In this image, you can see, from the right, a 16-bit ISA slot (occupied by a controller card), an 8-bit ISA slot, four more 16-bit ISA slots, and two of these strange slots that look like a 16-bit ISA slot, except with an 86-contact extension rather than the normal 36-contact extension. These are definitely not VLB extensions. I found this page on the motherboard, which says they are some sort of "32-bit external memory card." What are these slots? Are they compatible with 8-bit or even 16-bit ISA?

This is only one of this computer's oddities. The controller card, while sporting a typical floppy connector (the solder ends of which are visible at the top right), also has a strange 34-pin connector (keyed differently from the floppy connector and visible immediately below it) labeled "winchester", as well as two 20-pin connectors (visible below the "winchester" connector) labeled "drive 0" and "drive 1". The winchester and drive 0 connectors both connect to a very (physically) large Seagate? ST-251 hard drive.

  • Expansion memory cards on the ISA bus are not really anything odd in this time frame and go all the way back to 8088 machines but this 32 bit slot is kind of an oddity. It is also a really great idea. Most of the ISA bus memory cards that I remember seeing were usually 8 bit. 386 was really the breakover point where you were able to run a useful gui interface. The 386 was intel's first offering in a 32 bit cpu which was the standard until the first 64 bit Pentium 4's started poking out in 2004. The ammount of memory addressable by a 32 bit processor was just a huge jump. Soon after this jewel Commented Jan 25, 2021 at 0:13
  • This board is also discussed on VOGONs, see vogons.org/viewtopic.php?p=844188#p844188 Commented Aug 13, 2021 at 15:41

2 Answers 2


I found this page on the motherboard, which says they are some sort of "32-bit external memory card." What are these slots?

They are exactly that, memory expansion. This is a rather early 386 board from before memory modules became a thing.

The mainboard can be fitted with 1 MiB using 256 KiBit chips (41256), so any expansion has to go on cards. With (AT) ISA being only 16 bits, using an unmodified bus would slow down the 386 considerably. Not cool for someone who just paid an upper-end price tag. So they had to come up with some way to allow 32 bit wide memory expansions. Instead of making a whole proprietary slot or pin header, they decided simply to expand the 16 bit ISA layout to 32 bit - so in some way it is like VLB. Going that way allowed using these slots for regular I/O cards as well.

Are they compatible with 8-bit or even 16-bit ISA?

I'm pretty sure they are fully compatible at least with 8 bit cards - it might not be for 16 bit cards. All educated guesswork from the visible traces.

This is only one of this computer's oddities. The controller card, [...] also has a strange 34-pin connector [...] labeled "winchester", as well as two 20-pin connectors [...] labeled "drive 0" and "drive 1". The winchester and drive 0 connectors both connect to a very (physically) large Seagate? ST-251 hard drive.

(No, the ST-251 is not a large disk drive, it's only a 5.25 and in addition half height. Those were considered quite small at the time, top notch miniaturisation)

Again, it's exactly what's written: it connects to a Winchester Drive - slang name for a hard disk - and a ST-251 is much like the prototypical 40 MiB drive of that time. The interface is called ST-506 after the first Seagate drive using it. The 34 Pin connector holds all drive control signals, while the 20 pin connector delivers data.

So, nothing special here, just the regular disk controller in pre-IDE times.

  • And as usual a far more comprehensive answer than mine, which I'll delete. Commented May 15, 2020 at 4:05
  • 1
    More info about the memory slots: vcfed.org/forum/… Commented May 15, 2020 at 4:13
  • 2
    @manassehkatz-Moving2Codidact I am sorry, that was not my intention. Yours is quite valid on its own. It points out the most important information about the controller/drive interface. Worth keeping.
    – Raffzahn
    Commented May 15, 2020 at 11:23
  • 1
    Before I found out that the 'winchester' name came from an American rifle, I assumed it had something to do with IBM Hursley, located just outside Winchester in the UK :-)
    – dave
    Commented May 15, 2020 at 11:24
  • 1
    My answer was 100% valid. But it was not super-innovative (i.e., your including the ST-506 reference is not something you wouldn't have come up with on your own - anyone who has been in the business since the 1980s would know that part) and you provide far more detail. I've had my "great" answers in the past - but that wasn't one of them. Commented May 15, 2020 at 14:05

