Acorn Galileo (the computer?)/Dream Machine/Turbo Nutter

[flibble]

From an interesting hard disc image of a former Acorn employee I found a couple of documents about a computer hardware project that never saw the light of day but quite a lot of development thought went into it.

Trigger Warning: If you're an ARM/RISC OS 'true believer' Acorn's own opinion of its market position in 1994 will likely be difficult reading.

http://www.4corn.co.uk/archive//misc/ga ... dacted.pdf
http://www.4corn.co.uk/archive//misc/ga ... dacted.pdf

The context:

It's 1994, the Risc PC has just been launched with the ARM6, ARM7 and ARM8 cpus are on the roadmap for the next few years. The pentium is out, windows 95 (4/Chicago) is in testing. The AIM alliance (Apple/IBM/Motorola) has made good progress on the POWER and PowerPC architecture.

The machine:

A PCI bus based machine, with plug in cards for different CPU architectures and OS support, ARM/RISC OS, Pentiums, PowerPC and SPARC are mentioned. There's an ARM8 on the motherboard running TAOS that's used to handle switching between the different CPUs and OSs. With the suggestion of the 3D 'Rooms' interface for changing OS, in the style of Virtual Reality that was a buzz word at the time.

The outcome:

Information about TAOS and a demo of the Rooms interface were given out to members of the Clan

The machine itself was never released and the project likely mostly abandoned before too much more work was done on it.

The Phoebe, Risc PC 2, was developed instead and was a much less radical in terms of target audience and technology.

However the codename lived on, as a new Acorn OS (Embedded, microkernel, PMT, multi threaded) Galileo, probably originally intended to be the OS that would have been on the onboard ARM8. This OS also disappeared before any releases in the shutdown of the Acorn Workstation division in 1998.

[danielj]

This is brilliant. It's in the familiar vein of engineers who've thought up an amazing solution for a problem that no one's actually got

[jms2]

Bonkers... it just goes to show that there's a fine line between being a bit unconventional and total madness.

[davidb]

It's interesting in that there was someone (or many) at Acorn who would admit that x86 was going to outperform ARM and that the RiscPC was just a stopgap machine. As a user, it was exciting to see the RiscPC release in 1994, but it felt to me like a stopgap.

The proposed solution - to go all in on a multiprocessor workstation - seems a bit implausible given what else Acorn were known for at the time.

Warning for pedants: the presentation uses the name "Risc OS" for the operating system also known as "RISC OS". I guess we're allowed to use either of them now.

[algenon_iii]

1994, yeah kind of makes sense. Amiga and ST were finished by then basically and the PC was really coming into its own (i.e. Doom), with Win95 due to come out the next year. I think businesses were starting to make noises about kids leaving school without knowing how to use a PC.

The good thing is that at least some people realised there was a problem.

[Naomasa298]

Still, the ARM didn't do too badly. With Apple now going ARM, it's funny how things have come full circle.

[algenon_iii]

Still, the ARM didn't do too badly. With Apple now going ARM, it's funny how things have come full circle.

Whatever your opinion is of Apple they did put money into ARM Ltd in the early days and who knows whether ARM would have been as successful without the backing of a major US computer company. Now 30 years on they'll fully switch to ARM.

[BigEd]

The machine:

A PCI bus based machine, with plug in cards for different CPU architectures and OS support, ARM/RISC OS, Pentiums, PowerPC and SPARC are mentioned. There's an ARM8 on the motherboard running TAOS that's used to handle switching between the different CPUs and OSs...

Unexpectedly packed:

A scalable bank of processors
4 slots
each slot takes upto 8 ARM processors
or 4 Power PC or 2 Pentiums

On the one hand, supporting multiple CPUs, especially multiple ARMs, seems like a good way to try to get ahead, or keep up, in the performance stakes.

On the other hand, what about memory bandwidth? The genesis of ARM is on the observation that maximising use of memory bandwidth is a big win. I wonder if these CPUs have caches, whether on-chip or off-chip, and how large those caches are.

I also wonder at the bandwidth needed by the 1600x1200x32bits of video: you need bandwidth to get the pixels out to the display, and bandwidth to get those pixels updated with content.

Did anything noteworthy happen in the land of Acorn 32 bit offerings to make use of multiple ARMs, to separate graphics memory, to use caching to help with memory bandwidth?

[Boydie]

This rings a lot of bells, mainly from being drunk a lot with Acorn/ARM people around that time. Given the author’s style, I’d hazard a guess as to their identity if I didn’t know how rabidly anti-Wintel they were.

Can’t remember how far things got. ISTR it spawned the Hydra board from Simtec and Taos made it onto floppies, but I’m not sure if the two were ever formally introduced.

Edit: Check out the Archive Submissions section...

[richw]

Didn't Acorn make a bit of noise about the CHIRP (later PREP) standard, where they would use an off the shelf PowerPC board and plug in an ARM card. No idea how RISC OS would ever have booted!

[flibble]

Didn't Acorn make a bit of noise about the CHIRP (later PREP) standard, where they would use an off the shelf PowerPC board and plug in an ARM card. No idea how RISC OS would ever have booted!

Yep I remember the CHRP mention too, would have been about at this time or a little bit later I think.

I think the main problem this proposal, the CHRP one and finally the Phoebe had was no one sat back and said "But we won't sell very many if it costs too much"

[flibble]

Didn't Acorn make a bit of noise about the CHIRP (later PREP) standard, where they would use an off the shelf PowerPC board and plug in an ARM card. No idea how RISC OS would ever have booted!

Yep I remember the CHRP mention too, would have been about at this time or a little bit later I think.

I was wrong about the timing, checking my own archives, CHRP was 1996 (Galileo was 1994). The idea being RISC OS and Mac OS on the same hardware for Xemplar. However the title of the internal email suggests this wasn't a serious development...

http://www.4corn.co.uk/articles/emails/t/19960227-2.txt