Supercomputing adventures with a MegaST/4

[dml]

This is another weird sidequest in the sidequest tree and decided to document parts of it here, for better or worse.

I already had some success with programming the Transputer on the ATW800/2 and taking advantage of extra Transputers when added to the optional slots. So fully loaded, that's a maximum of three - already impressive for something that (nearly) fits inside the case, give or take a battery compartment here and there - but those paying attention will notice a header on the board exposing an external link...

The signals for a Transputer link are intentionally simple - a few 'global' control lines (reset, !error, analyse) and a serial pair.


So with those details in mind, it was not long before I ran across this thing on eWasteBay...

board-populated.jpg (715.15 KiB) Viewed 151 times

Of course this is not going to be 'plug and play' - it is a twin VME thing from a much bigger system and has onboard logic and unknown routing to sort out - but just the kind of thing that would be worth solving for the amusement of silly hardware expansions.

What's the worst that can happen?




I have already figured out some things, will post more soon...

[exxos]

I bet it could run doom twice!

[Steve]

I'm super curious as to what kind of things you're going to be able to do with all this power

[dml]

If I can make it actually work I'll probably figure out something to do with it still a few hurdles before I get there.

[dml]

The first thing I did after getting hold of the board was to see if the 8x TRAM-like daughter boards are actually TRAMs at all or if they are custom things which could create special problems figuring out signals etc.

The Transputers have a standard comms protocol but that doesn't mean the boards they are attached to are going to keep to a reference design.

There is local working memory present in zip package format on each board but there are also custom PALS on the boards and on the main carrier.

The most suspect detail is the fact these boards have 20-pin footprint with 10 on each edge while standard TRAMs have a 16-pin footprint with 8 on each edge. They are also 50% wider than a typical single TRAM. That's the first thing needing a closer look.

protram1.jpg (230.91 KiB) Viewed 116 times

protram2.jpg (473.81 KiB) Viewed 116 times

[dml]

With a bit of patience and a DMM I was able to confirm most of the expected signals are exposed on the inner 8 pins on each row, matching the standard TRAM footprint.

The 4 outer pins are 3 additional GND pins and one additional VCC (I was marking connections out with a sharpie as seen above :p )

So they appeared to be TRAMs with some extra power connections. A bit strange but ok.

I thought about all this for about 18 seconds then decided to stick one in my ATW800/2 to see what happens....

pt1.jpg (441.87 KiB) Viewed 107 times

pt2.jpg (744.88 KiB) Viewed 107 times

[Darklord]

If I can make it actually work I'll probably figure out something to do with it still a few hurdles before I get there.

"With great power comes great responsibility!"

[alexh]

Great find. Several employees here are former inmos engineers who say they have old boards in their attics and I've been begging them to bring them in / loan or sell them to me. So far just a few pictures no hands on

[rubber_jonnie]

Very cool find that, nice.

Just wondering, the boards seem to have sockets on top:

Screenshot 2026-08-27 at 11.57.54.png (491.17 KiB) Viewed 49 times

So is it possible to stack the boards one on top of the other to improve density?

If not then why have the sockets at all?

Also, doe this mean you're going to become a TRAM dealer

[Steve]

So is it possible to stack the boards one on top of the other to improve density?
If not then why have the sockets at all?

As far as I know, the cannot stack TRAMS. I *think* the sockets are so you can probe signals easily.