An S-100 80486 CPU Board

What university was this? I've heard IITs are pretty good.

tiny things with weak intel atom CPUs (same as you'd find in a cheap 7 inch tablet from 3 years ago) capable of the full i686 instruction set are not much more than $25...

The support chipset required is quite extensive. There is a market for them, but in the "industrial PC" form factor and not cheap because the quantity is so small.

(If you want to boot DOS then in addition to the cpu and DRAM you need a whole BIOS and emulated keyboard, interrupt controller, disc controller, and VGA controller)

Interesting aside - the Intel Quark line was seemingly based on the 486[1], albeit with a substantial process change (600 nm to 90 nm) and some modifications to support the Pentium ISA. So the 486 didn't just live on to be sold to industrial users as-is, it evolved and reappeared in some embedded systems running at a not too shabby 400MHz (originally the 486 ran at 25-100MHz). On a similar note, I think earlier Atom chips were largely based on the original Pentium (not Pro, not II or III) but maybe take that with a pinch of salt as I don't 100% remember where I read it :-)

[1] - https://semiaccurate.com/2013/10/28/intel-talks-little-quark

Intel attempted a GPU that was effectively a cluster of x86 cores with a massive process shrink.

https://en.wikipedia.org/wiki/Larrabee_(microarchitecture)#D...

Not quite as old as 486s, but damn close: The cores were all based on the P54C Pentium, in order execution only type processing cores.

I remember hearing about an Intel Dedicated GPU that was supposed to be all 486 cores, but my ability to Google it seems to be failing me so I suspect it was probably my conflating an IOE type cluster of cores of the Larrabee with the 486.

It appears that the Larrabee came in variants of 10 (at 65nm) 40 (at 45nm), 48 (at 32nm) cores on the dGPU.

https://www.anandtech.com/show/2580/6

In the end, Intel expected people to be able to run x86 code on these dGPUs...

The currently existing 486+ SBCs all list for approximately 200 GBP on eBay, unless you are lucky enough to grab one that's under valued (which I did a while back!).

The downside of these, are that they are designed to be used inside industrial equipment, typically have an ISA edge connecter, and are missing all the ribbon cables to connect them the peripherals (and power cables). So they become a DIY project, which is fun for some people, but not good for those who don't like building hardware. Additionally almost all industrial SBCs have no sound hardware on board (fixable by using an ISA backplane and another card, or via a PC/104 adapter).

The alternative is the old VIA Eden mini-ITX boards (mentioned elsewhere in this thread), if you get the right one, they are faster than a 486 but not TOO fast, AND come with SoundBlaster/Ad-Lib emulation on-board. I've got one as tiny DOS gaming PC and it works VERY well.

There's a 486 open core under a BSD-ish license: https://github.com/alfikpl/ao486 I have this on an FPGA and it can run DOOM albeit a bit slowly.

If I was going to play about on an architecture of my choice, I wouldn't choose the 486, I'd go for something more exotic. Perhaps this SPARC-powered kit: https://navspark.mybigcommerce.com/

A quick google indicates there are a small number of MIPS-powered Arduino-like kits out there. Looks like there's no equivalent for the SuperH, sadly.

The S-100 bus only outputs +8V, -18V and +18V. The idea is that the board itself will convert this voltage to what the components on the board needs to operate. The newly designed board have a newer voltage regulator that outputs 5V and the optional 3.3V from the higher voltage provided by the S-100 bus. You always need to do a voltage conversion on a S-100 card.

I've been working on something like that (actually with a Pentium), but the biggest problem is finding an FPGA with enough pins to connect to the CPU, RAM, peripherals, etc. Maybe I need to use a cheaper FPGA to mux down the 64-bit-wide data bus (which this one seems to do with a CPLD). No matter how you cut it, it quickly becomes a pretty hairy design.

since I can't edit the comment anymore by "bus voltage conversion" I meant signal level conversion.

A 486DX-50 was my high school graduation present, first non-68k non-6502 machine, my intro to the Intel world, and my first Linux machine (pre 1.0, even). I used it from 1991 to late 1997, and it was the machine that I did my first web development on, learned Python and C on, wrote my papers for my philosophy BA, wrote bad poetry on, and so on and so on.

I have never had a love for the Intel architecture but I have to admit that machine was amazing.

It's the website for the s-100 bus that the processors it talks about works with,(i.e, 80486 intel processors). It's a lot of nostalgic 1989 bus verbiage for uber geeks and engineers who make that stuff.

Appears the 80486 can be connected to the S100 bus directly given ability to read from adjustable size databus.

This means effectively an x86 style cpu can be part of a simple computer design

The S100 bus was the first standard microcomputer bus[1]. It was based off of a passive mainboard system, so the CPU was just another card that plugged into it. The overwhelming majority of microcomputers that shipped with CP/M (as well as many that shipped with other operating systems) were based off of the S100 bus.

While there was never a commercial S100 board with a 486 on it, this article shows how it is fairly easy to design and build one.

S100 predates the PC bus, which largely replaced it. The AT (a.k.a ISA bus) was a 16-bit version of the 8-bit PC bus. IBM tried to get everyone to switch to the MCA bus, but the high royalties they charged meant that pretty much everone stuck with the AT bus until PCI came out. In its server form of PCI-X, PCI maxed out at 64-bits and 133MHz (though 32-bits and 66MHz was what you would see in a PC), and was replaced by PCIe, which is still used today.

Graphics cards were treated specially at several points along the way, with the VL-bus, and AGP basically being ISA and PCI respectively with dedicated higher-throughput buses to avoid having to use the more expensive MCA and PCI-X when the graphics card was the sole device in the system that needed higher bandwidth.

The 386 was a 1000-1500 nano meter process, and even at that scale took 1/6th of the die space used by an RTX3090 today on 8nm process.

A 386 draws almost no power, much less than an Atom. At the end of the 386's market life, which I should point out ran for 22 years, it was being used in handheld devices that ran off 1 alkaline battery for months at a time, and those had radios in them.

Obviously you get more bang from something like an Atom but I don't think the 386 had any power saving tricks to learn from it.

Intel tried this with the Quark core, which was discontinued in 2019:

Compare the Quark Core Block Diagram: https://www.intel.com/content/dam/support/us/en/documents/pr...

with that of the 486 (Figure 3-2): http://datasheets.chipdb.org/Intel/x86/486/manuals/27302101....

20MB is microscopic for a 486! I had an 8088 with a 10MB drive and a 386SX laptop with a 120MB 2.5" drive, and neither was particularly special!

Let's dig up those Token Ring cards!

Why would you write this? He's building and testing boards with those processors today.