VBIOS Bit Rot? Saving a Pair of Matrox AGP Cards

[@philscomputerlab]

Good to know! While it hasn't happened to any of my Matrox cards, I will remember this just in case I do encounter a similar issue 🙂

[@marvintpandroid2213]

Bit rot in the diodes all down my left hand side.

[@xenon2Merchant]

Quite interesting. Seems that just for safety reasons we all should dump bioses from existing cards and save it for the future.

[@jonchapman6821]

I have a feeling that I’ll be revisiting this video in the future.

[@brainiac9579]

Thanks for making a video about this. I've encountered this, "bit rot" or corrupted vBIOSes, on AMD RX4XX and RX5XX cards, but wasn't aware it was a thing on Matrox cards.

[@mike306dt]

My G400 exhibited the same behaviour and a reflash fixed that one too! Another great video. Thanks :)

[@nicholsliwilson]

Mate! You’ve saved 2 of my G200’s from the e-waste bin so I can use them in retro projects. Thank you.

[@vswitchzero]

Fantastic. Your comment made my day! :)

[@stevec00ps]

Ah Matrox Files by gbm, I remember that name from the Matrox Users Resource Centre (MURC) forums! I don't remember this issue and I've been an avid Matrox fan boy for decades! Nice video :) (edit: in fact the MURC forums are still going so I just posted this on there!)

[@esc2dos]

I've got a couple of video cards that look good physically but don't work, I think I will try the 2 cards at once trick, really appreciate this.

[@ruthlessadmin]

Very cool. Never would have thought of this. Just in time, too, because soon I'll be starting the long and arduous process of testing (and repairing/refurbishing as needed) my entire collection. Cards like this would have gone straight into the spare parts bin without a second thought!

[@lordwiadro83]

Great knowledge - many thanks! The Matrox G200A is one of my favorite cards. I own 3 of them, one being an LE "slim" version. Luckily, they all still work.

[@PixelPipes]

Wow I had no idea about this issue. I haven't run into it (yet) with my G200, but it's good to have this video to refer back to should it happen to mine.

[@mikatorkkeli4932]

Company i worked that time had all matrox desktops and yes i had few diskettes with the recovery bios with me, it was so common that you needed couple extra disks to fix them all. Company had some 5000 desktops at the time :)

[@vswitchzero]

Thanks for sharing! Very interesting to hear historical perspective like this 👍

[@philscomputerlab]

I recently bought a Mystique 220 for a low price. No refunds and untested. The machine did POST but no image on the screen. I followed your link and found a reference to SETUP349 containing various BIOS files. But the update program would not run. I checked the commands and it worked with PROGBIOS -k -i auto. -k forces the flash. Shutdown machine, remove AGP Nvidia card and now the Mystique is working! I ran Final Reality and oh my, that card seems worse than S3 Virge??

[@vswitchzero]

Awesome! Really glad the video was helpful to get your Mystique 220 working again. This issue seems a lot more common with the early AGP Matrox cards (G100/G200 etc) but I've seen it happen with others too - including 3dfx V3s. Can't say I've ever really done much with Matrox cards as far as 3D acceleration is concerned. They've always been great 2D cards to pair with a Voodoo 2 though :-) .. I did get a G400 Max recently though. Looking forward to trying that one out.

[@philscomputerlab]

​@@vswitchzeroThanks I'm stoked didn't think it would work. G400 is nice card, with Direct3D actually usable. OpenGL not so much. Clear sharp image and should work with most games. Performance also not too bad.

[@jamnongko6285]

Thanks for sharing. My g200 is working again after sleeping a long time. From 🇹🇭

[@vswitchzero]

Nice! Glad it helped 😁👍

[@Raul_Gajadhar]

This is a very useful piece of info. Thanks for sharing that with us.

[@azbesthu]

Around 2005 I also got a g200 when I searched for a card to have a second monitor connected. So I already had a pci card there and win recognized it, flashing vbios saved it :) For some time I had a G200 agp + mill2 pci in my machine. I still have those :)

[@jikissgamer]

Very informative and helpful video! I remember something similar happening to those Voodoo 3 cards you repaired a while back. I have a few matrox and Voodoo 3 cards and will definitely refer to these videos if I ever need to flash their bioses.

