VOGONS


First post, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie

It started here...

The attachment Screenshot_20260427-201845.png is no longer available

I failed (miserably) at working with QFP 100pin de/soldering (a story for another beer) and had started looking for socketed 386DX boards and systems so I could scratch that itch of taking a 386 to a 486 like I did as a kid. This came up...

The attachment Screenshot_20260428-210534 (1).png is no longer available

Mitsubishi MP 386s. It's an Intel 80386sx-16 with an Intel FPU, 1MB (or is it 2MB?) onboard RAM, and no hard drive or video controller despite a floppy controller and plenty of ports (AT keyboard, 2x db9 serial, and parallel). This one arrived at my house with an ISA IDE card, an ATI LMS audio card, a WSS sound card, yet no video card. I couldn't find much detail on it other than there were some proprietary parts which do end up limiting its capabilities when unavailable.

But a socket is a socket.

The attachment PXL_20260517_165332770-1.jpg is no longer available

The TI486SXLC2 G50 WN is technically 3.3v but 5v tolerant. My crap desoldering was able to facilitate swapping the 32MHz crystal with an oscillator socket for which I've tried both 40 and 50MHz crystals. I also replaced the Intel 80387 with a Cyrix FasMath.

The challenge is this...

The attachment PXL_20260506_032908130.MP.jpg is no longer available

I think this is the header for a proprietary RAM module that is basically unobtanium now. Worse yet, that's your only path to 32bit memory upgrades as there are no slots of any spec on this board. I picked up an Acculogic RAMpAT! Plus which, even loaded up with 16MB 60ns of parity memory, doesn't compete when choked on an ISA bus.

But I'm having fun. I got a Powergraph x24 1MB board, a Sound Blaster 32 CT3600, slid a 32x CD drive where the 5.25" floppy was, hung a 2GB CF card off the controller, and boot to DOS 6.22, courtesy XT IDE ROM on the 3c509b.

Games like Tie Fighter and Doom, of course, run like ass when trying to hammer extended memory, but Wing Commander 1 and 2 are brilliant.

The attachment PXL_20260520_000143138.jpg is no longer available

I still need to sort out issues like the RAMpAT's dead battery, dialing in the best Cyrix utility for L1 cache and clock doubling, and dealing with the checksum error on the CMOS. That last one is a head scratcher. I replaced the battery, so the date/time memory works, but there's an error at boot up and no native way to get into the BIOS at POST. I found a couple generic Phoenix utilities that run from DOS to allow some editing, but neither clear the error.

Will keep searching and will post more photos later. It's really a nicely constructed beast.

Reply 1 of 18, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

I don't think you get a 32bit memory upgrade in an SX class system. It would most likely be 16bit... however that would be at CPU bus speed, not ISA speed, so 3x faster or so.

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 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie

Right, my bad, but yeah...the ISA kills DOS4GW.

Some more eye candy...

The attachment PXL_20260504_224506677.MP-1.jpg is no longer available
The attachment PXL_20260515_184023377-1.jpg is no longer available
The attachment PXL_20260516_235543274.jpg is no longer available
The attachment PXL_20260517_000114746.jpg is no longer available
The attachment PXL_20260519_233823116-1.jpg is no longer available

Reply 3 of 18, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++

Very modern design for 1988, build like a laptop. Gate arrays so no datasheet for chipset.
Do you have high resolution image of that ram expansion? 5 x 20 pin chips are suspicious, it might be something stupid like a PAL based full ram controller.
Options are:
Dynamic Ram 6 simm slots: 16 data + 11 address + 3 ras + 6 cas + we = 37 + grounds
CPU bus: 16 data + 23 address + r/w + oe = 41 + ground

50 pin connector is certainly too much pins for standard DRAM expansion, leading me to believe it might just expose CPU bus. That would let you expand ram with something silly like _one_ AS6C6416 8MB sram chip. Its expensive at $55, but cant beat simplicity 😀

Do you have any logic analyzers? or rpi pico? a good start would be slapping one on this connector and grabbing some logic captures. Making custom ram expansion looks pretty doable

https://github.com/raszpl/sigrok-disk FM/MFM/RLL decoder
https://github.com/raszpl/FIC-486-GAC-2-Cache-Module (AT&T Globalyst)
https://github.com/raszpl/386RC-16 ram board
https://github.com/raszpl/Zenith_ZBIOS Zenith Z-386 MFM-300 ZBIOS disassembly

Reply 4 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote on 2026-11-06, 15:26:
Very modern design for 1988, build like a laptop. Gate arrays so no datasheet for chipset. Do you have high resolution image of […]
Show full quote

