I think I'll just salvage CPU socket for this motherboard, it's been waiting for socket transplant for a few years now.

Make sure the pins in the socket are still ok, which I kinda doubt looking at the CPU's pins ...

Weird that they did put legit cache chips and didn't socket them, that's what they do most of the time. Maybe they're still fake ?

I had removed top plastic off the socket while washing MB, all of the pins were fine, except one, that was pushed in. Overall, CPU socket and ISA slots had preserved phenomenally. I also do find it strange, that legit cache is soldered in, probably they were trying to save every cent in manufacturing process.

Is it possible to peel off the big chip sticker? I'm rather curious why there even is one, rather than the chip itself being marked.

Probably nothing of interest, PCChips often did that in the 486 era, underneath there's just a single digit number and that's it. However the chipset of these m912 is really a UMC chipset. But they often put other stickers there.

Here you go:

These appear to be clones of UMC UM8498F and UM8496F.

Restored memory modules from this:

To this:

Interestingly, these modules are using ICs with ZIP through-hole type case, which are converted into surface mount type.

Also, removed chipset, but both chips had cracked at the underside, due to vapor pressure from absorbed moisture in the plastic... I hope that crystal didn't move when case poped and bond wires are still intact.

Chipset ICs after restoration:

I've also desoldered and tested SRAM ICs, now I know exacly why they've soldered BIOS and cache in. "TAG" and ICs at "BANK 0" are legit, but ICs at "Bank 1" are all fakes 😀

Can you visually tell apart which are which?

Fakes are just plastic with legs, measuring resistance between Vss and the rest of the pins, shows infinitely large resistance.

For the cracked chip you can simply replace it with a regular UMC chip, because this is what these are.

Could not find such UM8496/UM8498 chips to buy, maybe they have never officially existed? Tried to look for the list of all UMC chipsets, but found nothing.

Well it does, look at this https://www.ebay.fr/itm/UMC-UM8498F-Baby-AT-3 … 4.c100408.m2460

Mustn't have been very popular though

I've decided not to waste time trying to make PC Chips MB post and just salvaged its socket 😀

The recipient is prepared for socket transplant:

Reassembling socket after days of trying to whiten it:

Wasted nearly a week trying to whiten that socket. Kept it in hydrochloric acid for few days - no go, hydrogen peroxide and UV for at least 3 days - no go, only scrubbing it with hard tooth brush and polishing paste did help. Any ideas how could that chipped VLB slot be patched nicely?

I think I may have that exact PCChips board here somewhere 🤣 Need to do another clean up of all my hardware and see where it is.

very nice work, must of spent a bit of time on this. Did you use hot air to remove the chipsets to make the pins so clean?

Yes, technically it's possible to desolder QFP IC with just soldering iron, but not for its pins favor.

I desoldered a umc chip with a soldering iron only but the pins and excess solder make it look like a xmas tree, in my case it wasn't a problem because I replaced with a new chip which wasn't as rare as yours

Yours turned out perfect, did you use any new solder or just alot of flux and hot air?

Hot air and flux alone wouldn't do it, the oxide layer was just too thick for pins to tin, basically layers of lime had formed on them.

Pins had to be cleaned off thick layers of tin/lead oxides first, used diluted hydrochloric acid for that purpose. Once thick layers of solder alloy oxides had fallen off the pins, I've removed ICs from acidic solution.

After that, pins were still covered in very thin layer of oxide, so acid based flux had to be used to tin them properly, as tacky "Amtech" flux was not good enough. Used regular soldering iron in the tinning process, as some scrubbing was needed.

You're being very brave to restore such a piece of garbage of a motherboard. 😵
This is one of the infamous PCChips motherboards with fake cache and counterfeit chipsets. According to the Red Hill guide, the M912 was available not only with the UMC "sticker", but also other ones, including Intel. So basically they just used a chipset (or ripped off an already existing one), put a sticker on the chip according to what the customer wanted, and sold the M912 with any chipset they wanted, as if the fake cache chips weren't enough...

More info here: http://redhill.net.au/b/b-bad.html

I hope the traces of the fake chips actually lead to somewhere, otherwise you'll end up with a quite crippled 486 motherboard, if it ever comes to life. Oh, assuming the cache actually works. On fake cache motherboards, the BIOS was modified in order to keep the L2 cache feature disabled regardless of the user's choice. That's why they soldered the BIOS chip: not only to save costs but to also avoid that someone would discover the scam. Perhaps the motherboard might even POST without any cache at all, even though the option in the BIOS would be set to yes!

Ironically enough, when PCChips motherboards worked properly and used legitimate components, they were some of the fastest motherboards available on the market. But their shady practices and low quality construction just made them utter rubbish.

Yes, I know the history behind 486 PC Chips motherboards and company's practices of using fake cache and false chipset marking, IMHO that makes these motherboards somewhat unique and interesting in a way. For now I'll postpone the PC Chips MB restoration as I don't even have a CPU socket to test it. The funny thing is that I've previously bought a slightly newer version of the same PC Chips motherboard (M912 V1.7) as non working in an attempt to salvage its CPU socket. It turned out that MB was 100% fine, except 3.3V VRM transistor had blown, probably because of wrongly inserted CPU. Gah, I really wanted to pull out a ZIF socket out of it 🤣

Success:

This motherboard has a huge flaw in PCB design around CPU socket. It definitely doesn't meet power integrity requirements for 100MHz+ CPUs.

Any suggestions how could this be fixed W/O resoldering the whole thing?