The other 32-Bit Acorn system ....

[gordonDrogon]

You know the one - with the 32016 CPU that never seemed to go anywhere that I'm aware of...

I'm after any information about it other than what I've found online (Chris's Acorn site, wiki. etc.) In particular any user experiences, what worked, what didn't or just general thoughts - positive, negative, whatever.

No real reason other than my own curiosity right now.... but in the past few years, as a personal project (or series of them), I've written my own retro-style 32-bit OS that runs on an SBC with 16-bit CPU and 8-bit memory system (65c816 CPU) and have been looking at the next generation hardware, as it were - I'd initially discounted ARM for a number of reasons and had looked at 68K and RISC-V but recently I've been re-visiting the old Beeb and wondering why I'm bothering with these little SBCs when I can just do a software implementation in a PiTube widget... (or at least I think I can).

So having recently acquired a lot of old Acorn hardware and was reminded of Panos when I found some old Panos disks in a box and thought.... Acorn - you know it makes sense...

I have to say that one thing that puts me off using a Beeb as the host is the filing system nomenclature. Panos used the hyphen for filename suffixes... e.g. hello-c rather than hello.c and so on. There's a lot of "muscle memory" in there re. dot, slash, double dot and so on...

Cheers,

-Gordon

[BigEd]

Never used PanOS back in the day, or for anything serious, but there are some things you can do with the ns32k flavour of PiTubeDirect - compiling hello world in about three languages, for example. And there's a digits-of-pi program written moderately recently, which fits on a floppy together with a minimal amount of PanOS scaffolding.

As I recall, this copro was used in Acorn to run SPICE, the circuit simulator, written in Fortran.

[julians]

I remember reading an article about UCL's experience (I think...) around a smallish computer lab populated with Beebs and 32016 second processors using Econet for storage that was less than positive - from what i remember, possibly incorrectly as i can't find a reference to it now, mostly around poor performance and limited graphics, although possibly from what's likely to be an unfair comparison on cost at least, to some of the networked Unix workstations of that era.

But, I wanted one from the moment I saw the reference to the '16032' second processor in the user guide and the first thing I did with a PiTube was install Panos.

Julian.

[paulb]

I remember reading an article about UCL's experience (I think...) around a smallish computer lab populated with Beebs and 32016 second processors using Econet for storage that was less than positive - from what i remember, possibly incorrectly as i can't find a reference to it now, mostly around poor performance and limited graphics, although possibly from what's likely to be an unfair comparison on cost at least, to some of the networked Unix workstations of that era.

Maybe it was this article about Queen Mary College (QMC):

"Acorn's Supercomputer", Acorn User, September 1986.

[julians]

Maybe it was this article about Queen Mary College (QMC):
"Acorn's Supercomputer", Acorn User, September 1986.

That's the article I remember, thank you.
As per gordonDrogon's original post, I'd be interested to see any user recollections from 'back in the day' as my thoughts are that it was too little, too limited and too late to really be a viable commercial option for academic or commercial usage even if compared to a Sun 3 or I guess the closest 'equivalent' of say a Whitechapel MG-1 at I'd guess two or three times the cost with significantly better graphics, local or networked storage, OS etc unless you were pretty much an Acorn only establishment.

But, I still wanted one...

Julian.

[gordonDrogon]

Thanks for the replies. I've just now read that article - and like other stuff then, bit too little, too late, too expensive. Interesting they compare it to a CRAY of the time too.

ARM was around the corner too, so lets just quietly walk away from it... Shame they never went down the Transputer route, but I suspect the same might have happened.

I was just trying to get a feel for what happened back then with regard to my own 'retro' thing. I've started with 6502 systems, moved to 65c816, written a nice (to me) little single-user OS for it that supports local s/w development, (well, editing & compiling) now looking to port that OS to something else. For a long time I've been looking at RISC-V but my recent re-introduction to the world of Acorn has made me re-think it in many ways - mostly now as a co-processor (implemented in software as a PiTube cpu) leaving all the hard work of IO, etc. to the old Beeb..