[@Mq333]

Thank you very much! I just followed your instructions, and I fixed my old Matrox Millenium G400! Thanks!

[@vswitchzero]

Awesome! Great to hear! :)

[@Mq333]

@@vswitchzero Thanks. Greetings from Poland!

[@fft2020]

awesome article ! I have here 2 matrox cards with the same symptoms. I will give it a try!

[@DropDeadFrederick]

That happened to a G400 I owned back in the day that I had bought directly from Matrox. Barely had it for a month and it died like that so i sent it back to them under warranty. The replacement worked fine until I eventually retired it. My case was similar in that you couldn't see anything that used a DOS screen but once it got into windows and the driver took over everything worked normally. Up until now I had no idea that this was considered a common issue with these cards and I always assumed I just had gotten a defective unit. If it really is that common my hunch is that it's either something specific with the EEPROM used or there's a defect in the programming circuit of the cards that's causing it to apply a voltage to the chip high enough that makes it think it's being programmed even when it's not intended. That's just a guess on my part. Only way to find out is if a Matrox employee chimes in or you probe one of the cards directly.

[@jrodstech]

Oh ya! Just found this channel. I'm going to love this for sure.

[@pc-sound-legacy]

Interesting! I should check my old Matroxes, maybe they need a bios flash too

[@SobieRobie]

Thanks for the tip!

[@cybercat1531]

This is also an issue of sorts with the Intel P28F001BX 1mbit eeprom chips used on several motherboards. In particular these chips used at the time cheap NAND flash which is what gives them their fast programming speed. NAND = low cost, high density, high-speed program/erase, slow read NOR = high cost, low density, low speed program/erase, fast read NAND if not rewritten or powered up once in a while has a tendency to slowly "fade" to it's original un-programmed state. So these boards if not powered up or reprogrammed in a while (approx 10-15 years) end up with a corrupted BIOS rom. At least here a fix is as simple as just erasing and rewriting the chip. But a swap to the more reliable NOR flash winbond or ST chips is also good for absolute peace of mind. As for the matrox cards I believe there is a similar issue at play and it's ultimately worth replacing the eeprom chips on them with a modern more reliable type.

[@Xitrial]

I owned a Matrox Mystique, quite nostalgic

[@Alcochaser]

A lot of EPROMs from this era have this issue. I wouldn’t blame matrox as they used an off the shelf part that should have been good. One thing Matrox pioneered before it became mainstream was multi monitor support. If you look on the G200 it’s got an unpopulated second VGA port. This was for what they called dualhead. Later everyone started doing this, but Matrox was first. Matrox still makes cards, but I think they are built on modded AMD silicon

[@vswitchzero]

Totally agree. That’s a very good point and I wish I had included that in the video. It’s definitely not Matrox’s fault and in fact, they are probably the only manufacturer that at least tried to do something to avoid this with the bios guard utility. I’ve always had a soft spot for Matrox cards and loved that they’re Canadian too 🙂

[@pentiummmx2294]

yeah, woke up to my 486 board greeting me with Award Boot Block BIOS Checksum Error screen, at least i downloaded a bios that supports that board, on to the TL866-II and to flash a new BIOS to it and it's alive again

[@_derSammler]

I think the message in the end saying it was flahed 3 times is the total counter of re-flashes, not that it did run the process 3 times.

[@FoxMulder78]

Exactly, it shows the card has been programmed 2 times before, I bought these Matrox cards and all of them have been flashed before at least 1 time, a log count of flashes is kept on the EEPROM.

[@vswitchzero]

Thanks for sharing. That definitely makes more sense 👍

[@sapperlott]

