Disaster struck while working on this odd-ball 386SX VLB motherboard 

A couple notes: I tested the two sticks of RAM and they were 4mb each. I also tested the math co-processor and that works just fine in another 386SX board.

It's a common misconception that the 386SX is basically a 386 "with th 286 bus interface". While it is completely true that the number of address and data lines on the 386SX is just like on the 286, the *protocol* of the 386SX is very much like that of the 386DX, but it differs from the 286 bus protocol. For example, the 286 bus protocol generally allows the processor to output the address of the next bus cycle while the previous cycle is still being finished, and actually the address pins are *always* valid half a processor clock (one period of the CLK2 signal) before the bus cycle is started. On the other hand, on the 386 bus protocol, the address stays valid during one cycle unless the mainboard actively allows the processor to output the address of the subsequent cycle (using the "NAB#" signal - "next address to bus"). The 386SX uses the same timings and protocol details as the 386DX, it just has fewer data lines, it misses the top 8 address lines and has the extra line A1. Also, it only has two "byte enable" signals, whereas the 386DX has four byte enable signals. Furthermore, you are in fact correct that most 386SX chipsets also support 286 processors, but the chipset needs to be configured for the bus protocol to use. Most 286 chipset vendors added support for the 386 bus protocol in their newer chipsets to increase the market for their chipsets. A passive adapter from 286 to 386SX on the other hand won't work. As I don't expect that board to support VLB masters, it's up to the processor to determine the width of cycles.The VESA local bus is perfectly able to handle 16 bit cycles originating from a 32-bit processor, so the VESA local bus is also able to handle 16 bit cycles originating from a 16-bit processor. On the XT-IDE card, the IDE/ATA interface *requires* 16-bit cycles to be performed on that interface, which is something 8-bit ISA is not capable of, so that card needs to assemble two 8-bit ISA cycles into one 16-bit IDE cycle. I am very confident that your board will not run any VL card that requires bus operations to be performed as 32-bit cycles. VL VGA cards generelly do not require that (unless special accelerator feature do), as well as VL IDE controllers also work perfectly (also sometimes slower) with 16 bit cycles only. There is one thing the board has to perform, though: The high two bytes of a DWORD *must* be transferred on D16-D31, so the board needs to dynamically connect either D0-D15 of the VL slot to D0-D15 of the processor, or D16-D31 of the VL slot to D0-D15 of the processor. That's exactly what the four (74)F245 chips can do. In fact, stuff like this is also performed on some video cards, like Cirrus Logic cards. The CL-GD542x series are *16-bit* local-bus capable graphics chips! The VESA local bus has a signal to tell the processor that only a 16 bit cycle has been accepted by the device and the board/processor is required to re-issue a cycle for the remaining byte(s), if any. Logic on a CL-GD542x card detects whether one of the two low bytes is active, and if yes, it performs a 16-bit cycle using D0-D15 (or an 8 bit cycle with half of the data lines), otherwise it performs a 16-bit cycle using D16-D31 (or a subset). In essence, this means the 16-bitness of your board should not degrade performance of a Cirrus Logic card. This does not apply to the CL-GD5430 or CL-GD5434, though.

5:37 Those for chips are NOT latches, but high-speed bidirectional bus drivers. The model is F245, or 74F245, the F variant of the common 74LS245. The fact that there are 4 chips seems to indicate that the VLB is indeed 32-bit. However, there are 16-bit VL cards. The most famous example is the CL-GD5428 graphics card which use a 16-bit data bus and 24-bit address bus instead of 32-bit. A standard for 16-bit local bus may exist, but just rarely used.

Hi Adrian, sorry for making an announcement here, but I wouldn't if it wouldn't fit so good ;) Accidentally, I'm currently preparing a video about another 386-ish board, which also has VLB. That is a 386DX (unfortunately not a 386SX), but if everything is going as planned, I'll try to show some performance results as well. Btw. your board is not quite a 386SX, it is an extended version made by Cyrix/Ti. It is a reduced version of a 486, which was made to fit 386SX boards and is also known as IBM lightning.