Very modern design for 1988, build like a laptop. Gate arrays so no datasheet for chipset.
Do you have high resolution image of that ram expansion? 5 x 20 pin chips are suspicious, it might be something stupid like a PAL based full ram controller.
Options are:
Dynamic Ram 6 simm slots: 16 data + 11 address + 3 ras + 6 cas + we = 37 + grounds
CPU bus: 16 data + 23 address + r/w + oe = 41 + ground

50 pin connector is certainly too much pins for standard DRAM expansion, leading me to believe it might just expose CPU bus. That would let you expand ram with something silly like _one_ AS6C6416 8MB sram chip. Its expensive at $55, but cant beat simplicity 😀

Do you have any logic analyzers? or rpi pico? a good start would be slapping one on this connector and grabbing some logic captures. Making custom ram expansion looks pretty doable

I agree the build quality on the interior is pretty impressive for the era, on par with the best of Dell and IBM offerings at the time. What I have been able to dig up in my research is these (sold only in the US, with MAXY being the Japan-only counterpart) were professional/industrial focused machines. This one arrived with a soundwave analyzing board (8-bit!) and appears to have been raided of several components before I picked it up on eBay. I can't tell for sure whether this one had a memory board or not by inspecting the header for signs of wear, but it did have a hard drive connected to a pretty pedestrian controller card I've since upgraded (no onboard IDE). The CMOS battery was also a wild one I've not been able to precisely source, so I had to find an equivalent and have introduced a CHECKSUM error I can't clear...but at least it keeps good time now. 😎

I wish I had a better photo of the memory board, but sadly that is the best I've been able to turn up. No logic analyzers or RPI Pico here, but I'd be down to pick up some gear to pull readings and figure this out. I'm crap at soldering, but I can cobble together parts and gather readings with the best of them.

Seeing how much performance I can get out of this little engine is kind of a bucket list thing for me since, as of now, it's completely hamstrung paging memory across the ISA bus. Doom is hilarious until you make the screen size a literal postage stamp. 😆

Reply 5 of 18, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++

get a $5 pico and ill help you sniff ram expansion connector - will require getting matching plug or soldering bunch of wires on the back and running system out of case for a while

>introduced a CHECKSUM error I can't clear

dump bios and we can figure that one out too 😀 Hoe does the Bios setup look like?

There is zero info about Mitsubishi "walsy1-1" on English speaking internet. https://theretroweb.com/motherboards/s/mitsubishi-mp3300 looks similar but not really, like someone drew it from verbal description 😀

Does your case also state mp3300 on the back like https://wiki.preterhuman.net/Mitsubishi_MP_386S?

Was anything ever connected to JF2?

https://github.com/raszpl/sigrok-disk FM/MFM/RLL decoder
https://github.com/raszpl/FIC-486-GAC-2-Cache-Module (AT&T Globalyst)
https://github.com/raszpl/386RC-16 ram board
https://github.com/raszpl/Zenith_ZBIOS Zenith Z-386 MFM-300 ZBIOS disassembly

Reply 6 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote on 2026-11-06, 17:25:
get a $5 pico and ill help you sniff ram expansion connector - will require getting matching plug or soldering bunch of wires on […]
Show full quote

get a $5 pico and ill help you sniff ram expansion connector - will require getting matching plug or soldering bunch of wires on the back and running system out of case for a while

>introduced a CHECKSUM error I can't clear

dump bios and we can figure that one out too 😀 Hoe does the Bios setup look like?

There is zero info about Mitsubishi "walsy1-1" on English speaking internet. https://theretroweb.com/motherboards/s/mitsubishi-mp3300 looks similar but not really, like someone drew it from verbal description 😀

Does your case also state mp3300 on the back like https://wiki.preterhuman.net/Mitsubishi_MP_386S?

Was anything ever connected to JF2?

It is indeed an MP-3300, though the documentation and diagram from theretroweb.com only line up at a 20,000' level. That 32-bit memory slot is obviously misplaced, and there are no jumpers/switches on the board I can find anywhere. Mine arrived with nothing connected to JF2, and I haven't been able to find details on what it is for.

I'll dump the BIOS and message about that later today if I can, this weekend for sure. Which Pico do you recommend I pick up?

The attachment PXL_20260620_185444451.MP-1.jpg is no longer available

Reply 7 of 18, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++

2x cheapest standard pinout pico 1/2. You will also need 2x usb cables so either official picos + mini usb, or clones + USB 3 cables, whatever you already have.

