The whole site is a gem; great collection of 80s/90s RISC/SMP computers from a lot of the cutting edge vendors (Apple/NEXT/Sun/Motorola/etc). The author repairs them and keeps them actively working, in addition to their documentation of the systems on this site.
I believe the contemporary PowerPC port of Solaris 2.5.1 would also have a chance of working, as well as the OS/2 port. I think by 1996 there was a working Linux distribution with the nascent PPC port?
The last two suggestions you offer (OS/2 and Linux) absolutely would be the case. I’m not sure about Solaris, at least officially. I’ve never heard of it in the PPC discussion groups in the 90s/00s.
EDIT: A trusted friend in this space (part of the MorphOS/Amiga PowerPC community) told me that one would absolutely be able to run the PPC Solaris port. The proper reference machine was actually the RS/6000 (see: https://virtuallyfun.com/2020/05/28/powerpc-solaris-on-the-r...), which ran on the 601, so the PowerStack I and PowerStack II would be absolutely capable of running said Solaris port competently.
The Solaris port was literally for a single point release of Solaris (2.5.1), and was dropped the next update. Extremely bizarre, but it is widely mentioned within that single version, in the installer and everything. Apparently the code was later referenced to help port OpenSolaris 10 to PowerPC, the result of which is here: https://github.com/andreiw/polaris
Very mysterious, and since it's little endian, it only runs on these PReP/CHRP machines with LE-supporting firmware, like Windows used. I assume it'd work, if Solaris has the device drivers...
Correct - appreciate you adding the context about the PReP/CHRP reference systems. You can’t take the Solaris install and just run it on any PowerPC build, even if it is a 60x family machine.
Solaris could not run on RS/6000 as it had a completely different internal architecture, starting from the POWER (not PowerPC) CPU's, a proprietary IBM system bus and other proprietary, IBM made peripherals.
Solaris was ported to the PowerPC architecture (PPC 601 only, I think) and required a PCI bus to run on, but it was a one-off act and more of a mere curiosity rather than a serious investment, and the Solaris PowerPC port was quickly abandonded. Solaris could run on a Motorola PowerStack though as the PowerStack was essentially a PC with a PPC 601/603 CPU and with standard PC components (e.g. you could install a USB PCI card and attach a USB keyboard/mouse but I do not think AIX 4 had readily available PCI USB drivers, not initially anyway).
The initial point release could not run on the RS/6000. There are people who have successfully installed and used it in the following years; it does however require modifications and source that were created after the fact. I’ve personally installed Solaris on a PPC ThinkPad, which is even more different from a PowerStack than an RS/6000 is.
My mention of the RS/6000 was mainly to point out something which Spijdar explicitly called out, which is that these PowerPC ports (which wouldn’t be on most PowerStacks anyway as everyone I know ran Unix on them) targeted the PrEP/CHRP platforms explicitly. So just having a PowerPC chip of the right family isn’t enough.
> The initial point release could not run on the RS/6000. There are people who have successfully installed and used it in the following years; it does however require modifications and source that were created after the fact.
I am curious about this. RS/6000's had had a microchannel bus and architecture until mid late nineties – when IBM started the transition onto the PCI bus in their RS6K workstations in the earnest (but not the servers), and – if my memory serves me well – the first few generations of new, PCI-based RS6k workstations had some hardware compatibility issues. Solaris also required Open Firmware to access the device tree and to boot up, and PReP proved to be a problematic replacement.
> I’ve personally installed Solaris on a PPC ThinkPad, which is even more different from a PowerStack than an RS/6000 is.
The PowerStack was effectively a standard PC with standard PC components inside, but with a PowerPC CPU and it came in a fancy, futuristic looking case which were the only two redeeming qualities / discerning features about the PowerStack. There was nothing else that was special about the PowerStack, and that was the point Motorola, Bull, Be and (to a certain extent) Apple and IBM were trying to make: we also make standard PC's, just like everyone else out there, that have a faster and a better CPU (PowerPC) compared to Intel CPU's. But then Motorola could not deliver, and IBM was not all that interested, and Intel dropped the Pentium Pro. Sun's own Ultra's 5/20/20/60 were, essentially, the same thing: a standard PC with a fancy CPU enslosed.