People are here for your personality, expertise and your content - I don't think people are too bothered about a bad zoom on the hardware or bad cuts - loving this channel almost as much as your main one so just do what you enjoy and what's easier for you.

My first 486 was an IBM 486SLC2-66 with VLB slots. The 16 MB RAM limit wasn't an issue back then (I only had 4 MB in it anyway), and with 16 kB of L1 cache, it was almost as fast as a 486DX2 when running small CPU-intensive tasks -- but obviously the 16-bit external data bus slowed it down a lot when I/O was involved. Supporting 32-bit VESA Local Bus cards wasn't a problem; it just had to transfer the data in two 16-bit chunks at a time. My only complaint was that the little one-inch-square CPU had a tiny fan on it that was not only loud, but also didn't fit under the drive cage of the desktop case I wanted to put it in! So I was forced to get a tower case with more room above the motherboard.

Hi, that looks like an early "Blue Lightning" IBM 486SLC2. The rev after yours even lists its processor as a 486SLC2.

I like the setup. I feel like I'm sitting at the workbench with you watching you do your thing. I appreciate you making me comfortable. Disasters happen, been there done that. Good content.

The new setup worked really well. It is sad to have the motherboard killed in that manner. With that kind of power failure, any number of chips could have been burned out. 😕

It's a bit odd to refer to a 386SX as "basically a 286," since it still had the full 32-bit instruction set and the i386 protected mode (which was vastly different than, and superior to, the i286 protected mode). There's a *lot* of stuff that runs on the 386SX that wouldn't run on a 286, even if the overall performance isn't much higher for 16-bit applications.

1:20 I wouldn't call a 386SX "essentially a 286". Sure it has a similar bus and similar performance (on 16-bit code), but the instruction set was heavily updated and orthogonalized for the 386 (including the later SX). Actually more so than on any other new x86 processor. It would be more correct to say that the Pentium III was essentially a Pentium II (or Pro), for instance.

A real bummer that this board got toasted by a random PSU failure. But props to you for posting this "mission failed" video anyway! Always interesting to see the artifacts you turn up.

Gotta quibble with the equating of a 386sx with a 286. It's true that they share (almost) the same bus interface and tended to go into the same sorts of AT clones. But the 386sx was internally a real 386 core, 3x as many transistors, 32 bit registers, a paging MMU, emm386 compatibility, dos4gw, the works. Hell, you could (and many people did) run Linux on the 386sx. But it's true that almost all the chips were deployed in environments where they only ran AT software.

Thank you, Adrian. So nice to see that one isn't alone in having old stuff fail on you, damaging something you're working on. This actually makes my life more bearable. You're great! Get better soon!

Just to clarify "a 386sx is basically a 286", while true it has a 16-bit external data bus, it also has an internal 24-bit address bus. Plus it includes the 386 processor extensions to the x86 instruction set.

A 32 bit processor with external 16 bit bus using 32bits slots... a bit madness. wide>narrow>wide

@9:30: Yes, there were adapters from the 286 to the 386SX/486SLC bus. The probably most popular one was called "Make-It 486". They were, however, not just (mostly) passive adapters; they contained at least a CPLD and a PAL.

As I recall, VLB was basically just an extension of the 486's bus pins with some buffering thrown in to keep the processor stable. When the Pentium was released, VLB quickly fell by the wayside because it needed extra hardware to translate the 486's bus protocols to the Pentium's. At that point, it just made more sense to implement the platform-agnostic PCI bus.

Thanks Adrian another great episode. Cheers

Thank you for sharing this. Somehow it makes me feel less sad about my own failed repairs =)

Thanks Adrian for making a video even when feeling rough! 😊

Boy...I was working at a reseller from 1990 onwards...and I barely remember VLB. It was a very short-lived bus. A troubleshooting session with you is a great learning experience, regardless of the outcome.