J25 connector is imo unobtanium, looks very close to 50 Pin Mini D ribbon (used for laptop printer port sometimes?)
https://www.ebay.com/itm/195977201948?var=495835011357 https://www.ebay.com/itm/235271973398?var=535302794732
pin pitch matches, but wont mate unless original connector plastic case gets snipped away.
Cant find its matching pair in pcb version, only cable.
Ideal would be finding right angled PCI (old school, not pcie) connector, but those dont seem to exist either. How is the space above it in the case? is there space to plug card perpendicularly or is there disk, metal or something else above it? if there is space you could totally solder cut to size PCI connector, then potential ram board could be plugged into that.

Personally I would desolder J25 connector and
-directly solder cables to two picos for capturing
or
-2.54mm single row headers. Those wont fit soldered normally next to each other, have to solder one row and then pull off plastic before soldering next one. Pulling off plastic will be VERY annoying. Finally connect to picos with short dupond cables.

If/when signaling turns out obvious you can solder directly board to board using same 2.54mm header pins. If not can just snip them away.

https://github.com/raszpl/sigrok-disk FM/MFM/RLL decoder
https://github.com/raszpl/FIC-486-GAC-2-Cache-Module (AT&T Globalyst)
https://github.com/raszpl/386RC-16 ram board
https://github.com/raszpl/Zenith_ZBIOS Zenith Z-386 MFM-300 ZBIOS disassembly

Reply 8 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote on 2026-11-06, 21:15:
2x cheapest standard pinout pico 1/2. You will also need 2x usb cables so either official picos + mini usb, or clones + USB 3 ca […]
Show full quote

2x cheapest standard pinout pico 1/2. You will also need 2x usb cables so either official picos + mini usb, or clones + USB 3 cables, whatever you already have.

J25 connector is imo unobtanium, looks very close to 50 Pin Mini D ribbon (used for laptop printer port sometimes?)
https://www.ebay.com/itm/195977201948?var=495835011357 https://www.ebay.com/itm/235271973398?var=535302794732
pin pitch matches, but wont mate unless original connector plastic case gets snipped away.
Cant find its matching pair in pcb version, only cable.
Ideal would be finding right angled PCI (old school, not pcie) connector, but those dont seem to exist either. How is the space above it in the case? is there space to plug card perpendicularly or is there disk, metal or something else above it? if there is space you could totally solder cut to size PCI connector, then potential ram board could be plugged into that.

Personally I would desolder J25 connector and
-directly solder cables to two picos for capturing
or
-2.54mm single row headers. Those wont fit soldered normally next to each other, have to solder one row and then pull off plastic before soldering next one. Pulling off plastic will be VERY annoying. Finally connect to picos with short dupond cables.

If/when signaling turns out obvious you can solder directly board to board using same 2.54mm header pins. If not can just snip them away.

Copy that. Give me a week or so to acquire the gear, and I'll get to do the things. Could we go the route of soldering to the posts on the bottom of the board to avoid removing the header?

There's about 5.5" of space between the board and the top of the case, and another inch below the board to the bottom of the chassis.

BIOS dump link is in your PM.

Thanks so much for the assistance here. Really kind!!

Reply 9 of 18, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++
ralphmaif wrote on 2026-11-06, 22:09:

Could we go the route of soldering to the posts on the bottom of the board to avoid removing the header?

You do you 😀 imo desoldering would be much easier. Problem with back is you cant just solder long dangling wires as this is either CPU or RAM bus and those will create EMI and make computer crashy/not booting. For probing to make sense computer needs to run normally, ideally running test programs poking at ram.

Realistically you will never plug anything into it, maybe that ebay '50 Pin MDR' plug (https://www.aliexpress.com/i/2251832762922789.html) after cutting half of the connector off. ...hmmm, wait - pulling that plug out of metal shroud would work! didnt think of that at first 😀 Might need some arts and crafts, hot glue and cardboard, to make it align 😀 That would be much better than soldering wires at the back and in case of success connector can be later reused.

ralphmaif wrote on 2026-11-06, 22:09:

BIOS dump link is in your PM.

failed dump, its all FFs and 7Fs
try this ai slop https://drive.google.com/u/0/uc?id=1RVtJVhDKW … export=download https://www.youtube.com/watch?v=FAxQU07O4E4 bonus points afaik it also dumps CMOS

https://github.com/raszpl/sigrok-disk FM/MFM/RLL decoder
https://github.com/raszpl/FIC-486-GAC-2-Cache-Module (AT&T Globalyst)
https://github.com/raszpl/386RC-16 ram board
https://github.com/raszpl/Zenith_ZBIOS Zenith Z-386 MFM-300 ZBIOS disassembly

Reply 10 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote on 2026-11-07, 01:01:
You do you :) imo desoldering would be much easier. Problem with back is you cant just solder long dangling wires as this is eit […]
Show full quote
ralphmaif wrote on 2026-11-06, 22:09:

Could we go the route of soldering to the posts on the bottom of the board to avoid removing the header?

