I took a look at an early PC clone from 1987 that came out of e-waste, and the first obvious problem turned out to be memory. The case was rough on all sides, but the front panel was intact, and around back it had the usual clone-era mix of ports and cards. Once the cover came off, it was clearly an XT-class machine with a turbo button, a full-height floppy, an MFM hard drive, and three expansion cards already installed.
With the cards out, the hardware was easier to identify. The video card was a Paradise Basic EGA card, one card handled floppy, serial, parallel, and game port duties, and the hard drive controller was a Seagate ST10 for the ST-225 MFM drive. The combo card also had a barrel battery that had started leaking, so that went on the list for later. On the motherboard itself I found a Siemens 8088, four RAM banks, and a BIOS ROM labeled with DTK markings.
Before powering anything, I went ahead and checked the power supply and motherboard for shorts. I tested the AC input side, then the 5V and 12V output rails, and everything looked reasonable. With a spare hard drive connected as a load, the supply came up and the voltages checked out, and once the system was fully assembled again the machine actually posted on the first real test.
The floppy drive needed a quick sanity check because the latch handle wouldn’t turn until a disk was inserted, which turned out to be normal for that mechanism. I also showed a simple fix for loose disk labels using clear crafter’s tape. With the hard drive connected, the system recognized the ST-225 correctly and, after a long spin-up and boot process, it loaded DOS from the original MFM disk.
At that point I backed the drive up using an XT-IDE style CompactFlash card. I configured the CF card as C: and the MFM drive as D:, then started copying files across. That worked for a bit, but the old hard drive quickly started throwing sector read errors, and there was enough corruption that it wasn’t going to be trustworthy for regular use.
The memory issue showed up once I paid closer attention to the POST count. The board had enough chips installed for 640K, with a mix of 41256 and 4164 DRAMs across the four banks, but the machine only reported 128K in one configuration and wouldn’t complete a full count in others. I spent time working through jumper settings and eventually found better documentation, including a motherboard manual and jumper references that matched the board.
Rather than keep guessing, I burned a modern replacement BIOS ROM and tried that in place of the original chip. The first attempt stalled because the EPROM programmer needed a firmware update, and I also had to erase the EPROM properly with a UV eraser before it would program successfully. Once installed, the new BIOS gave much better diagnostics and reported a parity error at a specific memory address instead of just stopping during the count.
I then swapped the RAM chips around within the banks and tested again. The parity error moved with the chips, which narrowed the fault down to one DRAM chip rather than the surrounding logic. I ordered replacement memory, and the machine ended the session still running with the new BIOS and a confirmed bad RAM chip to fix next.
— Aaron
Links
- Super PC/XT Turbo BIOS
- Turbo XT Motherboard Manual
- “Fakebit World” by Malmen
- “T.I.M.E.” by Emil Rottmayer
- “Pathfinder” by Kubbi
Timestamps
00:00 – Intro
02:31 – First Look
06:26 – Disassembly
09:14 – Testing
13:30 – Backup
16:30 – RAM Problems
21:36 – Making A New BIOS ROM
27:52 – Testing The New BIOS
30:20 – Wrap Up