The note about the card having been flashed 3 times doesn't mean that the utility flashed it 3 times in a row. The PINS data structure (which holds card specific info like the card serial number etc) in the card EEPROM contains a field that records how many times the card has been flashed in total. Said PINS data structure also is the reason why it's a good idea to create a recovery disk using the Matrox utilities. If you get bit rot in there, there's no way to restore the card to its original state since that data structure is not part of the VBIOS update files (of course you can flash someone else's PINS data of the exact same model but then it'll still be different from the original contents). Another major headache are the Matrox cards that IBM re-badged for use in its RS/6000 servers / workstations. These cards (GXT120P aka Mystique 220, GXT130P aka G200 PCI and GXT135P aka G450 PCI) got re-flashed with a custom VBIOS by IBM (since Open Firmware expects FCode instead of x86 code in the PCI expansion ROM), so there are no other sources for the VBIOS to re-flash them (IBM never released VBIOS updates for these). So with one of these it's extremely important to create backups ASAP. Just now I dumped the VBIOS of two GXT130Ps and both had severe bit rot. Thankfully someone put a VBIOS dump of their card online, so I could do a three-way compare to figure out what the original content should have been. Dumping the VBIOS is reasonably easy in Linux (as root): - check the PCI bus ID of your card using "lspci" - cd /sys/bus/pci/devices/ - echo 1 > rom - dd if=rom of= Out of curiosity: What SOIC8 EEPROM was used on your G200? I've seen Atmel and STMicroelectronics chips being used. Both my problematic cards had the STMicroelectronics ones.

[@burntoutelectronics]

It wasn’t long ago that I’d never heard of the degradation of stored bits in EEPROMs and the like. One just assumes that it wouldn’t fail on its own seeing as it’s not being actively written to when in use. It would be interesting if background radiation, especially beta particles, would be a partial cause for this effect.

[@TomStorey96]

It's not really a matter of being written to, it's how the bits are stored in the EEPROM. They are stored as a charge (or lack of) within an insulated/floating gate of a transistor. Retention of the data depends on how well the gates hold their state, and usually they are rated for 100 years or so. But if they happen to be excessively leaky, perhaps as a result of a manufacturing defect, then the charge will leak away over a much shorter time. Reprogramming restores the contents but the faulty construction still exists so it will only happen again in the future, hence there was that tool to periodically reprogram the EEPROM. But as long as the EEPROM isn't internal to the video chip itself (fingers crossed it's just an external SPI or I2C device), it would be very much worth replacing it, and then it should live for another 100+ years without fault.

[@38911bytefree]

Eeeprom and fash data time retention is ensured to be at least 10 years but nothing else ... any datasheet will point this. They are not mask roms.... if you dont rewrite them .... they eventually will die. This applies to all EEPROM and flash and also flash based micrcontrollers. This is a good warning for thos with with very speicific or collectible gear where no .bins are available. You better read your EEEPROMS and FLASH while still working and save to files.

[@tiemanowo]

I just wonder. Maybe soldering more modern flash chip will prevent degradation?

[@drzeissler]

That is interesting....I will check my dead G200 PCI !

[@nexxusty]

"BIOSGuard"? My God that's ridiculous.

[@GizmoTheGreen]

had a dumpser dive PC that didn't boot couple of weeks ago. totally dead, no response to pwr pins. flashed latest BIOS from online, with a usb flasher, and it booted perfectly after that. so either person that threw it failed a bios upgrade (10+ year old pc, so doubt that's the case), or it bitrotted.

[@jamezxh]

By any chance do you have a parhelia? Would love to see a review of one of those

[@vswitchzero]

Definitely on my list of cards to acquire! I actually bought one on eBay a few weeks ago but the seller cancelled my order a few days later for some reason. Still on the lookout for a decently priced one to add to the collection 👍

[@briangoldberg4439]

that's interesting. i wonder however if it wouldn't be better to replace the EEPROM chip with something newer in order to future-proof it?

[@vswitchzero]

I wonder the same. I may look into this at some point. Thanks for the comment!

[@Druid_Plow]

I remember buying my first GPU. It was an ATI Radeon 9550 AGP. Halo, quake, doom, half life, "HD" movies. Lol

[@Igbf]

Out of curiosity, did you consider the possibility that the message "the bios has been flashed 3 times" is actually a total counter? I mean, that it shows 4 on the next time you flash it. It would be less weird for sure.

[@vswitchzero]

Thanks for your comment! Yes, not long after posting the video I realized that I was probably wrong. I definitely think that makes more sense. Especially since it knows the date of the last flash too. They must store that information somewhere else on the card. Cheers!

[@ljrretropcs]

I like this video. But also dislike it because I chucked a whole box of "dead" matrox cards last year :(

[@matthewday7565]

Now I'll have to find my non-functioning G200 and see if it can be revived

[@matthewday7565]

Very much later, plugged my card in alongside another.... Windows detects!, Now to prep those tools!