You do you 😀 imo desoldering would be much easier. Problem with back is you cant just solder long dangling wires as this is either CPU or RAM bus and those will create EMI and make computer crashy/not booting. For probing to make sense computer needs to run normally, ideally running test programs poking at ram.

Realistically you will never plug anything into it, maybe that ebay '50 Pin MDR' plug (https://www.aliexpress.com/i/2251832762922789.html) after cutting half of the connector off. ...hmmm, wait - pulling that plug out of metal shroud would work! didnt think of that at first 😀 Might need some arts and crafts, hot glue and cardboard, to make it align 😀 That would be much better than soldering wires at the back and in case of success connector can be later reused.

ralphmaif wrote on 2026-11-06, 22:09:

BIOS dump link is in your PM.

failed dump, its all FFs and 7Fs
try this ai slop https://drive.google.com/u/0/uc?id=1RVtJVhDKW … export=download https://www.youtube.com/watch?v=FAxQU07O4E4 bonus points afaik it also dumps CMOS

Yeah, that trick with the MDR might be an option and is worth trying out. I'll probably just need to get over myself and try desoldering to do it right.

Bummer on the BIOS dump! I'll give it another go with the link you sent.

Reply 11 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote on 2026-11-07, 01:01:
You do you :) imo desoldering would be much easier. Problem with back is you cant just solder long dangling wires as this is eit […]
Show full quote
ralphmaif wrote on 2026-11-06, 22:09:

Could we go the route of soldering to the posts on the bottom of the board to avoid removing the header?

You do you 😀 imo desoldering would be much easier. Problem with back is you cant just solder long dangling wires as this is either CPU or RAM bus and those will create EMI and make computer crashy/not booting. For probing to make sense computer needs to run normally, ideally running test programs poking at ram.

Realistically you will never plug anything into it, maybe that ebay '50 Pin MDR' plug (https://www.aliexpress.com/i/2251832762922789.html) after cutting half of the connector off. ...hmmm, wait - pulling that plug out of metal shroud would work! didnt think of that at first 😀 Might need some arts and crafts, hot glue and cardboard, to make it align 😀 That would be much better than soldering wires at the back and in case of success connector can be later reused.

ralphmaif wrote on 2026-11-06, 22:09:

BIOS dump link is in your PM.

failed dump, its all FFs and 7Fs
try this ai slop https://drive.google.com/u/0/uc?id=1RVtJVhDKW … export=download https://www.youtube.com/watch?v=FAxQU07O4E4 bonus points afaik it also dumps CMOS

I don't think desoldering is necessary now! 😎

Found the connector on eBay, and it seems to fit perfectly into the slot on the board, leaving 50 exposed pins to probe with the pico2 boards.

The attachment PXL_20261112_223650945.MACRO_FOCUS.MP-1.jpg is no longer available
The attachment PXL_20261112_224523411.MACRO_FOCUS.jpg is no longer available
The attachment PXL_20261112_223630400.MP-1.jpg is no longer available

Reply 12 of 18, by rmay635703

User metadata
Rank l33t
Rank
l33t

The ram soldered to the board appears to be very low density.
But the code should be 4 megabit with 20 chips thats a fair bit of ram, not a megabyte

If the onboard memory controller can handle it and the memory dipps come in a higher density it's possible you could desolder all that mess.

Would need tech and diagnostic manuals to validate first

Reply 13 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie

I think it's 2MB of RAM in aggregate between what's soldered to the front and to the back of the motherboard.

Reply 14 of 18, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++
rmay635703 wrote on Yesterday, 23:16:

The ram soldered to the board appears to be very low density.

in 1988 (year this design came out) it was highest density 😀 4Mbit afaik came to market in 1989

rmay635703 wrote on Yesterday, 23:16:

But the code should be 4 megabit with 20 chips thats a fair bit of ram, not a megabyte

20x 1Mbit https://www.datasheetarchive.com/datasheet/M5 … 41000&version=1
using up all available Address lines. Next up would be 16M x 1 and those have 4 more pins and require extra A10-11, or bodging 1M x 4 but no extra RAS for those.

rmay635703 wrote on Yesterday, 23:16:

If the onboard memory controller can handle it

thats the problem, it looks to be implemented using Gate Arrays and is undocumented

ralphmaif wrote on 2026-11-06, 22:09:

I don't think desoldering is necessary now! 😎
Found the connector on eBay, and it seems to fit perfectly into the slot on the board, leaving 50 exposed pins to probe with the pico2 boards.