Cheers,

-Gordon

[paulb]

Thanks for the replies. I've just now read that article - and like other stuff then, bit too little, too late, too expensive. Interesting they compare it to a CRAY of the time too.

The perspective is intriguing to me now. Back in the day, all of this was completely alien to us who were still school age, and so we were prone to reading articles like that and to take it all at face value. The title - "Acorn's Supercomputer" - sounds impressive, and we might say that the term only gained its modern significance later, but it sounds rather naive if not delusional to me now.

However, I can understand that people's experiences with computers had been framed by mainframes and minicomputers, and then suddenly someone puts a microcomputer on the desk that is as fast as the minicomputer of only a few years prior, if not the mainframe. Of course, if Acorn's relatively cheap workstation was a "supercomputer" then what were the competitors?

I personally wonder how much experience article authors had with other forms of computing, like actual, pre-existing workstations. Given the Cambridge flavour of the Acorn scene, maybe those who were at Cambridge might have something to add about the style of computing at Cambridge, what systems they were using, and what systems they missed out on.

ARM was around the corner too, so lets just quietly walk away from it... Shame they never went down the Transputer route, but I suspect the same might have happened.

Having been looking at workstations past, I found some quite ambitious solutions based on these National Semiconductor products only yesterday. Like a company in 1984 called Syte Information Technology designing a NS32032 workstation that could be combined with others to provide a kind of cluster of virtual environments. That company ran out of funding, but they definitely had a plan!

As for the Transputer, there were companies doing add-on boards for the Archimedes series, and I think that attempts by companies like Atari to turn the Transputer into the basis of a workstation - Atari's effort being largely designed by the two Perihelion businesses - were always likely to show that the market was rather more likely to accept the Transputer as an accelerator, bolted on to existing solutions instead. Something similar could be said about the Intel i860 which was hyped as a "supercomputer on a chip" for a time. Both of these products seemed to do well on accelerator cards by the likes of Microway.

I was just trying to get a feel for what happened back then with regard to my own 'retro' thing. I've started with 6502 systems, moved to 65c816, written a nice (to me) little single-user OS for it that supports local s/w development, (well, editing & compiling) now looking to port that OS to something else. For a long time I've been looking at RISC-V but my recent re-introduction to the world of Acorn has made me re-think it in many ways - mostly now as a co-processor (implemented in software as a PiTube cpu) leaving all the hard work of IO, etc. to the old Beeb..

A RISC-V system reframed in retro terms would be like doing a MIPS-based evolution of any given retro lineage, especially given the similarities and shared heritage. A project I considered was to interface a PIC32MX microcontroller to the Electron, which would be a bit like simulating what would have been a different expensive second processor to Acorn's ARM second processor. Of course, the Pi-based Tube emulation route allows such things to be explored, as well.

I stumbled across a mention of Xenix on the Acorn hardware yesterday as well. Yet more tales taking on their own life, I imagine.

[gordonDrogon]

Thanks for the replies. I've just now read that article - and like other stuff then, bit too little, too late, too expensive. Interesting they compare it to a CRAY of the time too.

The perspective is intriguing to me now. Back in the day, all of this was completely alien to us who were still school age, and so we were prone to reading articles like that and to take it all at face value. The title - "Acorn's Supercomputer" - sounds impressive, and we might say that the term only gained its modern significance later, but it sounds rather naive if not delusional to me now.

I suspect I'm slightly older (apologies for the personal history here!) - when I was in 5th/6th year at school in Edinburgh in about 1978 we had the Apple II. We were (or might well have been) a million miles from Cambridge. I grew up with the Apple II and the local computing centre (branch of Edinburgh uni) who had a 32-bit mini running a revolutionary multi-user OS (for the time) called Mouses and all written in a high level language called Imp77.