As for the motherboard not working with the XT power supply -- it may require +5V on pin 2 of the P8 power connector, while XT power supplies do not have anything connected to this pin. I haven't heard of any AT motherboards actually requiring +5V on pin 2 of P8, but it's just a thought.

Really enjoying the added production value of your OBS setup for the 2nd channel - awesome 👍

15:23 The VESA spec says the maximum frequency for VL-bus is 25 MHz if there are 3 slots, 33MHz if there are two slots, and 40MHz if there is only one slot. I think this has something to do with signal reflection, or just the load on the bus (in the load case it would be possible to run a 3-slot VL-bus at 40MHz if only one card is present, I don't know whether it is true or not). 50MHz is possible if the device is on-board instead of plugged in using a slot. However, few chipsets support 50MHz FSB 486s, and even fewer of those motherboards have an on-board VL device, so this is hard to verify. Plus, starting from DX2, intel stopped selling 50MHz FSB chips (because few motherboards support that!), the overall performance would be better on a 66MHz DX2 with 33MHz bus than a rare 50MHz DX with a 50MHz bus. Note that these are all guaranteed specs, in real life no one stops you from clocking a VL-bus with 2 cards at 40MHz just like no one stops you from overclocking. The only problem is that stability is not guaranteed outside of those frequencies.

Thanks for another awesome learning experience. Did not know any of that about VESA slots. Now I finally understand what latches are too!

I had a 486slc ibm...it was a weird chip might have been a blue lightning but it for sure was not a 486 chip, it had both vlb video and hd cards. the video was a number 9 vlb card... it was a very nice.

Hope you feel better.

Hi Adrian. I feel the same about your motherboard. I have 2 Packard Bell mobos that both suffered Varta damage. Both have the same small chip come off with corrosion. I bought the one in an attempt to fix the other, but that hasn't gone over well. You keep up the good videos

Very good video indeed ! Your experiences here immediately tells me I have to use an external crowbar circuit on the power lines when testing motherboards. The old power supplies are prone to have bad electrolytic capacitors inside. Commonly seen as the top of the caps bulging out or the bottom showing signs of leakage. When the bad capacitors no longer is able to hold a charge the output voltages drops under load causing the regulator circuit pushing the power side to it's limits. - and then disaster strikes ...

At the end, it looks like pins 80 and 79 as well as 75 and 76 are bridged on the 206. It didn't look bridged at the beginning of the video.

get well soon mate!

What we said at Microsoft at the time - and it was generally applicable - was "Remember: the S stands for Sux!" (Yes, that applied to Win32s as well.) But admittedly, there was one advantage to the 386sx from a development standpoint, in that you could treat it as a 386 and not have to deal with bloody segmenting. While technically it (and the 386dx) still had segments, they became settable and arbitrarily large, so you could set your segment to all of RAM if you wanted. All that aside, something I was told by others at the time was that they were in many ways the 286 that should've been, and if you ran 286-targeted code on them you'd get a really big performance uplift. I never tested that myself, but I can buy it - the 286 was absolutely not ready.

Just for the record - I have an old Motherboard from a Compaq SLT386 Laptop which actually refuses to boot without a battery. I would never disregard anything.

Kudos to you for the new camera setup ;)

This format with OBS and "uncut" it quite nice :) Take care and get well, Adrian!

Holy cats, I think I had that mainboard in my very first PC. The dual VLB slots and the soldered-on 80386 are really familiar.

