Thanks for the shout out! I'm glad that this heavily burned V2 board is back to life! 👍🏻 The software is still "work in progress" and it can do very little at the moment. Also all development is done on Linux, so this DOS-loadable package was rushed together and is not perfect. The idea is to get a Linux kernel that runs on 486 and up, the utility itself is compiled to run on 486 (but not tested on it). I've already fixed some visual bugs to make output clearer, added verbose and help options. VGA output from the Voodoo is not usefull at the moment. And I have trouble with direct access to upper 2 MiB of FB RAM and with accessing TMUs at all. Any help I can get is great, so feel free to let me know some magic that can allow that.
Woah. Didn't expect to see you here. I didn't watched your streams for last year or more. So I thought that you digged deeply into VCRs, cameras and so on.
Very interesting! I wonder if your tool would work for debugging Voodoo3 cards as well. I have one that has some weird memory problems as well, which persisted after replacing the memory chips with known good ones. And as a bonus I am also a Linux guy and would be comfortable running your tool natively.
Your software is a godsent already, even with what it currently does. Just think about it, any error it diagnoses can save a card. Any error it can't diagnose isn't worse than what we had before: dead cards on a shelf. I'd really like to test the current build, even if it had to be run natively on linux. PCI Slots exist on late enough boards to have a dedicated linux-based tesing setup that could run recent linux kernels.
that "YES! It works!" moment never gets old. Well done! Huge thanks to Maxime as well, waiting on the tool release to check what's wrong on a voodoo2 with crushed pins on the fbi :)
Haha, thanks! That tool is definitely a big help! And it was the one that made me replace the FBI chip. Unfortunate, but at least I didn't waste time on DAC and the buffer chip.
"PS/2 is not hot-pluggable, let's double check that" :D This is now your motto! :D This software is WICKED! It even shows you a text diagram of the board to tell you what pin you need to check. Super Super impressive! Let us know why the RAM test was with blue background. Well done in fixing this board!
I will check with Maksim if he has a clue why it was blue. I tried it a second time, and it was blue again. I cannot see any issues during game play though and the debugging output in Witchery also doesn't show anything.
My guess is the colours are not set by the tests and are based on uninitialised registers in the GPU. I feel like the testing doesn't bother to reset them when checking memory.
Great video, as always (and I really mean it)! I like your new film-making approach, it feels more natural and easy going. I really learned a lot from you about repairs and you are one of the reasons why I have retro bug now and why I now have PCs I wanted as a kid or teenager. I now regularly repair all my stuff I get from eBay. Thank you very, very much! Greetings from Montenegro!
So many seems to be impressed by your soldering skills, I say fuck that, soldering is easy, what's really impressive is your patience and fine motor skills when dealing with bent legs on various chips. Keep up the good work!
Haha, thanks! It is challenging to work on those small components. The microscope helps a lot, but I am thankful that my motor skills are quite good! Otherwise, this would be a totally different experience. Thanks again! 🙏
I am so happy you got this card running. This was one heck of a project, but you did an amazing job! And you make it seem like changing the 3DFX Chip was easy :D
Changing the 3Dfx chip wasn't that difficult to be honest. I think the most difficult part is to lift the chip without bending the pins. I should get one of those suction pens. A little bit of low melt solder might help a lot to remove those large chips.
The fact that Witchery! picked up exactly WHERE the problem was occurring on your voodoo 2 card is AMAZING! The saddest takeaway from your video: I SHOULD'VE NEVER SOLD MY VOODOO 5. Waaaaa!!! 😭😭
When PCI (and other) devices are listed as "Vendor 8086", then it's always Intel. Makes it easier to filter out devices when looking for a specific one (or better understand missing drivers in Windows Device Manager). It's a matter of "It's obvious once you know it"
Thanks, really enjoyed watching! Software showed no RAM, 'cause after de-soldering the two RAM chips the bits that are *always* used for data transfers didn't work. It's like having a road too narrow for a car: no matter how long the road is: the car can go zero length of it. The witchery software does testing on way lower level, so it reported specific pins that didn't work (impressive). After this video at least I know that RAM and DAC have separate buses :) A tip: diode mode can be used to validate trace continuity between chips (3Dfx - RAM, 3Dfx - DAC, etc.), it only shows clamp diodes (google "CMOS clamp diodes") between chips leads and supply rails.
One note. DAC and RAM have separate buses but not exactly. Actually DAC has three (3) buses. Address bus (3-bit) and data bus (8-bit) for setup process and data bus (16-bit) for actual pixel data for the display. The first and third ones are separate but data bus for setup has shared pins with FBI RAM. Those are intersected in a weird way. So you can not setup DAC when RAM refresh works and vice versa. That's why there is a bit in config space register for remapping some registers between normal workflow and DAC operations.
Great bit of detective work and a very nice result. It’s really nice to see yet another Voodoo card saved. I can see you put a lot of effort in to this
A-M-A-Z-I-N-G work Alex, congratulations on successfully giving this card a new lease of life! I'm so happy your dedication, hard work, perseverance and talent paid off in the end! It's a bummer that the FBI chip is beyond repair (...or is it? Maybe it will work with an external PU/PD resistor, like some suggested. Can't wait to see a video about it!). Either way, it still means that 2 TMUs and an FBI chip that were out of commission have made a triumphant return! You know what to do next: Go enjoy some Interstate '76´🚗🔫🚀 on it, you've more than earned it! And let this be another lesson to us all not to cheap out on our PSUs! I'm really looking forward to some MegaMonster™ content! I can't believe I am still playing the old classics that originally ran on these cards. Just last week I finished another replay of Vampire: The Masquerade - Redemption. 🧛♂😅 And of course many thanks to Макс for his awesome tool that takes a lot of the guesswork out of Voodoo repairs! I wish all your fellow countrymen were as kind and open-hearted as you, it gives me hope to see how our love for this old technology can bring us together in these dire times.
Haha, yes, I'll try Interstate 76. This Voodoo 2 was a great project. I learned a lot, transplanted a 3Dfx chip, and brought two Voodoo 2 cards back to life (the other one was from the scrapyard). So, yes, what a great ride! Thanks again for the opportunity to use your card and make those videos!
@@bitsundbolts I really envy you for seeing Interstate '76 with fresh eyes. Let me know if it's all rose tinted glasses on my side or if it's indeed still enjoyable almost (...holy s***!) 30 years later. 😅 I'm the one who has to thank you for fixing the card! It's definitely yours at this point, but I'll be glad to buy/trade it back in the future, once you've exhaustively run it through its paces!
Yay, now we finally know the FBI was damaged. The tool looks really nice and will save some of us a lot of time and effort for sure. Really looking forward to the release!
congrats to both of you, BuB! awesome series, happy to see it working again. these 3 parts were way way better than the latest marvel, dc, disney movies combined :))
Thank you! I enjoy those projects too. Unfortunately, I'm running out of broken Voodoo 1 and Voodoo 2 cards. Later models have BGA chips which I'm not yet confident to work on.
Thank you! I wasn't planning on releasing three videos about this card, those projects are unpredictable! Nevertheless, when such a project comes to a successful end, I am overjoyed to have learned something and I can give back the card to the owner who is watching my videos eagerly awaiting what is going to happen to his Voodoo 2 from 25 years ago 😄.
