Thingiverse website and search for "Stackable CPU tray holders for various sockets". You should be able to find stackable trays for various CPUs. You have to pick the one you need.

[@garysweetland32]

Where did you find the memory chips? I have an 8MB card I'd love to upgrade. 🥴

[@bitsundbolts]

I got them from AliExpress. There is a seller that seems to sell genuine parts. I bought them some time ago though.

[@SimonZerafa]

I'm not a 3DFX expert but if you set up an SLI system then would that show what might be going on? I assume you have tested your DOS/Tomb raider setup with a known working card? 😉 Best guess is a broken trace or via from the FBI to the RAM 😢🤷‍♂️

[@bitsundbolts]

Yes, I tested a known working voodoo 2. That is when I realized that I needed a different 3Dfx patch (which is probably the cause for the colors). However, I still have the issue with the 1MB on the FBI chip. I'll test the traces once more .

[@SimonZerafa]

@@bitsundbolts Also I wonder if those RAM chips had 12 + 5V coming in? The last time I saw damage that bad was a lightning strike on an analog modem 😣

[@bitsundbolts]

@SimonZerafa whatever it was, it must be some sort of overvolt from the power supply. Unfortunately, all this happened 25 years ago 😂 ancient history by now.

[@BrendanRaymondKoroKoro]

Discolouration could be due to damaged rgb amplifier chip?

[@bitsundbolts]

As it turns out, tomb raider has a dedicated patch for voodoo rush and voodoo 2 cards. Without that patch, the game looks like this. I updated the game with the correct patch, but now, no 3D scenes are rendered anymore. Only with the voodoo 1 patch, I can enter the game. And there is still only 1MB available to the FBI chip.

[@cricketol]

i wonder if the memery chips wired to the frame buffer could that be an issue?