However, I can understand that people's experiences with computers had been framed by mainframes and minicomputers, and then suddenly someone puts a microcomputer on the desk that is as fast as the minicomputer of only a few years prior, if not the mainframe. Of course, if Acorn's relatively cheap workstation was a "supercomputer" then what were the competitors?

At the time? Anything with a 68030...

In 1980 I went to uni and discovered Unix on a PDP11. Run as a sort of separate thing to the universities central computing facility - a horribly oversubscribed and under powered Prime. Also a pretty horrible branch of staff looking after them. Real preventers of making them actually useful for stuff we wanted to do. (Mostly internal politics, "land grabs", etc.) A few years later, still at uni. as a research associate thing, I was still finding value in the Apple IIs and then this new thing called the BBC Micro. (I was involved with research into industrial control, factory automation stuff like that). Our department bought their own Prime to run CAD/CAM software on, but it was still controlled by the central services. They were just not interested in networking or what lay outside, so we knew nothing.

I personally wonder how much experience article authors had with other forms of computing, like actual, pre-existing workstations. Given the Cambridge flavour of the Acorn scene, maybe those who were at Cambridge might have something to add about the style of computing at Cambridge, what systems they were using, and what systems they missed out on.

I think we moved to apollo workstations for the CAD/CAM - and in my little world I pushed for a network of Beebs with Econet, one per "station" on the factory floor (station being a CMC mill, lathe, etc.) I wrote all the code in BCPL and it was marvellous...

ARM was around the corner too, so lets just quietly walk away from it... Shame they never went down the Transputer route, but I suspect the same might have happened.

Having been looking at workstations past, I found some quite ambitious solutions based on these National Semiconductor products only yesterday. Like a company in 1984 called Syte Information Technology designing a NS32032 workstation that could be combined with others to provide a kind of cluster of virtual environments. That company ran out of funding, but they definitely had a plan!

As for the Transputer, there were companies doing add-on boards for the Archimedes series, and I think that attempts by companies like Atari to turn the Transputer into the basis of a workstation - Atari's effort being largely designed by the two Perihelion businesses - were always likely to show that the market was rather more likely to accept the Transputer as an accelerator, bolted on to existing solutions instead. Something similar could be said about the Intel i860 which was hyped as a "supercomputer on a chip" for a time. Both of these products seemed to do well on accelerator cards by the likes of Microway.

Fortune smiled on me and when I eventually left uni I worked for a company in Bristol making supercomputers out of the transputer, then transputer + i860, then Sparc... I was working with a lot of folk from Cambridge (some from early Acorn/ARM) It was all a bit of a mixed up place. Good though. I lost many brain cells to the i860 but programmed the transputer mostly in C and of-course, our desktops were all Unix...

I was just trying to get a feel for what happened back then with regard to my own 'retro' thing. I've started with 6502 systems, moved to 65c816, written a nice (to me) little single-user OS for it that supports local s/w development, (well, editing & compiling) now looking to port that OS to something else. For a long time I've been looking at RISC-V but my recent re-introduction to the world of Acorn has made me re-think it in many ways - mostly now as a co-processor (implemented in software as a PiTube cpu) leaving all the hard work of IO, etc. to the old Beeb..

A RISC-V system reframed in retro terms would be like doing a MIPS-based evolution of any given retro lineage, especially given the similarities and shared heritage. A project I considered was to interface a PIC32MX microcontroller to the Electron, which would be a bit like simulating what would have been a different expensive second processor to Acorn's ARM second processor. Of course, the Pi-based Tube emulation route allows such things to be explored, as well.

That's my thoughts too - I know its appeal would basically only to me. I like RISC-V as it reminds of of i860 and Sparc - slightly different from ARMs flavour of RISC. The issue I face is that I can't get a RISC-V system in the flavour I want - they are typically 32-bit microcontrollers with a few 10s of KB of RAM or 64-bit monsters designed to run Linux. (ESP32-C3/6 are almost good enough, but I'd like 1MB of RAM, not 400KB)

Which took me to the world of FPGAs then Beeb re-enters my life and I'm now quite (pleasantly) surprised at what people are doing with them - still, and creating a virtual CPU (not RISC-V, oddly enough) to fulfil my own needs seems relatively easy as the next step for my own project after losing far too many brain cells to the bloody awful 65c816.