[@spy.re-drobe20]

In term's of (d-ram_yield) how are the d-rams going to "properly" be (manufactured), if there is a producer of these chip's, it would be intriguing.

[@spy.re-drobe20]

Method needs re#ing, defining Such method's can be pretty tricky. (Ayah....)

[@RootiferasRetroGameplay]

I recently had something very similar with an NVIDIA Vanta, I thought it was the capacitors but that didn't help. I wonder if this is an option for nvidia cards.

[@omega552003]

@7:16 I cant find that track anywhere...

[@vswitchzero]

It’s an original track composed for the channel by Bertrand Guegan. It’s done using an old school adlib tracker. You can hear the full version of the track at the end of my Pentium Overdrive video. There is a link to his Sound Cloud page in the description as well.

[@omega552003]

@@vswitchzero fantastic and thank you for the information.

[@SidneyCritic]

The other card looks like it has the footprint for the extra RAM socket - lol -.

[@vswitchzero]

Yep! Looks like they just use the same PCB for both cards to save costs :)

[@retrogamestudios6688]

Confucius say.....man with soldering iron is cousin to the cow brander .....good work though boss

[@Hammersch]

Hi! 5:05 why does it say 15bit before you flash? then after 16bit?

[@vswitchzero]

The Matrox driver for Windows 2000 supports 15bit and 24bit colour options, which is pretty uncommon. I switched it to 16bit earlier and it just switched it back for some reason 🙂

[@Hammersch]

@@vswitchzero Okej thanks for replaying yes very uncommon :)

[@ProDigit80]

WHy does it default to 15bit color mode? I've never seen that before. Usually it's 256, 12 bit, or 16 bit.

[@vswitchzero]

Yeah I wondered the same. I’m assuming it’s a Matrox specific feature as I’ve never seen it with any other cards before 🤔

[@southernflatland]

I've written video drivers before, I can explain 15 bit. It's simply easier than 16 bit.15 bit divides cleanly by three. The results are still stored in a 16 bit space though, but go ahead and try dividing 16 by 3 (RGB). It doesn't work evenly does it? Sorry I'm probably not the best to explain this though, they always say the programmers aren't the best to write the documentation.

[@southernflatland]

@@vswitchzero VESA standards had options for 5/5/5 RGB or 5/6/5 RGB. 5/5/5 was 15 bit, 5/6/5 was 16 bit. Both got stored in a 16 bit space, the 15 bit form just like either ignored a bit, or at best used the lacking bit as a mask bit in more advanced modes/processing. But hey, you try the raw 16 bit mode for a bit. 5 bits red, 6 bits green, 5 bits blue... Math don't easily add up homie, and honestly the 15 bit mode was easier to work with. 5 bits red, 5 bits green, 5 bits blue, and an ignored bit to easily work around in a binary 16 bit addressing space.

[@mal2ksc]

@@southernflatland What worked best in practice was to run at 6 bits per channel, but either temporal-dither the red and blue, or just drop the least significant bit. Software certainly has no problem operating in a mode that's 18 bits wide, or maybe even 24, even if the VRAM is going to throw away a lot of it.

[@southernflatland]

@@mal2ksc I literally wrote VESA compatible video drivers and never once heard of an 18 bit mode. I mean sure it's possible I guess, but it's definitely no standard I've ever heard of. Back in the old days, it was basically 1, 4, or 8 bit palette indexing, or 15/16/24/32 bit high/truecolor. 18 bit would be an absolute waste of precious video memory, as it would have to be stored in a bare minimum of 24 bits, which itself was slow because it didn't align with memory segment boundaries. For speed purposes you wanted either 15, 16, or 32 bit, because those cleanly aligned on memory boundaries. 15 bit was a bit faster than 16 bit because it could use the same algorithms for all 3 color channels, ignoring one bit though, to align on binary memory locations. Your supposed 18 bit mode would have been blasphemy in video programming, both in terms of speed and also wasted memory. Edit: In 15 bit mode, the ignored 16th bit could be used for masking. Not like that was super common, but it was definitely a neat novelty of such modes.

