IBM 5170

Manufacturer

IBM

Model

5170

Kind

Motherboard

Year

1985

Condition

Partially working

Source

RC Table Sale

Acquired

2026-07-26

Reference

https://theretroweb.com/motherboards/s/ibm-at-type-5170,-type-1-6133920

Notes

RAM failure, SuperSoft reports protected mode fault.

Chipset

discrete

CPU family

286-class

Form factor

AT

Slots

2× 8-bit ISA, 6× 16-bit ISA

BIOS

IBM AT (1984 variant)

Ports

DIN keyboard

Not fitted in anything.

Part of IBM 5170

• Swap U43 and U44 (first 256K parity chips) for known-good chips and re-run the IBM POST.
• If it persists, get sheet 10 and trace the parity logic from MDPIN/MDPOUT to U43 and U44.
• Inspect the two new DRAM sockets and the U62 rework for bridges, lifted pads and damaged tracks.
• Note which positions have the new sockets.
• Check the SuperSoft Landmark AT manual for what "U126 8042 parity detected" actually tests.
• Get a probe ROM that shows parity check (61h bit 7) and I/O check (61h bit 6) separately, with a parity test per bank.
• Recheck what the POST card shows in the 201 loop (C0 against DD).
• Re-run the IBM POST and SuperSoft after each change.

No files for this one yet.

Acquired 2026-07-26 from RC Table Sale · held 2 months

2026-09-29

wanted for IBM 5170 (RH-V7QP)

2026-09-29

note This one is becoming fun, but I need a break. The IBM BIOS was stopping early in the POST with a 201 memory error. Interestingly, it was showing failed bit 00000000 with FF00. A very round and unlikely number. XTRAMTEST found bad RAM in bank 1, bits 4 and 5. SuperSoft Diagnostics failed at the protected mode CPU test, then hung on its memory test. A different cpu made no difference. I got Fable to build a 16-bit version of XTRAMTEST which after mating 4164 DRAMs to make 41128s and swapping them out came up clean, so basic byte and plain word access both worked. The IBM BIOS writes a word of 0000, then FF to the high byte, then reads the word back and expects FF00. I was getting 0000, so that proved the high byte write tongue twister wasn't working. A Fable-written byte/word probe ROM based on XTRAMTEST showed every byte access to a high byte landed 100h higher than it should. A byte written to 2469 turned up in the word at 2568. So A8 was being forced high whenever A0 was 1. Word and low byte reads were fine. That also explained March-U, two high byte addresses were ending up on the same cell while other tests passed. Hurrah for XTRAMTEST. I tested the 41128s with the ram tester and they all came up good, even the ones that I'd made. The 5170 schematics show MA6 comes through U62, a 74S51 that passes SA8 on normal cycles and SA0 during refresh, into U54 (74F158). U62 was letting SA0 through on normal cycles as well, so MA6 carried SA8 OR SA0 at row time. I swapped U62 for a tested one that I had scrounged from a scrap board. Free, and feeling chuffed about my scavenging old chips paid off again, and with XTRAMTEST passing with flying colours, I was on a high. HOWEVER..... SuperSoft now stops at U126 8042 PARITY DETECTED and the error counter flies like the £ numbers on a diesel pump. This passed before the S51 swapout, so another fault to find. My soldering looks good but I am going to have to do the simple stuff like running over it with a multimeter beeper first. I'm a good few days into this board now though, and WAY beyond my own concentration span, so it's being filed under "projects for another day". 

2026-09-29

cropped 2 photos

2026-09-29

tuned a photo

2026-09-29

deleted a photo

2026-09-29

rotated a photo

2026-09-29

added 5 photo(s)

2026-09-29

deleted a photo

2026-09-29

added 1 file(s)

2026-08-01

Chipset: discreet → discrete

2026-07-29

flagged a photo as a reference image

2026-07-27

duplicated to RH-7WWA

2026-07-27

fetched a photo from the reference

2026-07-27

created