Thanks for all these great videos! It would be cool and i think useful if he updated it to send the debugging output that goes to the console to also output on the computer's serial (COM) port. that way you could plug in a serial terminal and watch in real-time the output of all these tests as the main screen is flashing bars! In any case this is amazing and if my voodoo card ever has an issue, or if i just want to check this tool out, I look forward to testing it when he releases it. thanks!!!
shorts on data lines can be caused by burnt internal pull up or pull down resistors and in some cases you can suplement it externally either on the controller chip or on the memory module alone if pin is "dangling" i would assume voodoo has pulldowns since it doesn't work on chip burnt by overvoltage so if you didn't want to replace the 3dfx chip you could bridge pin 2 with resistor ~100-500k (i think you could check it on any working chip to ground or power if i am wrong) and do a getto fix if it is just a pull up on the bus it would work if not it would be just swaping the controller like you did but i don't advice to put bad chip back to test my theory :P
Well, I still have the chip and the donor board. You're the second person suggesting this fix. So, you mean I just place a 100K resistor between 3.3V and the broken D0 (Pin 2) line in question? I could try that on the donor board.
@@bitsundbolts if you had an oscilloscope you cuould mesure logic level voltages on memory or just get the value from memory datasheet if it is 5v chip you might have 5v logic levels, if 3,3v then 3,3v the biggest thing is to figure out if memory is shorted to ground, dangling or is there interna pullup without pull down or vice versa . if the damage is just on pull ups it is as simple as bridging data pin with a resistor to whatever voltage it should be pulling to, if it is internal damage on decoding logic or the output of a logic gate in a buffer or something, this would not fix or break anything it was just a silly idea about how to potentially getto fix a broken chip kind of zombie voodoo
Well, anything might be possible with those Voodoo cards 😉. I might try to revive that donor card. I just need to get my hands on the 3dfx chips and could try the Brocken chip once more. I'm just a bit worried due to the high frequency. I'm facing similar issues on my 16MB SIMM modules with the level shifters.
Super nice tool, especially the clear information on where the chips are and what might be the problem. Did Maxim create all kinds of errors on his test rig to know the errors or is he just a god among mortals.
I agree! The information and error description of that tool are amazing! I think he did simulate errors by purposefully modifying Voodoo 2 cards. It's a slow process and deserves a huge shout-out to Maksim's work.
Multiple hours were spent buzzing out connections, stepping through adresses in software while probing some pins, shorting stuff on working cards to see how it will affect results. And then I've bought a dozen of broken V2s and started to run tests on those.
Great tool, it makes debugging way way easyer. Instead testing for long hours for shorts, continuity, open traces or bad solder contacts on each pin, this tools in 10 minutes gives you the needed info about what is not right. Who would suspect that the FBI chip have one data bit stuck low? You could have find the problem but it would took much more time and more than a DMM.
Correct! And I couldn't spot anything with the multimeter. I tested so many pins on the memory chips. Diode mode, continuity mode, etc. I wasn't able to figure it out. In the end, I would have to guess and I would have found it. The tool gives you the right hints and will save you time!
@@bitsundbolts In the end you would had to use an oscilloscope, logic probe, logic analyzer to see in real time what happens with the adress/data bus on the memory chips or FBI. Tools that most mortals do not have and are expensive and hard to get (good ones, I mean). I wonder if this tool will be free and what will be able to do in its final revisions, maybe full test any Voodoo card or I am expecting too much? I mean, and you also said it, most of the documentation for those chips is not available, so is voodoo magic...
@sebastian19745 Unfortunately, the tool supports only Voodoo 2 cards for now - I might have said Voodoo 1 as well in the video. That was a mistake then. The tool will be open source and hopefully, with more people contributing, Witchery will support all Voodoo cards. But it will take time as you said - it's all voodoo and witchery ☺️
Great work on this repair! I love your channel and your work :D I notice you use an OSSC as part of your video capture setup, what capture card you use for the hdmi output?
Thanks! For HDMI capture, I use a very cheap USB capture card I found on Amazon. I had to go through many before I found one that works well and responds quickly to resolution changes. I had an Elgato Cam link 4k as well, but surprisingly, that somehow didn't work well (especially with certain resolutions under DOS). The USB capture device was around 15 USD and is silver. Unfortunately, there is no guarantee that it's still the same device. I'll add a post to my website about it in the coming days.
Awesome job! That really is a useful tool for diagnosing problems. Sad that it was the FBI chip, but not unexpected. I'm surprised it even worked that much! So thanks to Maksim for helping save this Voodoo 2! Anyway, you have far more patience than I do for straightening pins, and a much steadier hand! I definitely would've put it in the too hard basket once I worked out the memory was fine. At least if I come across a board with a chip that has mangled pins (or a chip with mangled pins), I could donate it to you so I can just watch you fix it instead! 🤣
Thank you! Ha, I hope you will have plenty of Voodoo 1/2 cards - definitely keep them if you find them and I can try my luck on any of them. The ones that won't be recoverable stay with me for spare parts, the working cards can go back to you :)
Great job! I guess the burned chip damaged the FBI chip and made it report bad memory, no matter what chips you replaced. One more Voodoo card lives on! This encourages me to pick up a multimeter and test out a Sapphire Radon 9250 card I've got that isn't detected by the AGP slot, I've got 2 so I'll be able to compare values between the good and the bad card.
@@bitsundbolts Unfortunately I don't know how to check it. In fact, I don't even know which one is the BIOS chip 🤔 I might try googling the inscriptions to see if I find it, then try to swap them. But before the soldering iron gets involved, I'll try to measure each component's value and compare to the good card.
Maybe there is a flashing tool you could use to see if it detects the BIOS chip. I haven't done this, so I have no experience. Use an old PCI graphics card if you have and keep the Radeon 9250 in the system. Try to flash the BIOS. It might work. I have two videos on my channel where I flash once a GeForce 2MX and once a Voodoo 3 using DOS flashing utilities. Both cards suffered from a corrupted BIOS. After flashing, both cards started to work again.
@@bitsundbolts Actually I've done that already, I've used a PCI card to boot from a USB floppy drive, and used FLASHROM.EXE for ATI cards but it doesn't detect it. I've used this utility to flash Radeon 92xx cards with modified firmware and turn them into ArcadeVGA cards which can connect directly to a RGB SCART TV at 15khz, when something does wrong I just insert a PCI card and flash the original firmware back. It worked successfully with a MSI card and a Sapphire card, but this one (which is EXACTLY identical to another working Sapphire card) isn't even seen by the system, it just ignores it and either boots directly into the onboard graphics or just beeps if the board doesn't have integrated graphics. Tonight I'll try to do a continuity test on some of the components and compare between the good and the bad cards to see if I can find any difference.
Well done! I am impressed with your effort and detail level! Congratulations on a success! (regarding the crng inthe outpout I think it is the Linux Random number Generator sending messages to stdout)
I haven't done any overclocking on Voodoo 2s yet... There are models with 110 MHz memory, but those memory chips are hard to come by. Not sure if it's worth it. And if I'm not mistaken, the Voodoo 2 chips are running at 90Mhz - so, there should be some headroom for overclocking anyways. I'll try that maybe on one of my own cards in the future.
@@bitsundbolts I remember pushing the old Voodoo1 chips from stock 50 to 62 Mhz easily. Even Powerstip was displaying a flame icon next to the speed slider. Looking back, I wonder how many ns were those chips. Anyway, thanks for the video, I'll probably try Witchery on the ol' V1 just to check the tool when it is released. Were it not for your video I would probably never hear about a new tool for my 28yr old card, heh.
@Thomsonicus I'm afraid Voodoo 1 cards aren't supported yet. I may have said it wrong in the video. But I am sure support can be added quickly after the tool has been released. I never got any Voodoo 1 stable past 57 MHz. I had 35ns memory chips on those cards. I guess you were lucky with your model!
if this ends up working w/ Voodoo 3's i might be able to eventually find out what's going on w/ this Gateway branded Voodoo 3 1000 AGP i have on hand. i'll haveto keep an eye out for when the Tool's Released.
At the moment utility works only with V2. In theory V1 is pretty similar so it probably will be supported later. For V3 I briefly looked at the datasheets and I don't know yet if its support will be easy. No promises on that yet.
It will be released as open source as per Maksim. He's still working on it, but maybe, with all the positive comments, it may be released sooner than later.
@@bitsundbolts at least it checks the FBI ram chips at the moment - to exclude those 8 chips if your trouble shooting a artifacting/broken card - it will be great when it can check the TMU ram chips aswell - better to just remove one bad chip than the whole 24 ha ha - BTW i knew urs would need that replacement FBI chip - as it was directly connected to the blown ram - but what a great tool to prove it first
I am curious what the pin looks like underneath the ceramic now. If the damage isn't too deep, you could maybe dig it out and re-solder it internally lol
The main lesson here is to go back in time and make sure people don't buy cheap PSUs. Had a bad PSU in 2005 go pop and it charred a pin on the 24 pin connector but luckily it didn't burn my P4 Northwood and Radeon 9800 Pro , was a PSU that came with a case and I thought i had gotten a deal, turns out that any time you get a PSU along with something, it's likely sketch.
The PSU and the cooling are the key components i never compromise on, even on cheap systems, it's just not worth it endangering an entire system to save a few bucks/dollars/euros. Brand name or nothing for that, people who don't like that can buy their computers at the food discounter and leave me alone with it.
Yes, there were a few suggesting that the FBI chip is broken. That is one of the chips that is hardest to replace and I only did it after there wasn't any other choice. It's very unfortunate since it seems to be just one of the lines that killed the chip. But it was coming from that one memory chip in the bottom right corner that had that big hole burnt in the center.
i do wonder what can you use to clean up very stubborn solder paste ? i 'v tryed isopropanol but often its just dilute the paste and make a more of a mess. I could use hot water with soap and a tooth brush but thats add some other inconveniences, and even then it isnt cleaning everything. Any tips for any chemicals? Also very interesting and good video, i could almost feel the anxiety passing trou :-P but big reward in the end.
Unfortunately, I don't know how you could deal with your solder paste. Maybe try to heat it up with a hot air station to 60-80 degrees (so the solder doesn't melt, but the flux will lower its viscosity), and then try again with a cloth dipped in IPA. I assume you have those tiny solder balls everywhere on the board?
@@bitsundbolts my hobby involve crafting electronic dohicky or prototyping so tiny is debatable, its not as small as what you working with, but my issue is when i need to melt and fuse solder with some stubborn metals contacts that work nicely with paste, but paste is such a mess and sticky. mostly i use clear pcb liquid solder flux, but i don't want to use all of it and im looking for alternatives, like paste . I will try the heat and clean . or maybe i should get a bottle of Acetone.
46:29 that looks like a linux kernel message, crng is a cryptographic number generator, it seems like the init was quite late so the kernel just spat that message out in the middle of the console, imo unless the test uses crng it's unrelated to the test and nothing to worry about
It wouldn't change much, but I wondered why haven't you checked the continuity between suggested FBI pin 64 and pin 2 of U6 & U21? Clearly, the databuses of these chips are in parallel. The 'crng init done' with a timestamp during the last Witchery run, was just a Linux kernel debug message (you can see a lot of them during Linux startup).
That donor card deserves to be framed as saving of X and Y cards. Maybe in the future there will be a process to can open that FBI chip and repair the die issue with that memory channel.
Haha, there are already suggestions on what I can try to save the FBI chip. But now, I am also considering saving the donor card - I just need a few NOS 3Dfx chips.
I did suspect the FBI chip was the problem, this is a bit of a sad repair as the broken chip is one less of a finite supply rather than capacitors or other simple components that you can get brand new. But I still loved that "YES" moment when it worked! Thank you for another journey, and thank you to Maxim? / fagear for a nice tool.
Nice job! You shouldn’t be that much surprised by the 0MB without chips or with faulty U6 access. The FBI chip uses 64-bit bus to access memory so you need 4 working chips to get the first 2 MB of FB memory. In this case I guess it failed memory test on first byte so effectively crashed the system and reported 0 MB. I wonder (but maybe I missed it in previous episodes) how scope view of faulty bit looked like? If the chip was fried it should never reach out L or H and be a bit different from other bits.
I did not scope every data pin. It is in an unfortunate location and difficult to reach with the probe. I do have the Brocken FBI chip though - I could resolder it to the donor card and test it with a scope now that I know where the issue is. I can just test one of the corner pins of the FBI chip.
Man, I wish I had this tool. I have an even stranger Voodoo 2 (also with an FBI issue) on which I replaced all memory chips (with new ones) and also the FBI itself with a NOS chip and the problem still persists... And my card looks like new, no broken traces, no missing components (it sat on a shelf for 2 decades, in an antistatic bag). Edit: btw, "Random: crng init done" is output from the Linux kernel (Cryptographic Random Number Generator initialization) which is displayed on the console output with a certain delay. It's not related to Witchery. 🙂
tool for voodoo diag is not public / free? Is usable also for other 3dfx cards? 🤔 good work voodoo working again 🥰 I am waiting atm for two V2 12MB so i hope i will try SLI mode 🤗
I wonder if it's just the chip select line for that bank of 2 ram chips. If that's all it is, maybe a little bodge of some logic chips to make up the missing signal based on the other 3 would work. Just a thought. Ping me if you want to try something like that
Hey, I absolutely would. However, the shipping cost would make this endeavor very costly if I'm not mistaken. I believe it would come close to just buying a 12MB card locally. It only would make sense if the card travels with someone in their vacation luggage I'm afraid.
This is what Maksim explained to me as well. U6/U21 seems to be the first MB, If this isn't there, 0MB are shown. For me, one chip (U6) wasn't working, so I got to 1MB, but no further because there was an issue.
Thank you! Yes, I did put quite a bit of time into this Voodoo 2 card. But in the end, the card is saved, I learned a lot, and the owner gets his card from his childhood back!
Maybe you already did this and I just missed or forgot (I'm very forgetful), but did you check the pins of the FBI chip for loose pins before replacing?
@@bitsundbolts Ah, ok. I was thinking of going back and checking, but I didn't feel like skipping through previous parts. I know I watched them as I gave them a like, I just forgot already.