And who knows, maybe someone will like it other than me...

-Gordon

[paulb]

I suspect I'm slightly older (apologies for the personal history here!) - when I was in 5th/6th year at school in Edinburgh in about 1978 we had the Apple II. We were (or might well have been) a million miles from Cambridge. I grew up with the Apple II and the local computing centre (branch of Edinburgh uni) who had a 32-bit mini running a revolutionary multi-user OS (for the time) called Mouses and all written in a high level language called Imp77.

I did follow some Retrocomputing Forum links to the work at Edinburgh, and I imagine that much of that was gradually forgotten by the broader industry in the course of time. I didn't do computing at school, contrary to early expectations, but Acorn's 32-bit machines were concurrent with my own passage through sixth form, at which point my school was generally still relying on RM Nimbus 286 machines for information technology classes. However, the maths teachers had sensibly repurposed the school's Beebs, and the science teachers had bought their own Archimedes models.

Of course, if Acorn's relatively cheap workstation was a "supercomputer" then what were the competitors?

At the time? Anything with a 68030...

1986 is slightly too early for the 68030, but anything with a 68020, certainly. And by 1988, Apollo had quite a few processor options, for example.

Fortune smiled on me and when I eventually left uni I worked for a company in Bristol making supercomputers out of the transputer, then transputer + i860, then Sparc... I was working with a lot of folk from Cambridge (some from early Acorn/ARM) It was all a bit of a mixed up place. Good though. I lost many brain cells to the i860 but programmed the transputer mostly in C and of-course, our desktops were all Unix...

Even in the early- to mid-1990s, there were plenty of people doing research into parallelism and using things like the Meiko Computing Surface, as I recall from the research projects in my computer science department. While I am sure they continued with their work, the emphasis on parallel programming techniques was probably weakened by the incredible acceleration of mainstream hardware, although some clever techniques were being adopted, including specific forms of parallel execution, that really allowed people to enjoy the benefits while sticking to traditional paradigms.

That's my thoughts too - I know its appeal would basically only to me. I like RISC-V as it reminds of of i860 and Sparc - slightly different from ARMs flavour of RISC. The issue I face is that I can't get a RISC-V system in the flavour I want - they are typically 32-bit microcontrollers with a few 10s of KB of RAM or 64-bit monsters designed to run Linux. (ESP32-C3/6 are almost good enough, but I'd like 1MB of RAM, not 400KB)

I'm not really on top of the breadth of RISC-V products out there. The problem sounds similar to that with ARM or MIPS, though: microcontrollers with a paucity of RAM or SoCs that need a lot of extra effort. I have had some exposure to a MIPS-based SoC that has several megabytes of RAM in the package, and although such chips need quite a bit of effort to put on a board, there are people trying to make more easily usable modules for "maker" audiences. I think that probably applies to SoCs of other architectures, as well.

[scruss]

The issue I face is that I can't get a RISC-V system in the flavour I want - they are typically 32-bit microcontrollers with a few 10s of KB of RAM or 64-bit monsters designed to run Linux. (ESP32-C3/6 are almost good enough, but I'd like 1MB of RAM, not 400KB)

The Milk-V Duo comes with at least 64 MB of RAM
Manufacturer page: Milk-V Duo | Extremely Cost-Effective Ultra-Compact Embedded Linux Development Platforms

[gordonDrogon]

Fortune smiled on me and when I eventually left uni I worked for a company in Bristol making supercomputers out of the transputer, then transputer + i860, then Sparc... I was working with a lot of folk from Cambridge (some from early Acorn/ARM) It was all a bit of a mixed up place. Good though. I lost many brain cells to the i860 but programmed the transputer mostly in C and of-course, our desktops were all Unix...