👍 Nice. I think I looked at those but didnt think it would fit due to sides, glad I was wrong 😀

https://github.com/gusmanb/logicanalyzer/blob … /Schematic1.jpg
first pico pin 1 and 2 linked, and connected to pin 2 on second pico for chained triggering. Confession time - I havent used their software with more than one pico so no idea how well it handles chaining, looking at Multidevice menu it does seem pretty automagical once you setup which one is master.

edit: IMPORTANT, before soldering picos to connector pins measure every pin on running PC to check if there isnt 12V or even worse -5V -12V on any of them! That would be an unfunny surprise 😀

We can start with one max size capture while running hmm this: save to walk.txt

a 100
push ds
mov cx, 10
xor ax, ax
mov ds, ax
xor si, si
mov bl, [si]
inc si
jnz 010a
add ax, 1000
jnc 106
loop 104
pop ds
int 20

r cx
17
n walk1m.com
w
q

mov cx, 10 <- tis 10 is 16 repeats of teh full 1MB read loop, can change to longer if it happens too fast
then debug.exe < walk.txt
then run walk1m.com , click capture (max speed, max buffer size) and cross fingers we get something that makes sense (I hope pico2s will be able to capture without buffers)
Result should be CPU reading this code wherever it was loaded interleaved with reads of ever increasing addresses. pico buffers are small so we will only capture a small fragment of this, but should give an idea of what we are dealing with.

https://github.com/raszpl/sigrok-disk FM/MFM/RLL decoder
https://github.com/raszpl/FIC-486-GAC-2-Cache-Module (AT&T Globalyst)
https://github.com/raszpl/386RC-16 ram board
https://github.com/raszpl/Zenith_ZBIOS Zenith Z-386 MFM-300 ZBIOS disassembly

Reply 15 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote:

Confession time - I havent used their software with more than one pico so no idea how well it handles chaining, looking at Multidevice menu it does seem pretty automagical once you setup which one is master.

While we are confessing, I will 💯 need a lot of guidance on all this 🤣🤣

I'll look up and pull down the app needed, test connectivity to the picos, and ensure they chain successfully.

Beyond that, I'll be like a 5yo at the DMV. 😎

Reply 16 of 18, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++

1 draw yourself a map of J25 pins on the plug you got. A1-26 B1-25 so you know whats what 😀
2 turned off: buzz in meter continuity mode all pins against PC ground, 5V supply, 12V -5V -12V supplies. Note pins that beep.
3 turned ON measure voltages on all the pins on connector, be careful not to short anything with multimeter 😀 note any pins above 5V

if everything is <= 5V
4 connect just one pico for now, ground to ground, couple data lanes to non ground/power pins and run few test captures

https://github.com/raszpl/sigrok-disk FM/MFM/RLL decoder
https://github.com/raszpl/FIC-486-GAC-2-Cache-Module (AT&T Globalyst)
https://github.com/raszpl/386RC-16 ram board
https://github.com/raszpl/Zenith_ZBIOS Zenith Z-386 MFM-300 ZBIOS disassembly

Reply 17 of 18, by Intel486dx33

User metadata
Rank l33t++
Rank
l33t++

We use to have an American car dodge colt that had a Japanese engine made by Mitsubishi
It was my father’s daily driver to work.
Thats why he bought it was because he wanted a small
Commuter car for work and around town back in the 1970s
All our other vehicles were BIG American cars with V8 engines .
The Mitsubishi had a small 4 cylinder engine.
But it was very reliable and good on gas.
I remember I learned how to drive in it too.
It had a 4 speed stick shift transmission .

Reply 18 of 18, by ralphmaif

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote on Today, 03:39:
1 draw yourself a map of J25 pins on the plug you got. A1-26 B1-25 so you know whats what :) 2 turned off: buzz in meter continu […]
Show full quote

1 draw yourself a map of J25 pins on the plug you got. A1-26 B1-25 so you know whats what 😀
2 turned off: buzz in meter continuity mode all pins against PC ground, 5V supply, 12V -5V -12V supplies. Note pins that beep.
3 turned ON measure voltages on all the pins on connector, be careful not to short anything with multimeter 😀 note any pins above 5V

if everything is <= 5V
4 connect just one pico for now, ground to ground, couple data lanes to non ground/power pins and run few test captures

Got it, that's helpful -- thank you!

I'm going to pick up some wire for the harness, header pins, and a breakout board for safety's sake. This looks like what I need to do.

EDIT: looks like I can just get Pico's breakout board with the headers already there. That saves some time!