Photos of the manual of (nearly) that board appeared on the VOGONs forum in https://www.vogons.org/viewtopic.php?p=844188#p844188 . They clearly spell out that these slots are 32-bit memory slots combined with 8-bit ISA slots. The pinout is included. A single slot has the pins equivalent to two banks of (double-sided) PS/2 SIMMs. So a board with two of those slots might support RAM equivalent to four PS/2 SIMMs. The manual is for a slighly different variant that only has one slot, though. It should be straightforward to design a passive adapter card for 2 PS/2 SIMMs (or four banks of 4 30-pin SIMMs) for that slot.

Old guesswork answer:

To add to Raffzahn's answer: I strengthen the "might not be compatible to 16-bit cards" to "are most likely not compatible to 16-bit cards". As I see the traces on the photo, there is nothing that supports the impression these slots are ISA 16-bit compatible - but also no proof they are not.

A reliable tool to learn more about that board is a continuity tester. I already probed a lot of computer hardware using continuity beepers of multimeters, and despite randomly applying open-circuit voltages of up to 3V to undefined pins, I never damaged any electronic component by it. If you happen to have a quickly responding continuity tester, you can easily probe connections to a pin by having one probe at the connector you are interested in, and sliding the other probe alongside possible pins that might be connected to your connector.

For your board, I see two likely possibilities on how the memory expansion slot is connected:

  • It might be some kind of local bus slot, i.e. directly connected to the 80386 address and data pins. Address decoding (like row/column multiplexing, determining whether the board should respond to that address) would be on the memory card in this case. You should be able to beep out address and data to the 80386 pins directly or connected through 74F245 or 74F244 buffer chips.
  • It might be some kind of 32-bit memory module connector (like the 72-pin PS/2 SIMMs) with already decoded address and RAS/CAS lines. In this case, you should be able to beep out the data pins again to the 80386 local bus (and again possibly through the buffer TTL chips), whereas the address pins and RAS/CAS pins are connected to a memory decoder, which is often part of a VLSI (the technology "very large scale integration", not necessarily the manufacturer with the same name) chip, on which you might find a datasheet in the internet.

The only thing to add to the hard drive controller information is that there were different kinds of "winchester" controllers, most prominently MFM and RLL controllers. The hard drive model number indicates it should be connected to an MFM controller, and have a capacity of 40MB. If your computer shows a 60MB hard drive, someone swapped the original MFM controller by an RLL controller, which manages to write data more densly to the drive, with only moderately higher requirements on the drive itself. This is an unsupported configuration, and while it works for some controller/drive combinations, there are numerous reports of reliability problems around the internet caused by using RLL controllers with MFM drives.

  • Yeah. I’d be absolutely shocked if the initial section was not 8-bit ISA given the trace layout. But the memory expansion part appears to have zero traces in common with the 16 bit extension connectors, so a more direct connection to the main bus seems likely, especially since it was designed for memory expansion. Commented May 16, 2020 at 13:15
  • This is in stark contrast to the other 8 bit ISA location, where there exists through holes to support the 16 bit extensions. I’d be absolutely shocked if that would not work if populated. Indeed, I’d not be surprised if the reason it was unpopulated was merely because the system shipped with a factory installed 8 bit card in that slot that they were not anticipating that users would remove. Hence no need to populate. Commented May 16, 2020 at 13:20
  • @KevinCathcart Wish I knew more about what this machine went through! This was my mom's 386 that she bought new in 1989. It must've had a video card at some point but none of us know what happened to it. The 8-bit slot is also one of two that's missing a placeholder shield, so I do agree that it was probably a factory card that was removed.
    – robbie
    Commented Jun 16, 2020 at 5:38
  • When I was just starting college I made pocket money by buying borderline obsolete components at flea markets and building low-end systems on the cheap. I remember getting a couple of boards like this and 16 bit cards didn't work in those slots. Commented Aug 16, 2021 at 17:12

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .