Atari Transputer Workstation | Hacker News Reader

Atari Transputer Workstation

en.wikipedia.org

63 pointsbane28 comments

One fine morning I opened the San Jose Mercury News to see that Atari had announced a new Transputer-based workstation.

It was especially exciting because I was working for Atari Corp, and our little engineering team (of maybe 20 hardware and software people) had no inkling we'd been working on it!

This was a project shot out without permission by a group in the UK. They hadn't bothered to tell us.

[I rather liked the Transputer architecture. Bizarre, but very interesting]

Funny because I worked at Inmos at the time and we like "ATW, WTF??"

Wow - this brings back memories... When I was an undergrad, I took a summer job at Perihelion writing demos for the ATW.

From what I recall, things weren't going great, and most of the staff was grumbling about the Atari side of the machine. Since Atari was where the latest funding had come from, there was a strong desire by management to show how the Atari hardware (display, disk interface, etc) was essential to the whole set-up. Whereas from an engineering point of view the Atari front-end was pretty much dead weight (vs. the Transputers that hung off the back).

Less than a year later, Perihelion went into liquidation, and a friend and I bought the contents of the building (boxes like 'Misc Electronics #14' and 'Papers #5'). We spent the following summer holidays using the customer list (i.e. the Papers) to flog the remaining Transputers (in the Misc Electronics)... Fun (though slightly morbid) times!

Ah. The transputer. Did my doctoral thesis using CSP converted to occam for the transputer. Fun times.

There's some of my occam code here: http://blog.jgc.org/2009/08/in-which-i-resurrect-13-year-old...

Oxford had some transputers from INMOS. The story I heard was that they'd been manufactured with the die in the wrong way round meaning the pinouts were all messed up. So a bunch (16 IIRC?) had been wirewrapped together to make some hideous but working machine.

T212s I seem to remember.

I did some work on a transputer at Tao --- we had a nine-node demo system; eight transputer nodes, and one Pentium node, running as an asymmetric multiprocessing cluster. The demo program drew Mandelbrot images by farming out scanlines to each processor node. Unfortunately, the Pentium was so much faster than the transputers that by the time each transputer node had rendered one scanline, the Pentium had done all the rest...

Later, the transputer metamorphosed into the ST20, which was sold as a single-processor embedded microcontroller, used a lot in the TV and set-top-box market. It actually worked pretty well.

One exciting thing we discovered: on at least on transputer system (and the details are from memory, so possibly incorrect here), if you asked it to do a shift right by 0 bits, the processor would hang. We eventually decided that the microcode that implemented this was doing:

    while (--shift)
      value >>= 1;

So, if you passed in 0, in the first iteration the shift rolled round to 0xffffffff, and kept going; with interrupts off. IIRC we calculated that it would probably take about a fortnight to run. I don't think we ever tried it to see, though.

Atari has some fun computers during this era that were overshadowed by the (arguably better) Amiga. The Falcon was a favorite of musicians.

Inmos sold a T-400 board when I was in college. It was another example of a stack machine. Occam was an interesting programming language. I still have the book for that.

The ST was another musician's favorite, because of the MIDI ports.

Arguably the original musicians favorite. (From 1985 until Atari Corporation folded up in 96.)

The Falcon came later and was unsuccessful compared to the ST.

> "The Falcon came later and was unsuccessful compared to the ST."

Agreed, though the Falcon was a big step up in terms of audio hardware due to the addition of the DSP.

To hear the difference the DSP made, search on YouTube for 'ACE Tracker Falcon'. This is my personal favourite:

http://youtu.be/xxXCGSJ9nT0

I like this one too, and it's probably a better technical showcase:

http://youtu.be/kXzMkO83jA8

In comparison, here's a decent ST chiptune. Musically it's still enjoyable to listen to, but the difference in audio hardware compared to the Falcon should be fairly clear:

http://youtu.be/RnsMz-vZFHc

I never did and never will understand why an ATARI ST family, meant to compete directly with Amigas, had a three channel sound chip on the level of the C=64 SID. What were they thinking?

The AT&T DSP was light years ahead of the rest of personal computers' sound capabilities, and is in fact still competitive today. Amazing piece of work.

The Atari ST was released (in 1984) before the Amiga 1000 (in 1985) and was designed around being a low cost 16-bit computer. The Amiga 500, which was a much closer competitor to the Atari ST in terms of target market, was released in 1987. Back at that time, 2 to 3 years was a big jump in terms of the processing performance you could get for your money.

Sadly (I say this as an Amiga fan), the Paula sound chip was never upgraded, I believe it's the same in all Amigas, even the AGA ones. In some ways it's a testament to the capabilities of the Paula sound chip, but it's also just another example of Commodore squandering their technological lead. There were plans to go down the DSP route with the A3000+, but that machine never materialised, leaving the Atari Falcon as the best sounding 16-bit computer (IMO, of course).

The C64 SID is generally considered the best soundchip made.

But with a DSP you can do not only sound, but this (Atari Falcon demo) https://www.youtube.com/watch?v=_cKRZ8QgH5o

> "The C64 SID is generally considered the best soundchip made."

Best for the time, I'd agree, but I'd suggest it's been surpassed since. Of course the SID still has its charm, when I say surpassed I mean in terms of flexibility.

IIRC the designer of the SID went on to design some of the early high-end PC sound cards.

It was a quantum leap, just like the iPhone was to phonemarket, nothing before or after comes close.

And Bob Yannes went on to found Ensoniq after designing the SID at Commodore.

> "or after comes close."

What do you base that on?

I refer to the leap in it self, not the actual chip with that comment. There is for sure better things today :)

It wasn't that it was unsuccessful, it was that so few were made before ATARI folded. There weren't enough to go around.

Another company made the Falcon after Atari folded.

I saw an Amiga transputer ray tracing in person. It was way beyond most desktop performance at the time.

http://www.amigahistory.plus.com/prototypes/transputer.html

Ah yes, the transputer. Such an interesting, if somewhat flawed, concept. Inspired a lot of interesting things. Given how cheap small FPGAs are these days, you could easily build a transputer "board" which was probably faster and cheaper than the original boards were. Put together a few dozen and start writing parallel software for fun (and certainly no profit :-)

There is a "startup" XMos co-created by chief architect from Inmos (who introduced transputers). XMos microcontrollers share a lot with transputers design and are really fun to play with.

Cores are RISC, there is at least 8 of them, but you can link multiple uC into a single system. Their variation on C, xC, is using channels and guards as in occam transputer language.

You can get easily a cheap dev-kit [1] and play with the tech.

[1]: https://www.xmos.com/support/boards?product=17441

I couldn't find anything on how much RAM it has.

It's based on the XS1. For the follow-on, the xCORE-200, the flyer mentions twice the memory. Since the smallest xCORE-200 has 256 KiB, I'm guessing the XS1 might have 128 KiB.

There's a dev kit based on the xCORE-200, for 10 X the price ($150), but it includes a 1 GbE interface which makes it a lot more interesting IMO.

Interesting architecture. Makes a ton of sense for embedded apps (which obviously was the point).

The good old days, when hardware vendors innovated rather than sticking an Intel reference design inside a slightly different case.

An interesting read:

It's the story of the UK's attempt to become a force in the emerging microprocessor industry of the 1970s, told by the man who was the driving force behind it.

In the first part Iann Barron describes how Inmos came into being and the thinking behind the revolutionary transputer.

In the second part, which starts at the point when Inmos had just secured its second slice of Government funding, Iann Barron charts the company's decline and fall.

http://www.transputer.net/fbooks/iandt/inmos_transputer.html

Wow - I had no idea that Atari was even in this space. Talking about Transputers, I spent all of 1989 working on a Meiko computing surface with 40 T800 Transputers. Had a glorious time writing code in Occam for the traveling salesman problem.

Yet another wonderful machine killed by the wintel standard...

I wonder if MESS cam emulate it and if someone, somewhere, preserved enough of HeliOS to explore it.

Edit: it looks like someone did http://www.classiccmp.org/transputer/helios.htm

Reply on news.ycombinator.com