Zdróweczko Adrianku 🍷

Might be worth looking into some sort of crowbar overvoltage protection circuit that shunts all the rails if one of them goes out of spec

When I see VLC buses on 386, I usually check first that they are actually Vesa Local Bus and not Orchic Local Bus and other Local Bus from the same era (same connector, different pinouts)

Two thoughts on the board: 1) is it possible that the different pitch of the beep codes with and w/o the video card was already sign for a short or problem in the PSU? 2) AFAIK use of the 32bit extension on VLB cards was optional, i.e. if the card is inserted into a normal ISA slot, it still works, yet slower. Therefore the VLB slots on this board might be just for better physical support of a VLB card - probably only connecting ground pins of the 32bit extension.

what a shame it didnt work but them are the breaks sometimes. still worth the watch keep doing these

Thanks!

I remember buying my first PC compatible, a 486. The local computer store was selling two with "enhanced" slots: one with VLB and another with EISA. Taking home all the printed documentation and a borrowed Byte magazine (this was before the Internet @ home was a thing), I did all the pros & cons. Eventually, I order the EISA board and all the necessary cards. I must say that my machine was super fast, compared to the run-of-the-mill 386s & 486s at school. Although I didn't like that the EISA cards & BIOS required special software to configure them, but all 8 slots supports 16bit ISA / 32bit EISA. The VLB mainboard only had 2 VLB slots, and the rest were 16bit ISA -- so EISA for the win.

With the finding of the broken wire to the OPTi chip that makes me think the beep was actually 1 beep short repeated.

The PEEL18CV8 was a competitor to the 16V8 and other GAL chips. I haven't seen one in several years, not since I lost the only one I had when spilling some chips off my workbench. Their proximity to the VLB slot extension leads me to believe they're part of what makes this work. I know some versions of the 486SLC chips had "System Management Mode" so it's possible it emulates a 32-bit bus using software and extra hardware.

Thank you because of you i started making diy tech YT videos

When you zoomed in on the opti chip, it looked like there was some rogue solder or something bridging some pins further along the edge (bottom edge, to the far right, near the corner.. as looking on the video) and then also, the larger chip above, the bottom right corner (again as looking from the video) - looked like there was a rogue piece of metal or wire going from the via or something...

Still satisfying! Things happen.

I vividly remember owning such a board when i was a kid. Because it was my very first own PC. A Highscreen branded desktop with 386-SX 16 and a whole 1024 k of RAM. Also a 40Mb HDD and 2 Floppies. But mine had a Phoenix BIOS, not AMI. I also remember that i had to get a VLB I/O card to get the fast UARTs to drive my 14.4k modem. The ISA ones only had 9600 bps UARTs on them. The big advantage was that the VLB ran with the full 16 MHz CPU clock, twice as fast as the ISA slots. And if i remember correctly, the CPU handled the two busses separately. That way my fancy terminal programs screensaver didn't mess up my Zmodem downloads anymore.

Last week I had a headache... then a headache and a sore throat... then it turned into a full-blown headcold that I'm still fighting. It's out of my sinuses, now, but lingers in my lungs. So you might have another week ahead.

Might be worth going around some of the components and seeing if you can find any obvious shorts and see if it is as simple as just a couple of shorted capacitors or resistors.

I had a PSU go bad while working on my parents' IBM PS/1 Consultant years ago that took a video card (ATi Mach64 ISA), network card (D-Link DE-220), and motherboard with it. Whatever voltage spike went through the system was enough to crack the main chip on the video card. The computer would still POST with another video card, but when in the BIOS setup, every other character was scrambled. It was a very sad day.

I had a 586-133MHZ system with VL-B, it was the Image 128 graphics card. Anyways I always had issues with the card popping out of the extended slot all the time. It was a very tight fit, amazing the card nor motherboard never got damaged.

What a beast, my folks had me contribute my life savings in grade 5 to buy a 286 SX, was it 16 MHz?. I remember the video card had pin sockets wit unoccupied chips so you could add your own 512k memory...

I remember Alaris from back in the day. Bought a mobo at a computer show once, and the vendor said "they were an OEM for IBM." I guess based on what you found in the search, he was speaking the truth. You've solved a mystery for me... that started more than 20 years ago!!