If you email me, when I’m back in the states in a week I can look through my notes about the previous such installs we’ve done, check my PPC software library and determine any modified or alternate software packages, and boot up my personal RS/6000 and run through the process myself again and fully document it.
I am happy to share my findings regarding what all needs to be done to install Solaris on this particular system family.
So did Solaris! ... For some bizarre reason. I would love to hear an explanation for that, given SPARC is big endian only. I guess it was based on the x86 port?
My fun fact is all PowerPC processors except the 970 "G5" CPU can do little endian! And the Linux kernel is actually capable of running LE processes on a BE kernel [1]. There was a bug for a long time in Linux where the kernel signal handler assumed all 32 bit processes are BE, but after patching that, 32 bit "ppcle" executables work.
[1] Well, you can load mismatched endian processes, but syscalls and ioctls need data in the kernel endianness, so you would need a very weird libc and userland to actually use this
So the previous computers in this class (for the 88k chips) they released a version of System V; the PowerStack also had a Unix, which was a (modified) AIX. The PowerStack II supported Windows NT (I remember using one in the 90s) and had official Motorola discs for this purpose. I’m not sure if the PPC NT was (officially) supported on the PowerStack; I’m not sure it’d be extremely performant on the non-e 603s. OS/2 would have been supported given IBM was the other chip producing partner in the AIM PPC alliance.
Today one’s best bet is a stable PPC Linux release or NetBSD.
Latest upstream kernel should work, because every time I send patches to arch/powerpc I manage to break one legacy platform or another, so there's still people testing them.
> OS/2 would have been supported given IBM was the other chip producing partner in the AIM PPC alliance.
Officially speaking, no, OS/2 PPC never supported any Motorola machines, only a limited selection of IBM machines. [0]
It probably could have been made to support Motorola PowerStack without too much work, maybe just a few extra drivers. But OS/2 PPC was essentially cancelled by IBM before it was even finished – it appears the "release" IBM made was simply a case of "we promised people we would release this, so let's just ship it unfinished so we can tell everyone we keep our word".
You’re correct. Supported was not the proper word choice here; given that I do know people who ran OS/2 on these systems in the 90s, and have done such installs myself, it’s a relatively painless process compared to, say, Solaris or NT on a PPC system that diverges from PReP/CHRP.
TL;DR: While IBM developed OS/2 to work on PPC, they never explicitly promoted or provided direct support for OS/2 on Motorola’s PPC. I was sloppy.
I would be shocked if anyone used it seriously - I think there was a novelty in simply having explored the software. Same goes for people who install the Solaris port.
Contrary to the other replies, I know of nobody who ran NT on the PowerStacks. You needed to install a special little endian firmware (ARC) to boot NT, which involved exchanging two PLCC EPROMs, typically by a Motorola FSE. Tried it on my Atlas board (same hardware in an AT form factor board) just for giggles… it ran even worse than NT on my Multia Alpha back then :).
The most common OS for all PowerStacks I have seen (quite a few) was a Motorola-specific version of AIX 4.1.3/4.
Of course, there was also a Linux port, but the machines were too expensive (compared to a regularPC) for a typical Linux use case.
I’d say from the discussions I’ve had with people who owned PowerStacks when they were initially being sold, 90% or so ran Motorola’s AIX when initially bought (as would be expected).
The interest in NT/Solaris/etc on these machines as an install option is a more contemporary phenomenon, I think, and comes from these PowerPC ports not really getting traction and therefore being an extremely rare OS on an already rare hardware platform.
> You needed to install a special little endian firmware (ARC)
Was this the same ARC on Alpha gear? I got a PDW tower for cheap on ebay a while back, got it to boot... A few times before the I/O board crapped out, and I remember one of the two firmwares was called ARC. (the other being SMRC I think? For OSF/1 and VMS and anything else)
It looked pretty similar to the Alpha version, but I have no idea if the code bases were related. The later PowerStack II machines (in more-or-less regular PC mini-towers) were able to just reflash the firmware, there was a tool called "Arcinst.exe" IIRC.