I only tested the traces with a multimeter. Some of the data pins are difficult to access with a multimeter. Especially the one in question is right next to the 5V supply pin. I still have the 3Dfx chip and I could solder it to the donor board if I ever attempt to fix that one too. Then I could try and see if it would be possible to spot the issue with an oscilloscope. And now I know the pin in question is at a corner of the FBI chip.
I've seen this suggestion! That would be awesome if it would work. I just wonder what resistor value to use. The capacitance of the memory is about 5pf. Not sure about the 3dfx chip. I need to get a value that works at the frequencies of the data signal. I guess I could experiment. Start with a large value (e.g. 10K) and work my way down.
2 questions: first: why would you try to repair a so badly burned card instead of making it the donor and repair the previous donor it couldn't be more broken could it? second: would the FBI chip work with an external pull up on that pin? Its a pity you did not test that now its to late, or may be you could test the on the donor?
The donor card I have was handed down to me and it had many parts removed or replaced. Additionally, the spot for TMU0 has a few broken pads. So, the PCB of the card with the exploded memory chips has a PCB in better condition as well as all the original SMD components. And I think reviving such a card is a greater achievement than reviving the donor card. To answer your second question, I am planning on reviving the donor card once I get an FBI and TMU chip. I will try the external pull-up resistor. Unfortunately, I have to solder the FBI back to the board, but I'm willing to do that. It would be cool if this TMU could be saved by an external pull-up. Why didn't I test it while I had the chance? I didn't know or it didn't cross my mind to do this. I'm not that experienced with electronics repair, but each time I learn something else thanks to my audience!
Hm, interesting... I still have the chip - and a donor board... and a pullup resistor... But could you elaborate on your "questionable fix"? I assume it is just a shot in the sky to add a pull-up?
@@bitsundbolts Well, by not having a short on that pin we can assume that the pin protection diodes that are connected between the pin and the VCC and GND respectively are not shorted. The only thing that is wrong is probably internal mosfet of that output's gate that turns that pin ON (connects it to VCC). So if the other mosfet is OK that means that this output can in fact pull down the pin but there is no active device to turn it ON (pull up). So by adding a resistor we can provide that pull up however we don't know if the rise/fall time will be adequate (I once managed to make this work but not at 90MHz for sure) or we don't know if the output is sometimes put in it's third state (high impedance) so this repair attempt will be quastionable at least.
Is this tool also works for Voodoo5 cards? Ohh I hope Maxim will make a tool for Voodoo5. Please mention to him. Soo badly needed! Especially that there is not even a mojo for it. Thanks.
I am sure he gets this a request to support every Voodoo card out there, but it must be incredibly difficult to support them since there is very little documentation for some of those cards. If he releases the source code, I am sure the community will slowly build on it and create support for more cards over time. We just need to be a bit more patient.
i dont know why. maby it's just the mood i'm in but i heard the typing and was like ... nope not working then slap slap slap.. Well wrong keyboard :p lauging my but odff. Thanks for the blooper.
I also repaired broken Voodoo 1 by fixing 2 FBI chip traces and some pins. Nice work. I prefered to not use hot air, becasue the pads can fall off. Whole restoration video is on my channel.
Thanks for the shout out!
I'm glad that this heavily burned V2 board is back to life! 👍🏻
The software is still "work in progress" and it can do very little at the moment. Also all development is done on Linux, so this DOS-loadable package was rushed together and is not perfect. The idea is to get a Linux kernel that runs on 486 and up, the utility itself is compiled to run on 486 (but not tested on it).
I've already fixed some visual bugs to make output clearer, added verbose and help options. VGA output from the Voodoo is not usefull at the moment.
And I have trouble with direct access to upper 2 MiB of FB RAM and with accessing TMUs at all. Any help I can get is great, so feel free to let me know some magic that can allow that.
Woah. Didn't expect to see you here. I didn't watched your streams for last year or more. So I thought that you digged deeply into VCRs, cameras and so on.
Very interesting! I wonder if your tool would work for debugging Voodoo3 cards as well. I have one that has some weird memory problems as well, which persisted after replacing the memory chips with known good ones. And as a bonus I am also a Linux guy and would be comfortable running your tool natively.
great piece of software. excelent work Макс.
Are you planning to extend the tool to work on Voodoo5 cards? That would be amazing, since there is absolutely no software available for them.
Thanks.
Your software is a godsent already, even with what it currently does. Just think about it, any error it diagnoses can save a card. Any error it can't diagnose isn't worse than what we had before: dead cards on a shelf.
I'd really like to test the current build, even if it had to be run natively on linux. PCI Slots exist on late enough boards to have a dedicated linux-based tesing setup that could run recent linux kernels.
that feeling when you've been following Maxim for over 7 years and you find out that he watches the channel that you watch.
that "YES! It works!" moment never gets old. Well done!
Huge thanks to Maxime as well, waiting on the tool release to check what's wrong on a voodoo2 with crushed pins on the fbi :)
Haha, thanks! That tool is definitely a big help! And it was the one that made me replace the FBI chip. Unfortunate, but at least I didn't waste time on DAC and the buffer chip.
"PS/2 is not hot-pluggable, let's double check that" :D
This is now your motto! :D
This software is WICKED! It even shows you a text diagram of the board to tell you what pin you need to check. Super Super impressive! Let us know why the RAM test was with blue background.
Well done in fixing this board!
It's hot-pluggable in Linux.
@@pvc988 I thought it could be damaged if hot-plugged? (note "could")
@@tony359 Yeah. It doesn't have proper mechanical provisions for being hot-plugged. But I've never, ever damaged any PS/2 port or device that way.
I will check with Maksim if he has a clue why it was blue. I tried it a second time, and it was blue again. I cannot see any issues during game play though and the debugging output in Witchery also doesn't show anything.
My guess is the colours are not set by the tests and are based on uninitialised registers in the GPU. I feel like the testing doesn't bother to reset them when checking memory.
The random crng is coming from the kernel, not the tool. It just appeared in the middle of your output but is unrelated.
‘dmesg -n 1’ will disable these messages if dmesg is available in that initrd
I have no words. I'm really amazed by the tool, your soldering skills once again, and the final result. Congratulations!
Your amazing soldering skills made it look easy!
Oh yeah, Max is a wizard in any kind of retro tech, he has in-depth analysis of old hardware and such.
With this deep understanding of Digital Circuitry and Computer Programming he can rule the world. 😆
Great video, as always (and I really mean it)! I like your new film-making approach, it feels more natural and easy going. I really learned a lot from you about repairs and you are one of the reasons why I have retro bug now and why I now have PCs I wanted as a kid or teenager. I now regularly repair all my stuff I get from eBay. Thank you very, very much! Greetings from Montenegro!
Thank you! I hope you enjoy your newfound hobby ☺️
So many seems to be impressed by your soldering skills, I say fuck that, soldering is easy, what's really impressive is your patience and fine motor skills when dealing with bent legs on various chips.
Keep up the good work!
Haha, thanks! It is challenging to work on those small components. The microscope helps a lot, but I am thankful that my motor skills are quite good! Otherwise, this would be a totally different experience. Thanks again! 🙏
I am so happy you got this card running. This was one heck of a project, but you did an amazing job!