Interesting content as usual and you do a great job of explaining what you present. I have a question nonetheless. How is Portland surviving without a local Walmart?

Wow that is frustrating having a PSU destroy good hardware! Thanks for proceeding to share it with us even though it ended badly.

Hi Adrian, I had a 386DX back in days and I had experience with this configuration as well. Using a very similar Cirrus Logic card the graphical speed was like half. Not many software used the graphical acceleration that time, but with the SX series it's was like no acceleration at all. On price wise it was a 25% cheaper (the whole PC) but it definitely not worth to save on this boards. This is my first PC was 386DX and I saved on the Sound Blaster compatible 8bit sound card instead. I also tried to save on the Quantum Hard Drive, but that was a failure as I had to replace 3 times in one year. Anyhow I like the video and thanks for your hard work.

Bummer, I was hoping you'd be able to save it!

watched this feeling a bit down with a cold..so sympathies ... hope you are feeling better.... as for the board .....ratssssssssssss grrrrrrrrr

I remember seeing some systems that came out *before* the VESA Local Bus that included various sorts of proprietary bus extension slots. I suppose one of these got enough industry backing that it became VLB (some of the ones I saw looked much like the VLB). If I knew where my CompuServe email archives were I could look up when it was, since I had emailed John Dvorak asking about them. One of my VLB systems also had a drive controller that plugged into the second VLB slot, and took regular 30-pin SIMMs to add cache to the controller. Now I wonder just what happened to all these boards and such when I upgraded them (I haven't seen them since even *before* our house fire, which is where I lost a lot of old components).

Boards of this vintage usually don't get my interest often but this is one of those rare examples that I didn't know existed not that VLB 386 is new to me but for the SX wasn't expected at all.

This was entertaining nonetheless! Thank you :)

The differences between the 286 and 386SX were a lot more than a few lines. Protected mode worked, it was 32 bit internally so could run 32 bit code (unlike the 286) but had a hobbled data bus to 16 buts to make it cheaper.

Oh Boy!, this video brings back my early memories of PC days of working for a Clone-PC company, and yes there were some very cheapo motherboards, and as posted below the chinese card makers used very sloppy insertion specs and the cards could move front to back inside the slots, you had to line up the pins yourself and then push the card down into the slot, and sometimes it would move between pins and not post. Oh was I glad when this died off, then PCI bus, and then the kinda lousy AGP slot era. it was a bandaid ontop of more bandaids, and most just were regular cards that you could put in a VLB or AGP slot and no better performance was noticed. a lot of problems go away when you put MR Bios on these old boards if available VOGONS has a zip file with a ton of Mr Bios files.

IIRC this motherboard was sold at b&h photo. It was a dirt cheap 486 motherboard and I was almost going to buy it but upgraded my 386sx with the cyrix piggyback 486slc. That is when I realized the cheap IBM 486 motherboard was a 486slc not a real 486

My first Pentium computer was (someone else's) homebrew with a Canon Pentium 66 motherboard with ISA and JUST Vesa bus, no pci. It was a dog, hardly faster than my 486DX, 100 I think, also with VLB. I eventually sold the Canon MB to a fellow for $200, way more than it was worth, because he was a collector of rare and unique computers and components.

This is why I don't like using old AT power supplies, but use ATX's with AT adapters, the fear they'll fail and take hardware with them. Same with original C64 power bricks. But I don't have wizardry repair skills either, hence I'm extra cautious and in awe of those who's got em' 🙂 Regarding 386SX, I think along with more MHz, it's also got double the amount of transistors, so still quite beefy compared to a 286. But would never have guessed it existed with VLB either. I hope you feel better Adrain, and thanks for making and sharing the video despite everything 😊

I actually have a Pentium MB with VLB on it. I have a 200MMX om it with a Sound Blaster AWE 32 VLB board.