[@kokodin5895]

be crazy, repopulate missing memory connector and upgrade memory in them both as a experiment

[@JarrodCoombes]

@5:03 - 15 bit high color?

[@vswitchzero]

Yeah very unusual. I've only ever seen this with Matrox cards in Windows NT/2K. The Windows 9x drivers don't have this option, but 24-bit colour is also an option, which is interesting.

[@primus711]

I would switch the chip out obviously they used cheap junk and a utility needed to keep flashing it would wear it down more

[@vswitchzero]

Yeah that’s a good point too!

[@marvintpandroid2213]

Bit rot, I know how that feels.

[@PROSTO4Tabal]

The person who is reading this comment , I wish you great success , health, love and happiness !

[@jaybird57]

I paid like $200 in 1997 for a 4mb mystique.. yikes.

[@southernflatland]

Why did you screw in the cables? That itself was like the BIGGEST form of failure I ever saw in my shop. Some fool forgets it was screwed in, then rips the port right off the motherboard/video card... Seriously DON'T screw VGA cables in, they're already grounded quite fine without the screws. Unless you just like ripping ports off the board...

[@TomStorey96]

Seriously, don't ever eat any food because one person died of food poisoning one time... And don't ever drive a car because another person crashed... Etc etc. Really?

[@southernflatland]

@@TomStorey96 Why in the world do you think that ALL modern connectors are designed with quick disconnect? USB? HDMI? Lightning Port? Etc etc... Because of this very reason, they actually LEARNED from the mistakes of the past. This has happened WAY more than one time in history, more like tens of thousands of times. People are simple and they sometimes forget simple things. And one of those simple things was apparently forgetting they screwed in their cables which led to countless boards getting destroyed when they decided to move their computer or devices somewhere else, or just simply kicked their feet at the wrong angle under their desk. Engineers recognized this problem and did away with the cable screws on newer ports. So intelligent people might heed my words of caution to just leave the older cables unscrewed, it'll still work perfectly fine while also avoiding said risk.

[@TomStorey96]

​@@southernflatland Tens of thousands of connectors may have been broken, but how many tens or hundreds of millions werent? The numbers are nearly insignificant. 10,000 out of even 1,000,000 is already 0.01%... One "fragility" has simply been replaced with another. USB etc connectors still break off boards due to being stressed side to side or up and down, so I dont really buy that as *the* reason for changing them. The screws can also be used to anchor the connector to something structural like a faceplate or bracket, rather than relying solely on the pads the connector is soldered to, so I would argue that such older connectors are mechanically stronger than modern connectors. I would also argue that connectors with screws are bigger and bulkier and not compatible with the nature of modern electronic devices which are smaller and thinner. There are also other considerations beyond the physical form factor and how robust it may be, such as signal integrity - older connectors werent made to handle the huge data rates of modern signalling standards, for which more compact and appropriately shielded connectors and cables are required. Manufacturing capabilities are also much better than they were some 50 or so years ago, making it possible to manufacture smaller connectors. I would further (and finally) argue that it is more of a convenience than protecting consumers against accidentally moving something with a cable still attached. If that were the case we wouldnt have flimsy micro USB connectors on phones, and Ive heard of wayyyyy more of those breaking than I heard D-sub connectors breaking. And how annoying is it when you kick a cable out from the back of your PC under the desk and have to go crawling down there to plug it back in? Doesnt happen when the connector is screwed in...

[@southernflatland]

@@TomStorey96 You can argue your thoughts all day long until your face turns blue. I'm speaking from over 10 years shop experience. Those screw anchors are like NEVER secured to the faceplate, they're anchored to the circuit board and it absolutely destroys the board when screwed in cables get ripped loose. At least quick disconnect cables have a fair chance at coming loose in a pinch of such an incident. The only time you really want an anchored connector is in a permanent weatherproof installation. But hey, you do you, nobody ever listens to the technician anyways. Sucks for all the folks that destroyed their motherboards and graphics cards though.

[@southernflatland]

@@TomStorey96 Also, as far as cables accidentally getting kicked loose... Yeah it's annoying, sh!t happens though. It's MUCH better for the cable to come loose than it is for the port itself to rip off the board though. Inconvenience is WAY better than damaged hardware.