And you make it seem like changing the 3DFX Chip was easy :D
Changing the 3Dfx chip wasn't that difficult to be honest. I think the most difficult part is to lift the chip without bending the pins. I should get one of those suction pens. A little bit of low melt solder might help a lot to remove those large chips.
The fact that Witchery! picked up exactly WHERE the problem was occurring on your voodoo 2 card is AMAZING!
The saddest takeaway from your video: I SHOULD'VE NEVER SOLD MY VOODOO 5. Waaaaa!!! 😭😭
Fantastic progress, and effort!🙃😊🙃
Thank you
Geil, schön den freien Tag mit 3DFX-Reperatur beginnen.
When PCI (and other) devices are listed as "Vendor 8086", then it's always Intel. Makes it easier to filter out devices when looking for a specific one (or better understand missing drivers in Windows Device Manager). It's a matter of "It's obvious once you know it"
He just forgot to add pcidb file to rootfs image, it was meant to be as in lspci output (and it is fixed now AFAIK)
Nice to see it working again, got to love the retro community 👍
Yes indeed!
Thanks, really enjoyed watching!
Software showed no RAM, 'cause after de-soldering the two RAM chips the bits that are *always* used for data transfers didn't work. It's like having a road too narrow for a car: no matter how long the road is: the car can go zero length of it. The witchery software does testing on way lower level, so it reported specific pins that didn't work (impressive). After this video at least I know that RAM and DAC have separate buses :)
A tip: diode mode can be used to validate trace continuity between chips (3Dfx - RAM, 3Dfx - DAC, etc.), it only shows clamp diodes (google "CMOS clamp diodes") between chips leads and supply rails.
One note. DAC and RAM have separate buses but not exactly. Actually DAC has three (3) buses. Address bus (3-bit) and data bus (8-bit) for setup process and data bus (16-bit) for actual pixel data for the display. The first and third ones are separate but data bus for setup has shared pins with FBI RAM. Those are intersected in a weird way. So you can not setup DAC when RAM refresh works and vice versa. That's why there is a bit in config space register for remapping some registers between normal workflow and DAC operations.
Great bit of detective work and a very nice result. It’s really nice to see yet another Voodoo card saved. I can see you put a lot of effort in to this
Thanks! I'm happy each time I can save a card like this - and I learned a lot too!
A-M-A-Z-I-N-G work Alex, congratulations on successfully giving this card a new lease of life! I'm so happy your dedication, hard work, perseverance and talent paid off in the end! It's a bummer that the FBI chip is beyond repair (...or is it? Maybe it will work with an external PU/PD resistor, like some suggested. Can't wait to see a video about it!). Either way, it still means that 2 TMUs and an FBI chip that were out of commission have made a triumphant return! You know what to do next: Go enjoy some Interstate '76´🚗🔫🚀 on it, you've more than earned it!
And let this be another lesson to us all not to cheap out on our PSUs!
I'm really looking forward to some MegaMonster™ content! I can't believe I am still playing the old classics that originally ran on these cards. Just last week I finished another replay of Vampire: The Masquerade - Redemption. 🧛♂😅
And of course many thanks to Макс for his awesome tool that takes a lot of the guesswork out of Voodoo repairs! I wish all your fellow countrymen were as kind and open-hearted as you, it gives me hope to see how our love for this old technology can bring us together in these dire times.
Haha, yes, I'll try Interstate 76. This Voodoo 2 was a great project. I learned a lot, transplanted a 3Dfx chip, and brought two Voodoo 2 cards back to life (the other one was from the scrapyard).
So, yes, what a great ride! Thanks again for the opportunity to use your card and make those videos!
@@bitsundbolts I really envy you for seeing Interstate '76 with fresh eyes. Let me know if it's all rose tinted glasses on my side or if it's indeed still enjoyable almost (...holy s***!) 30 years later. 😅
I'm the one who has to thank you for fixing the card! It's definitely yours at this point, but I'll be glad to buy/trade it back in the future, once you've exhaustively run it through its paces!
Nice! :). Now that is a tool that i need, i have the same Voodoo 2 but rev B that i would like to test, and also a Voodoo 1 Orchid.
Only Voodoo 2 cards are supported as far as I know. But still, it is a very impressive tool!
Congratulations! There was so much frustration, but it finally paid off. I enjoyed watching it.
Yay, now we finally know the FBI was damaged. The tool looks really nice and will save some of us a lot of time and effort for sure. Really looking forward to the release!
Вот уж чего не ожидал, так это увидеть тут канал Макса 😀!
congrats to both of you, BuB! awesome series, happy to see it working again. these 3 parts were way way better than the latest marvel, dc, disney movies combined :))
Haha, great to hear! Thanks for your time and happy you enjoyed those videos so much!
Excellent video. I watched all 3 and enjoyed the series. The repair videos are definitely my favorites.
Thank you! I enjoy those projects too. Unfortunately, I'm running out of broken Voodoo 1 and Voodoo 2 cards. Later models have BGA chips which I'm not yet confident to work on.
0.5 V is about one PN junction worth of voltage drop. So one diode junction, or one PN junction in a BJT or transistor.
Silicon PN junction. Other materials have different drops.
Awesome work! Awesome testing tool too! Super thorough!
I am impressed that you are so happy to have saved another component from the dump
Thank you! I wasn't planning on releasing three videos about this card, those projects are unpredictable! Nevertheless, when such a project comes to a successful end, I am overjoyed to have learned something and I can give back the card to the owner who is watching my videos eagerly awaiting what is going to happen to his Voodoo 2 from 25 years ago 😄.
Foi uma longa aventura recuperar essa placa de vídeo! Parabéns! Muito sucesso ao canal! Abraços do Brasil.
This tool will be very useful in future. Nice to see a another V2 is working :)
This is awesome. Happy to see you got it working!
It works!!!! Just amazing 🤩 so wonderful this tool exists!!
Nice work, good to see this card back to life!
You need some low temp solder for removing SMD chips. ChipQuik and the like. It makes the job much easier.
That is amazing. 👏
Thanks
Thanks for all these great videos! It would be cool and i think useful if he updated it to send the debugging output that goes to the console to also output on the computer's serial (COM) port. that way you could plug in a serial terminal and watch in real-time the output of all these tests as the main screen is flashing bars! In any case this is amazing and if my voodoo card ever has an issue, or if i just want to check this tool out, I look forward to testing it when he releases it. thanks!!!
shorts on data lines can be caused by burnt internal pull up or pull down resistors and in some cases you can suplement it externally either on the controller chip or on the memory module alone if pin is "dangling"
i would assume voodoo has pulldowns since it doesn't work on chip burnt by overvoltage
so if you didn't want to replace the 3dfx chip you could bridge pin 2 with resistor ~100-500k (i think you could check it on any working chip to ground or power if i am wrong) and do a getto fix
if it is just a pull up on the bus it would work if not it would be just swaping the controller like you did but i don't advice to put bad chip back to test my theory :P
Well, I still have the chip and the donor board. You're the second person suggesting this fix. So, you mean I just place a 100K resistor between 3.3V and the broken D0 (Pin 2) line in question? I could try that on the donor board.