I used to own a 386 server board pc.. the actual processor and ram were on a daughter board with VLB... the motherboard had 3 slots for VLB boards (1 for the processor board)

I'll have to check the VLB board I have in the garage, see if it's 486 or 386.

I tested local bus vs pci on a 486 SX 50mhz (so basically a 386, and I guess the bus was running at 25mhz with a multiplier of 2). The game Screamer has a "test your graphics card speed" option. I don't remember the numbers but the PCI card was faster than a local bus card. Both had 1mb, the PCI card was a Cirrus 542x or 5430, I can't remember, the local bus card was a "chips 64300" (the drivers in Windows called it "wingine").

I had a 486sx w/ math co & a VESA video card. It had higher video resolution (800 x 600) compared to my former ISA VGA. I didn’t notice much, if any speed increase.

I have this exact board, the CPU is an IBM 486SLC2 - mine running at 66MHz. My first 486 had this board in it (not the exact one I have now), they were often sold in bargain basement PCs where they could use the cheapest parts but still market as a "486". Although it's really a 386SX as you said, it's surprisingly fast as it also has 16KB of L1 cache, double that of a 486DX2. Probably on par with a 486SX-33 as far as speed. As for the VLB, I have a Cirrus Logic VLB in mine which runs decently but I haven't done any comparative benchmarks, but I mean to do so one of these days. I have tried installing VLB IDE and SCSI disk controllers in the board but have been unsuccessful in getting anything but a video card working in them

We watch these videos because we always learn something. A bad board sure it's sad but things happen and sometimes things go to shit. I have a idea what have maybe killed the Cpu. When you lifted the board and it stopped beeping a few times in thinking maybe some pins on the back made contact with the case of the psu. Anyhow I love your videos you are awesome. Keep it up and I'll see you in the next one. Cheers from Sweden

26:41 are there some bridges on the upper pins of the Opti chip (75 to 80, lower right in image)? There look to be posibly 3 small blobs of solder. They could of course just be non-conductive dust / dirt particles.

You assume that adding VLB slots is for speed increase, it could just be added for compatibility with VLB cards. Also, while some 486 had PCI slots the initial PCI version was broken. Most cards for the 486 were either EISA or VLB.

i've run VLB at 50 MHz (486Dx50 cpu) it was suppported but finding VGA and multi IO card that were happy at that speed was tricky

Sorry board got fried, that can't have helped you current cold! Thanks for uploading anyway, helps to see that sometimes even you can have issues like this.

Can’t win them all. I’m curious. I remember back in the 386-486 days testing boards and half of them had some “feature” where they would NOT boot without a battery. I remember in part because in recent years restoring old Mac’s it shocked me that people just yanked the batteries and started machines up without replacing them.

Power supplies killed so many computers back in the days!

I just sold an IBM Blue Lightning SLC with VLB - it was also a 16 bit processor at the core. I've seen 386 SX VLB mobos, albeit with a socket 3 upgrade path.

For those that are learning from this video, the 386SX isn't a 286 with extra signals. As it is correctly said later ,it's a 32 bit 386 CPU internally with a 16 bit external data bus, whereas the 286 is 16 bits internally and externally and is missing 386-specific CPU instructions (of course).

I like the format, keep it up! Sorry about your Mobo!

VLB 2.0 was released in 1994, with a 64-bit bus, and a bus speed of 50 MHz. So with the Pentium 64-bit data it was able to use the later spec, but with PCI coming out for the Pentium VLB went out of favor

The first computer I ever put together for myself (after being pretty badly ripped off buying my first PC, an essentially useless 286, from a guy at the mall just selling a computer at a table) was with a 486 SLC. I found a motherboard for sale in a catalog that had the processor already seated and thought the specs sounded good enough. I worked a lot as a teenager and ended up buying a pretty high end VLB graphics card to use the extended slots. This was all so I could run Mortal Kombat so that program was the benchmark. It would run great for a few minutes but would always lock up no matter what I did. I think I even remember the requirements for the card exceeding my system specs although I bought it anyway. I thought that SLC stood for 'Single Line Com', although I may be remembering that incorrectly. It ran everything else fine, though. I had a lot of fun with the system. The 286 I sold to a friend who insisted that I sell it to him although I assured him that it could play nothing but older games and had to emulate color. It came with a small monochrome monitor. The seller gave me a disk to boot the system and one to emulate Hercules CGA graphics. I think he included a copy of Pharos Tomb. I sold it to him and used the money to buy a VGA monitor and the first parts to my new system. Eventually I had a nice gamepad, sound card, graphics card, etc. and put it all together myself. I got Saturday detention a lot which was held in the high school library. I would go to the job finder computers in the back and reset them then enter DOS to play games I would sneak in from home. I remember the librarian catching me playing a demo of Desert Strike and being surprised that the computers could be used for anything other than the job finder program. I explained to her that it was just a computer and that it had DOS and Windows although nobody ever used them for that. I even showed her 'CTRL ALT DELETE'. She was not angry with me in the slightest but rather impressed. I later brought blank disks in and copied Windows and DOS for my own use. This was back when Windows often was laughed at for being useless and I would complain about how DOSSHELL was better if someone really needed to click on icons. Since then I have made so many computers and done so much with them. It all started with a 486 SLC. 😎

Leopard might be an Alaris SLC board. Had a very similar one up until about a week ago. You're looking at an Alaris 33 MHz 486 SLC with a 16 bit external data path. I made a video about a similar computer years ago.

I miss the Dr. Debug boars. Nowdays they are built in the motherboards.

Boom, unexpected combination - what kind of chipset is on there ?

The Alaris Leopard, i have the same card with a 486SLC2 66MHZ ibm cpu. It can run at 80 MHZ by playing with the jumpers in an undocumented setting. I recently got a VLB card and will test if there is an improvement with it.

I had a 386SX 25MHz back in the early 90s.

1996: I had a 486 MB with 3 PCI bus plus ISA. My processor was an excellent AMD 486 DX4 120 MHz.... Great times, great memories

When I finally switched from my souped up Amiga 1200/030 - I specced up a 486DX2-66 with VLB for SVGA graphics and hard drive cards. Good times for the very short lived VLB!

Tracking the power-good trace from the PSU might be us... haha, I was typing this while continuing to listen to the video and there you go. It was that being part of the problem. Next thing I would verify that is present is the osc is running.

Reminds me of those later 486 motherboards with PCI slots, had a friend that bought my first Voodoo card and had it running on that, could even play Half-Life, well a bit poorly.

I remember than the 486SLC or 486DLC version were a 386SX and a 386DX with co-processor. The performance of the 486DLC-40Mhz was similar to Intel 486DX-33Mhz but with un chip very cheap. I am not sure but there were a VLB mainboard for the first Pentiums 60Mhz and 90Mhz or a mainboard with PCI and VLB combined.

My Dad's school had some 486SX-25s with VLB (DEC LPV) and 486SX-33s with PCI (DEC Venturis), although I don't think I ever compared I/O speed on them. Also, the IDE controller in the LPV had a silicon bug that could cause data corruption if used in 32-bit mode, anyway.

You can definitely upgrade a 286 machine with a 386SX. I've got an IBM PS/2 model 50 with a Kingston SLC Now! upgrade, which is a 386SX plus some extra chips. Thanks for the tip about the retro web website, I hadn't come across it before. That motherboard search looks really useful, so I'll keep it in mind next time I need to get some details on an old board. Anyway, that's really bad luck with your power supply. But I guess it's better that it failed on a weird random board rather than something more noteworthy.

I've actually got a stack of those boards that came from little IBM PS/2 computers. They all have some level of battery damage, though some not so much, but at some point I'm going to be getting rid of them. So ya, I'm not surprised that they came up in your search as being associated with IBM.