VOGONS


First post, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

Hi folks,

quick backstory, this FPU was in a damaged PT-319A type motherboard with SARC chipset and TI486SLC CPU, 386SX class. I don't know if it was a drop in bonus or somehow had been working in the board like that before other damage (battery) as I bought the board cheaply to mess around with and hopefully repair. However, I wanted to try an FPU in my Epson ActionNote, so pulled it.

To find... pin missing and two mashed together. Now there's some out of line pins on the other side also. I could presume that this is just a rather mishandled FPU. However, this motherboard and chipset is kinda notorious for being "quick and dirty" and some hacks have to be done to get cache working and other things. So this leads me to wonder if this is an intentional pin delete and short to "fix" something for this motherboard... or whether it's just broke.

Note, this is a different ULSI FPU than the one that already didn't work in the ActionNote if you are stalking me 😁

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 1 of 10, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

Wow, gimme a sanity check, finally dug up a pinout, unthanks to the help of artificial ignorance and botsplaining getting in the way.

https://www.digchip.com/datasheets/parts/data … 7/387SX-pdf.php

Seemingly the messed up pins are redundant Vss and not connected. It's got a small hope in hell of working.

edit, no wait, let me do that again... think I was hoping too fast, got the orientation wrong...

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 2 of 10, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

Yah it's pin 15 that seems to be AWOL D13 ... maybe could rip 17 out of the substrate and stuff it in there though 🤣 that's NC

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 3 of 10, by Dan386DX

User metadata
Rank Member
Rank
Member
BitWrangler wrote on 2026-05-26, 21:11:

Yah it's pin 15 that seems to be AWOL D13 ... maybe could rip 17 out of the substrate and stuff it in there though 🤣 that's NC

Kek, you could try that. Would it not be feasible to fashion a ghetto pin just for testing?

90s PC: IBM 6x86 120Mhz. 128MB/6GB. ATI Rage Pro 3D.
Boring modern PC: R9 3900X, RX 7800XT. 32GB/1TB.
Fixer upper project: NEC Powermate 486SX/25. 16MB/400MB.

Reply 4 of 10, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

I was just thinking I have some super tiny screws in an eyeglass repair kit, could maybe screw one into there, engaging the lead frame and making contact, and having it as a stud.... hmmm.

I gotta remember where I put my extra tiny drills and my nerves though.

Edit: found my drills, 1.05 mm is smallest... I think pitch on these 387s is officially 50 mil, so 1.27mm, so that's straight down the middle with 0.11 clearance each side... hmmm... if a drink can is about 4 thousandth thick... that's about 0.1mm... so jig with a sliver of drink can each side to keep centered... as long as the print on it isn't too thick... ... ... Also though I had some imperial drills, 1/32 is skinnier... gotta see if I have a micrometer still to check screw diameters... might have to put it on an anvil flat with a dial gauge... if this substrate is ceramic though it won't drill very nice.

Last edited by BitWrangler on 2026-05-26, 22:22. Edited 1 time in total.

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 5 of 10, by Dan386DX

User metadata
Rank Member
Rank
Member

Without knowing the exact dimensions of the legs, I'd be looking at old/dead socketed ICs I have lying around, see if I could pull something off a DIP chip perhaps.

Hope it works though, it'll allow you to run the 386 version of Quake, which is turn-based 😁

90s PC: IBM 6x86 120Mhz. 128MB/6GB. ATI Rage Pro 3D.
Boring modern PC: R9 3900X, RX 7800XT. 32GB/1TB.
Fixer upper project: NEC Powermate 486SX/25. 16MB/400MB.

Reply 6 of 10, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

Spec say 50 mil, eyeball, 7/8" divided by 17 pin, 1.3mm, okay close enough.

So preliminary measurements are in, shit is small...

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 7 of 10, by ekkiller

User metadata
Rank Newbie
Rank
Newbie

One of the pins on the HT21 chip was broken, I repaired it like this.

144KM……

Reply 8 of 10, by Intel486dx33

User metadata
Rank l33t++
Rank
l33t++

What software do you use to test a math-co processor ?
I have a 387 in my AMD 386dx-33 motherboard but I never tested it.
The bios reads it’s working but I never tested it.

I

Reply 10 of 10, by kixs

User metadata
Rank l33t
Rank
l33t
Intel486dx33 wrote on 2026-05-27, 03:28:
What software do you use to test a math-co processor ? I have a 387 in my AMD 386dx-33 motherboard but I never tested it. The bi […]
Show full quote

What software do you use to test a math-co processor ?
I have a 387 in my AMD 386dx-33 motherboard but I never tested it.
The bios reads it’s working but I never tested it.

I

You can use any sysinfo app like: NSSI, Checkit...

Visit my AmiBay items for sale (updated: 2026-03-23). I also take requests 😉
https://www.amibay.com/members/kixs.977/#sales-threads