Even in the early- to mid-1990s, there were plenty of people doing research into parallelism and using things like the Meiko Computing Surface, as I recall from the research projects in my computer science department. While I am sure they continued with their work, the emphasis on parallel programming techniques was probably weakened by the incredible acceleration of mainstream hardware, although some clever techniques were being adopted, including specific forms of parallel execution, that really allowed people to enjoy the benefits while sticking to traditional paradigms.

Well - and at the risk of drifting way OT.... Just for the record, I worked for Meiko at the time and helped build, test and install many Computing Surface systems... A question often heard from prospective buyers: "I have this 20 year old FORTRAN code - make it go faster" ...

There was (and still isn't I believe) no magic solution to parallelism - you had to re-write your code to utilise parallelism, or link it with a pre-written standard library to do it for you. That approach was very common, so from what I recall (not my department) there was a small team re-writing the standard math libraries (BLAS, etc.) to make it appear "magic". The addition of the i860 was because customers wanted more speed and Inmos couldn't deliver... a 40Mhz RISC CPU that could do a multiply and add in a single cycle (cough, cough) was pretty fantastic at the time. The transputers were just fancy comms chips after that, then Meiko dropped the transputer entirely in-favour of their own Sparc CPU and custom comms chips.

Hardware gets faster, memory gets larger, but there is a limit and so we still need parallel processing.

This:

was the last big system I helped build - hard to give scale to it, but it's 2m high. "L" shaped as the comms wires go diagonally under the floor... About the computer power of a modern cellphone these days (with the electrical power of a small town)

One day I'll do a little video of my project although the idea/concept isn' something that would attract the retro-gamers so it's no x16, etc. ...

-Gordon

[paulb]

That's the article I remember, thank you.
As per gordonDrogon's original post, I'd be interested to see any user recollections from 'back in the day' as my thoughts are that it was too little, too limited and too late to really be a viable commercial option for academic or commercial usage even if compared to a Sun 3 or I guess the closest 'equivalent' of say a Whitechapel MG-1 at I'd guess two or three times the cost with significantly better graphics, local or networked storage, OS etc unless you were pretty much an Acorn only establishment.

On the topic of the MG-1, I was reviewing Udo Möller's excellent site about the NS32000, when I found a reference to Queen Mary College adopting the MG-1. See page 5, "Focus on QMC":

"Whitechapel News"

In this case, it was a DEC minicomputer that they were replacing for the 1986/1987 academic year, although the "far-sighted investment" apparently dated back to 1984.

[jimster]

Queen Mary did not keep those MG-1’s for long. I studied computer science there starting 1990 and they had built a new computing lab building and it was kitted out with Apple Mac ci’s runnning A/UX (Apples version of SVR2 Unix). Blimey I even remember Richard Bornat mentioned in that article who taught computer language theory and compilers!

[paulb]

Queen Mary did not keep those MG-1’s for long. I studied computer science there starting 1990 and they had built a new computing lab building and it was kitted out with Apple Mac ci’s runnning A/UX (Apples version of SVR2 Unix). Blimey I even remember Richard Bornat mentioned in that article who taught computer language theory and compilers!

Yes, I imagine that those MG-1 models were obtained at a discount given that the MG-200 was imminent, but as the references to "cost effectiveness" in that brochure suggest, both of these models would have been looking unattractive compared to newer products from the likes of Sun.

I haven't dug up much in the way of performance figures for the NS32332 of the MG-200, but fairly credible remarks in the Wikipedia article suggest that it was struggling to keep up with the competition. The NS32532 arrived in 1987, and although it seems to have been competitive with the 68030, the performance-seeking end of the industry was already migrating from the 68000 family to various RISC architectures.

Whitechapel were sensible and far from alone in adopting the MIPS architecture, but it seems that they probably needed more investment to make a success of it, and I guess that this was one of those situations where investors decide to cut their losses and take their money elsewhere. The loss of the supplier would have seen those deployed workstations heading for the door within a couple of years, which is a shame but entirely understandable.