How to build a compact x86 / EISA / ISA architecture PC

Question

This question is based on a previous question I asked (x86 based Retro/Nostalgia PC the size of a Raspberry Pi), and also I guess to some extent it fits in with this question regarding ISA bus on newer computers

I'm still on the hunt for hardware to create a compact x86 based computer, but based on dated technologies such as 486/586 and ISA expansion for use with legacy graphics and sound cards (SB16).

I have researched:

• SBC (Single Board Computer)
• Nano-ITX, Pico-ITX
• ISA/EISA Back Plane

Since these technologies are rather outdated, building a tiny equivalent would not be a realistic investment for any technology company nowadays (except for those in the market where legacy embedded systems are being used).

So one idea I have come up with to build a compact x86 based PC is to use a small ISA backplane (maybe just 2/3 ISA slots).

Slot configuration:

1. Single Board Computer (Legacy, 486/586, RAM, I/O, maybe graphics)
2. Graphics card (if the SBC does not have this on-board), or additional I/O (PS/2, RS-232)
3. Original Sound Blaster 16

My first concern here is that I don't know how ISA backplanes work. I fully understand traditional PC architecture, where the motherboard controls I/O to the expansion slots, however in this configuration, since the motherboard is essentially a card on the backplane, I'm not sure if/how it would be able to communicate with other hardware (graphics card / sound card) which are on separate slots on the backplane.

If anyone is knowledgeable in this area, I basically want to know if the purpose of an ISA backplane is to allow communication with other devices when used in conjunction with a SBC.

Answer 1

I looked at the market for this stuff about 8 years ago. My finding was that there were very few options, because the ISA technology was just considered way too obsolete then. My goal of getting a high end Pentium 4, but with ISA slots, was being a challenge. I also wanted to support the most RAM that is widely supported by 32-bit software, etc. I made notes at notes for a new ISA computer; in particular the charts in newcomp.htm and almomobo.htm

From what I could tell, the ISA back planes were not widely praised for high compatibility. It seemed like they were designed for industrial use, possibly for things like hardware that controlled electronic machinery by sending out simple commands. In such cases, a company cared about things like standardization (which may have required older cards to connect to other older equipment), but not necessarily speed and supreme performance. So I really aimed for support directly on a motherboard.

IIRC, my selection was the WinIP-06046. Unfortunately, the results were not very satisfying. I note this in my documented experience (in the "again.htm" file). I tried using a sound card and it could be detected, but trying to actually pipe audio data to the sound card did not result in pleasant tunes coming from speakers. This was also with a sound card that wasn't fully verified to be recently working. Still, I was left thinking that at the end that the hardware did not seem to be very compatible. It seemed like actually using a card required some sort of timing that didn't work as well with the chips that were so ridiculously fast (compared to ISA standards).

I've heard great things about DOSBox, which I saw you were unhappy with (based on a comment in the other question you hyperlinked). I just saw my friend run DOSBox a couple of days ago, and it did not handle Future Crew's Second Reality well. However, despite that imperfection, I don't think things are likely to better trying to merge modern tech (like a Raspberry Pi) with older tech (like an old ISA sound card), even with conversion hardware like a backplane board. Such conversion tech often doesn't work like we would dream it to, even if things do physically plug in.

You may also want to check out ramlimit.htm where I discuss some struggles trying to use older software on newer hardware. Some workarounds are documented, but my findings was that even those workarounds didn't always function as documented.