@@bitsundbolts if you had an oscilloscope you cuould mesure logic level voltages on memory or just get the value from memory datasheet if it is 5v chip you might have 5v logic levels, if 3,3v then 3,3v
the biggest thing is to figure out if memory is shorted to ground, dangling or is there interna pullup without pull down or vice versa . if the damage is just on pull ups it is as simple as bridging data pin with a resistor to whatever voltage it should be pulling to, if it is internal damage on decoding logic or the output of a logic gate in a buffer or something, this would not fix or break anything
it was just a silly idea about how to potentially getto fix a broken chip kind of zombie voodoo
Well, anything might be possible with those Voodoo cards 😉. I might try to revive that donor card. I just need to get my hands on the 3dfx chips and could try the Brocken chip once more. I'm just a bit worried due to the high frequency. I'm facing similar issues on my 16MB SIMM modules with the level shifters.
'crng init done' is a Linux thing. Just happened to be mixed together with Witchery output.
Excellent series, and @fagear's Witchery is clearly a power tool. Great work.
I mean "powerful" tool. If I edit my comment I'll loose the coveted heart!
I guess it's both! A power tool and a powerful tool!
well done, and well done on the debug tool, awesome works
Great job !!!, You put lot of effort to work this card 👍🏻
OK you did an awesome job getting this working!!! Witchery is awesome too!
Thank you! I absolutely agree! Witchery is an amazing tool and I hope it will evolve into something great!
nice work - hope I can work on mine like this at some point
Super nice tool, especially the clear information on where the chips are and what might be the problem. Did Maxim create all kinds of errors on his test rig to know the errors or is he just a god among mortals.
I agree! The information and error description of that tool are amazing! I think he did simulate errors by purposefully modifying Voodoo 2 cards. It's a slow process and deserves a huge shout-out to Maksim's work.
Multiple hours were spent buzzing out connections, stepping through adresses in software while probing some pins, shorting stuff on working cards to see how it will affect results. And then I've bought a dozen of broken V2s and started to run tests on those.
This tool is absolutely amazing
Really beautiful work!
What a great series of videos.. thanks for sharing.. cheers.
Thank you! Glad you enjoyed those videos!
Great tool, it makes debugging way way easyer. Instead testing for long hours for shorts, continuity, open traces or bad solder contacts on each pin, this tools in 10 minutes gives you the needed info about what is not right. Who would suspect that the FBI chip have one data bit stuck low? You could have find the problem but it would took much more time and more than a DMM.
Correct! And I couldn't spot anything with the multimeter. I tested so many pins on the memory chips. Diode mode, continuity mode, etc. I wasn't able to figure it out. In the end, I would have to guess and I would have found it. The tool gives you the right hints and will save you time!
@@bitsundbolts In the end you would had to use an oscilloscope, logic probe, logic analyzer to see in real time what happens with the adress/data bus on the memory chips or FBI. Tools that most mortals do not have and are expensive and hard to get (good ones, I mean).
I wonder if this tool will be free and what will be able to do in its final revisions, maybe full test any Voodoo card or I am expecting too much? I mean, and you also said it, most of the documentation for those chips is not available, so is voodoo magic...
@sebastian19745 Unfortunately, the tool supports only Voodoo 2 cards for now - I might have said Voodoo 1 as well in the video. That was a mistake then. The tool will be open source and hopefully, with more people contributing, Witchery will support all Voodoo cards. But it will take time as you said - it's all voodoo and witchery ☺️
Great work on this repair! I love your channel and your work :D I notice you use an OSSC as part of your video capture setup, what capture card you use for the hdmi output?
Thanks! For HDMI capture, I use a very cheap USB capture card I found on Amazon. I had to go through many before I found one that works well and responds quickly to resolution changes. I had an Elgato Cam link 4k as well, but surprisingly, that somehow didn't work well (especially with certain resolutions under DOS). The USB capture device was around 15 USD and is silver. Unfortunately, there is no guarantee that it's still the same device. I'll add a post to my website about it in the coming days.
@@bitsundbolts great, thank you!
Awesome job! That really is a useful tool for diagnosing problems. Sad that it was the FBI chip, but not unexpected. I'm surprised it even worked that much! So thanks to Maksim for helping save this Voodoo 2!
Anyway, you have far more patience than I do for straightening pins, and a much steadier hand! I definitely would've put it in the too hard basket once I worked out the memory was fine. At least if I come across a board with a chip that has mangled pins (or a chip with mangled pins), I could donate it to you so I can just watch you fix it instead! 🤣
Thank you! Ha, I hope you will have plenty of Voodoo 1/2 cards - definitely keep them if you find them and I can try my luck on any of them. The ones that won't be recoverable stay with me for spare parts, the working cards can go back to you :)
@@bitsundbolts Deal! Except I've already got enough Voodoo cards, so I'd probably just let you keep them!
Great job! I guess the burned chip damaged the FBI chip and made it report bad memory, no matter what chips you replaced. One more Voodoo card lives on!
This encourages me to pick up a multimeter and test out a Sapphire Radon 9250 card I've got that isn't detected by the AGP slot, I've got 2 so I'll be able to compare values between the good and the bad card.
If you can, check the BIOS chip of the card. It might suffer from bit rot.
@@bitsundbolts Unfortunately I don't know how to check it. In fact, I don't even know which one is the BIOS chip 🤔 I might try googling the inscriptions to see if I find it, then try to swap them. But before the soldering iron gets involved, I'll try to measure each component's value and compare to the good card.
Maybe there is a flashing tool you could use to see if it detects the BIOS chip. I haven't done this, so I have no experience.
Use an old PCI graphics card if you have and keep the Radeon 9250 in the system. Try to flash the BIOS. It might work. I have two videos on my channel where I flash once a GeForce 2MX and once a Voodoo 3 using DOS flashing utilities. Both cards suffered from a corrupted BIOS. After flashing, both cards started to work again.
@@bitsundbolts Actually I've done that already, I've used a PCI card to boot from a USB floppy drive, and used FLASHROM.EXE for ATI cards but it doesn't detect it. I've used this utility to flash Radeon 92xx cards with modified firmware and turn them into ArcadeVGA cards which can connect directly to a RGB SCART TV at 15khz, when something does wrong I just insert a PCI card and flash the original firmware back. It worked successfully with a MSI card and a Sapphire card, but this one (which is EXACTLY identical to another working Sapphire card) isn't even seen by the system, it just ignores it and either boots directly into the onboard graphics or just beeps if the board doesn't have integrated graphics. Tonight I'll try to do a continuity test on some of the components and compare between the good and the bad cards to see if I can find any difference.
Awesome! Where can I get this Tool?
Nice tool and cool i told you it be the memory. But now i might be the soldering 😊 hope you solve it :) love the channel
Well done! I am impressed with your effort and detail level! Congratulations on a success!
(regarding the crng inthe outpout I think it is the Linux Random number Generator sending messages to stdout)
Thank you! Yes, you are correct. I was told the crng line is generated by the Linux kernel.
Far out, I haven't seen linux booted via loadlin in *years*. Decades even
Last time I used loadlin was with Slackware 3.6 circa 1998 :)
Zipslack!
IT WORKS !!! WHAT A RELIEF
Amazing work!
Success! Nicely done!
PS I Wonder if inatalling faster memory chips would enable you to OC the mem?
I haven't done any overclocking on Voodoo 2s yet... There are models with 110 MHz memory, but those memory chips are hard to come by. Not sure if it's worth it. And if I'm not mistaken, the Voodoo 2 chips are running at 90Mhz - so, there should be some headroom for overclocking anyways. I'll try that maybe on one of my own cards in the future.
@@bitsundbolts I remember pushing the old Voodoo1 chips from stock 50 to 62 Mhz easily. Even Powerstip was displaying a flame icon next to the speed slider. Looking back, I wonder how many ns were those chips. Anyway, thanks for the video, I'll probably try Witchery on the ol' V1 just to check the tool when it is released. Were it not for your video I would probably never hear about a new tool for my 28yr old card, heh.
@Thomsonicus I'm afraid Voodoo 1 cards aren't supported yet. I may have said it wrong in the video. But I am sure support can be added quickly after the tool has been released. I never got any Voodoo 1 stable past 57 MHz. I had 35ns memory chips on those cards. I guess you were lucky with your model!
Amazing job !
if this ends up working w/ Voodoo 3's i might be able to eventually find out what's going on w/ this Gateway branded Voodoo 3 1000 AGP i have on hand. i'll haveto keep an eye out for when the Tool's Released.
At the moment utility works only with V2. In theory V1 is pretty similar so it probably will be supported later. For V3 I briefly looked at the datasheets and I don't know yet if its support will be easy. No promises on that yet.
The crng is a Linux kernel message. You can identify them because they are prefixed with a timestamp. You can ignore it...
wow what seriously handy debug tool - i so need this - where and when can i download it
It will be released as open source as per Maksim. He's still working on it, but maybe, with all the positive comments, it may be released sooner than later.
@@bitsundbolts at least it checks the FBI ram chips at the moment - to exclude those 8 chips if your trouble shooting a artifacting/broken card - it will be great when it can check the TMU ram chips aswell - better to just remove one bad chip than the whole 24 ha ha - BTW i knew urs would need that replacement FBI chip - as it was directly connected to the blown ram - but what a great tool to prove it first
Amazing video !
I am curious what the pin looks like underneath the ceramic now. If the damage isn't too deep, you could maybe dig it out and re-solder it internally lol
I am afraid that the damage is inside the silicon chip. There is nothing I can do :(
The main lesson here is to go back in time and make sure people don't buy cheap PSUs. Had a bad PSU in 2005 go pop and it charred a pin on the 24 pin connector but luckily it didn't burn my P4 Northwood and Radeon 9800 Pro , was a PSU that came with a case and I thought i had gotten a deal, turns out that any time you get a PSU along with something, it's likely sketch.
Very true! I remember those cases with PSUs. And you are right, most of them were garbage and you shouldn't use them if you liked your PC.
The PSU and the cooling are the key components i never compromise on, even on cheap systems, it's just not worth it endangering an entire system to save a few bucks/dollars/euros.
Brand name or nothing for that, people who don't like that can buy their computers at the food discounter and leave me alone with it.
Cool! I really need that tool
I am sure Maksim is going to release it at some point. I will probably mention it when it is public.
Good job. One or two more video of this kind and you can start a Netflix show 😂
Told you the FBI chip. ;) My guess is that the protection diode on that FBI chip data line shorted to VDD from the overvoltage.
Yes, there were a few suggesting that the FBI chip is broken. That is one of the chips that is hardest to replace and I only did it after there wasn't any other choice. It's very unfortunate since it seems to be just one of the lines that killed the chip. But it was coming from that one memory chip in the bottom right corner that had that big hole burnt in the center.
Excellent! Maybe you can bridge the two Voodoo 2 cards and run 1024 x 768. You do have a bridge, right?
I need to make one. I think I have everything I need. I'll try to make a cable today.
Nice Work!
i do wonder what can you use to clean up very stubborn solder paste ? i 'v tryed isopropanol but often its just dilute the paste and make a more of a mess.
I could use hot water with soap and a tooth brush but thats add some other inconveniences, and even then it isnt cleaning everything.
Any tips for any chemicals?
Also very interesting and good video, i could almost feel the anxiety passing trou :-P but big reward in the end.
Unfortunately, I don't know how you could deal with your solder paste. Maybe try to heat it up with a hot air station to 60-80 degrees (so the solder doesn't melt, but the flux will lower its viscosity), and then try again with a cloth dipped in IPA. I assume you have those tiny solder balls everywhere on the board?
@@bitsundbolts my hobby involve crafting electronic dohicky or prototyping so tiny is debatable, its not as small as what you working with, but my issue is when i need to melt and fuse solder with some stubborn metals contacts that work nicely with paste, but paste is such a mess and sticky. mostly i use clear pcb liquid solder flux, but i don't want to use all of it and im looking for alternatives, like paste .
I will try the heat and clean . or maybe i should get a bottle of Acetone.
46:29 that looks like a linux kernel message, crng is a cryptographic number generator, it seems like the init was quite late so the kernel just spat that message out in the middle of the console,
imo unless the test uses crng it's unrelated to the test and nothing to worry about
Yes, confirmed. This message is not related to the tool. As you said, it's a kernel message that popped up.
It wouldn't change much, but I wondered why haven't you checked the continuity between suggested FBI pin 64 and pin 2 of U6 & U21? Clearly, the databuses of these chips are in parallel. The 'crng init done' with a timestamp during the last Witchery run, was just a Linux kernel debug message (you can see a lot of them during Linux startup).
Oh, I did check continuity multiple times. It's not the fault of the PCB, pads, or traces. The continuity is there. The fault is within the FBI chip.
@@bitsundbolts yeah, I know, but you just haven't shown that. On the video you've only shown voltage drop from these points to GND.
With the new nanometer this chip can be remanufactured and used. I think it was a very strong screen layer and was invincible at the time.
That donor card deserves to be framed as saving of X and Y cards.
Maybe in the future there will be a process to can open that FBI chip and repair the die issue with that memory channel.
Haha, there are already suggestions on what I can try to save the FBI chip. But now, I am also considering saving the donor card - I just need a few NOS 3Dfx chips.
I did suspect the FBI chip was the problem, this is a bit of a sad repair as the broken chip is one less of a finite supply rather than capacitors or other simple components that you can get brand new.
But I still loved that "YES" moment when it worked!
Thank you for another journey, and thank you to Maxim? / fagear for a nice tool.
He is using witchery, burn him! 😁
Fascinating Video, as always! Thank you very much!
Nice job! You shouldn’t be that much surprised by the 0MB without chips or with faulty U6 access. The FBI chip uses 64-bit bus to access memory so you need 4 working chips to get the first 2 MB of FB memory. In this case I guess it failed memory test on first byte so effectively crashed the system and reported 0 MB.
I wonder (but maybe I missed it in previous episodes) how scope view of faulty bit looked like? If the chip was fried it should never reach out L or H and be a bit different from other bits.
I did not scope every data pin. It is in an unfortunate location and difficult to reach with the probe. I do have the Brocken FBI chip though - I could resolder it to the donor card and test it with a scope now that I know where the issue is. I can just test one of the corner pins of the FBI chip.
Man, I wish I had this tool. I have an even stranger Voodoo 2 (also with an FBI issue) on which I replaced all memory chips (with new ones) and also the FBI itself with a NOS chip and the problem still persists... And my card looks like new, no broken traces, no missing components (it sat on a shelf for 2 decades, in an antistatic bag).
Edit: btw, "Random: crng init done" is output from the Linux kernel (Cryptographic Random Number Generator initialization) which is displayed on the console output with a certain delay. It's not related to Witchery. 🙂
Maksim will probably release his tool at some point. Hope you can wait for it to be available! And thanks for watching!
tool for voodoo diag is not public / free? Is usable also for other 3dfx cards? 🤔 good work voodoo working again 🥰 I am waiting atm for two V2 12MB so i hope i will try SLI mode 🤗
The tool is not yet public, but it will be released as open source some time in the future.
well done !!
I'd try to scope on the pin 2 to see if it's just kinda low and has something pulling in down but can't get it high enough to register
Gz Sir super rep :)
Any news on a possible publication of witchery? Even an unfinished version.
I'm not sure if there is a set publication date yet. However, I have seen first tests on Voodoo 1 cards.
@@bitsundbolts great, thank you for taking the time to reply :)
I wonder if it's just the chip select line for that bank of 2 ram chips. If that's all it is, maybe a little bodge of some logic chips to make up the missing signal based on the other 3 would work. Just a thought. Ping me if you want to try something like that
Awsome very nice job
Thank you!
@@bitsundbolts if I were to send you a 8mb card would you look at it and upgrade the memory ?
Hey, I absolutely would. However, the shipping cost would make this endeavor very costly if I'm not mistaken. I believe it would come close to just buying a 12MB card locally. It only would make sense if the card travels with someone in their vacation luggage I'm afraid.
Amazing! :D
Look at that! People love this series "Ride of the Voodoo Valkyries"! 3h and quite so many comments. Amazing!
I'm very happy that this Voodoo 2 repair is interesting to so many people.
🔈 *turn your sound up*
Is it just me or is there no download link for Witchery available? Would love to use that tool!
Witchery isn't available yet. The author is releasing the tool once he's comfortable it works properly.
В принципе я давно подписан на Максима. Не знал , что у него есть такой софт. Это круто.
P.S. by the way my version of issue eas right.)
I suspect certain memory is more important than other memory. Maybe that's why 0mb shows up even on the new chip.
This is what Maksim explained to me as well. U6/U21 seems to be the first MB, If this isn't there, 0MB are shown. For me, one chip (U6) wasn't working, so I got to 1MB, but no further because there was an issue.
@@bitsundboltssome systems require a certain amount of ram before the rest becomes available. I don't know about the voodoo
Good job as usual but it looks like you put a lot of time into this Voodoo 2 card.
Thank you! Yes, I did put quite a bit of time into this Voodoo 2 card. But in the end, the card is saved, I learned a lot, and the owner gets his card from his childhood back!
Maybe you already did this and I just missed or forgot (I'm very forgetful), but did you check the pins of the FBI chip for loose pins before replacing?
Yes, I did. I reflowed the solder of all pins in part 2, of this series.
@@bitsundbolts Ah, ok. I was thinking of going back and checking, but I didn't feel like skipping through previous parts. I know I watched them as I gave them a like, I just forgot already.
As I'm say on previous video - you must check data lines on memory chips by oscilloscope! I'm think you will be found this issue))
I only tested the traces with a multimeter. Some of the data pins are difficult to access with a multimeter. Especially the one in question is right next to the 5V supply pin. I still have the 3Dfx chip and I could solder it to the donor board if I ever attempt to fix that one too. Then I could try and see if it would be possible to spot the issue with an oscilloscope. And now I know the pin in question is at a corner of the FBI chip.
@@bitsundbolts some broken chips give very low or distorted signal. It can happen that you need just pull up resistor to 5V and it will work again ))
I've seen this suggestion! That would be awesome if it would work. I just wonder what resistor value to use. The capacitance of the memory is about 5pf. Not sure about the 3dfx chip. I need to get a value that works at the frequencies of the data signal. I guess I could experiment. Start with a large value (e.g. 10K) and work my way down.
@@bitsundbolts You can watch signal on oscilloscope and add resistor 10k-1k-100 Om, until signal become good enough...
2 questions:
first: why would you try to repair a so badly burned card instead of making it the donor and repair the previous donor it couldn't be more broken could it?
second: would the FBI chip work with an external pull up on that pin? Its a pity you did not test that now its to late, or may be you could test the on the donor?
The donor card I have was handed down to me and it had many parts removed or replaced. Additionally, the spot for TMU0 has a few broken pads. So, the PCB of the card with the exploded memory chips has a PCB in better condition as well as all the original SMD components.
And I think reviving such a card is a greater achievement than reviving the donor card.
To answer your second question, I am planning on reviving the donor card once I get an FBI and TMU chip. I will try the external pull-up resistor. Unfortunately, I have to solder the FBI back to the board, but I'm willing to do that. It would be cool if this TMU could be saved by an external pull-up.
Why didn't I test it while I had the chance? I didn't know or it didn't cross my mind to do this. I'm not that experienced with electronics repair, but each time I learn something else thanks to my audience!
Curious to know if you’ve heard from the Witchery developer? :-) Is there anyway to get access to a beta version? Got so many Voodoo cards to repair.
I'm in touch with him. There is a lot of development going on including support for more 3Dfx models (Voodoo 1 specifically).
@@bitsundbolts Wow that is fantastic! I smell a follow up video coming down the pipeline ;-)
💯
Because there was no short on the D0 of the FBI you could have tried a questionable repair by adding a pull-up resistor from the 3.3V line.
Hm, interesting... I still have the chip - and a donor board... and a pullup resistor... But could you elaborate on your "questionable fix"? I assume it is just a shot in the sky to add a pull-up?
@@bitsundbolts Well, by not having a short on that pin we can assume that the pin protection diodes that are connected between the pin and the VCC and GND respectively are not shorted. The only thing that is wrong is probably internal mosfet of that output's gate that turns that pin ON (connects it to VCC). So if the other mosfet is OK that means that this output can in fact pull down the pin but there is no active device to turn it ON (pull up). So by adding a resistor we can provide that pull up however we don't know if the rise/fall time will be adequate (I once managed to make this work but not at 90MHz for sure) or we don't know if the output is sometimes put in it's third state (high impedance) so this repair attempt will be quastionable at least.
Is this tool also works for Voodoo5 cards?
Ohh I hope Maxim will make a tool for Voodoo5. Please mention to him.
Soo badly needed! Especially that there is not even a mojo for it.
Thanks.
I am sure he gets this a request to support every Voodoo card out there, but it must be incredibly difficult to support them since there is very little documentation for some of those cards. If he releases the source code, I am sure the community will slowly build on it and create support for more cards over time. We just need to be a bit more patient.
i dont know why. maby it's just the mood i'm in but i heard the typing and was like ... nope not working then slap slap slap.. Well wrong keyboard :p
lauging my but odff.
Thanks for the blooper.
Haha, yes... Sometimes, I'm impatient. No USB support in that Linux distro.
I also repaired broken Voodoo 1 by fixing 2 FBI chip traces and some pins. Nice work. I prefered to not use hot air, becasue the pads can fall off.
Whole restoration video is on my channel.