The regular Alpha firmware was SRM (or SRM console), this was used to boot OSF/1 (aka Digital Unix or Tru64) and VMS. Linux was able to boot both from SRM as well as ARC console. IIRC you had to reconfigure some other stuff when switching between the firmware versions, since the PALcode (think of low-level privileged machine code similar to SBI on RISC-V in M mode today) was different between SRM and ARC.
Both PPC and Alpha were rather open platforms back then, the PPC due to the IBM/Motorola/Apple alliance, Motorola's famously good databooks and the PowerOpen (PrEP/CHRP) standards, the Alpha also due to DEC's history in providing excellent documentation and especially John "Maddog" Hall's (https://en.wikipedia.org/wiki/Jon_Hall_(programmer)) great efforts to support the open source OS communities while he was head of the Unix group at DEC. I still have a TurboChannel Alpha machine (DEC3000) he sent me back then.
Good old times sigh. Now all you can hope for is a badly machine-translated and incomplete databook for an obscure Chinese RISC-V SoC...
First of all, thanks for the info in this thread - the context is great and helps fill in a bunch of the story that the site glosses over.
Secondly, I’ve been teaching my nephew and niece Smalltalk on the Raspberry Pi thanks to Crosstalk. So I can finally thank you for bringing that port into the world.
You're very welcome - great to see this information is still valuable to somebody out there. As a student in the '90s, it was fascinating being able to play around with all sorts of exotic hardware and from time to time grab some of it from the university's scrap heap. One was able to learn so much...
I still try to keep some of the old machines running. It becomes harder to find spare parts, but due to great projects by the open source hardware and software community (such as the various SCSI disk emulators and also system emulators such as qemu), some of it has also become much easier. Now I just need more time...
And I'm really happy to see that you are using Crosstalk for Smalltalk's original purpose! It's amazing to see the response to my little side project - high time to update it a bit (and add a RISC-V port). I hope your nephew and niece like it!
I'd like to echo the thanks for taking the time to write everything down here -- I was only born in the late 90s, so these machines are all before my time, and I love learning more about them.
I'm stuck with exploring the odd workstations I could get before the last reasonably priced ones dried up, and my modern POWER9 workstation. Things are better in a lot of ways now, but there's something magical about computers that power-on to Forth prompts and have ring binders full of detailed pin-outs and board schematics...
ARC was a specification, more like Open Firmware than specific implementation. NT before 6.0 can in fact boot only using ARC - the NTLDR code is responsible for emulating ARC on PC BIOS systems.
It originated on MIPS, iirc, and SGI MIPS machines used a related firmware.
On Alpha, there was AlphaBIOS firmware (first as an option that you needed to reflash, then as single mostly integrated package with SRM) which fulfilled the ARC standard combined with BIOS-like setup program.
Absolutely. I remember drooling over SGI, NeXT, PowerStacks, even the BeOS SMP Hobbit box, all throughout the 90s. I wanted a RISC workstation, ideally a SMP one, so badly. Unfortunately it just want a realistic proposition.
Meanwhile my actual computer was either a Mac IIe or a 133 MHz Pentium Packard Bell beige box. In retrospect, this was probably a benefit - I really focused on learning how to do low level systems programming to make my code performant (I was really into demo scene).
Still, my interest in RISC and PowerPC did eventually pay off when my first startup company sold a million handheld PowerPC-based portable media players in China in the mid-2000s. I got to meet a lot of the original PowerPC engineers and team members and, as Freescale was our official partner, was able to tour the production facilities. My unhealthy obsession somehow eventually paid off - and now I can play with these machines in my free time.
Company I was working for in the 1990s bought a PowerStack II machine to test on. This was the same but in a regular PC looking case. Of course being difficult the guy assigned this tried to put NT on it. After a week he couldn't work out how to get it to boot NT so the thing ended up in the corner of the office until it was chucked in a skip in the early 2000s.
We were a Sun shop mostly until Windows 2000 came out and then I left when I saw the death of Unix on the horizon.
It's interesting how for me year 2000 was the flourishing of Unix, not death: by that time, all the servers around me were running Linux or FreeBSD, and my desktop / laptop since then was invariably either Linux or sometimes MacOS.
I didn't have to touch Windows at work until 2020.
What is meant by "the time the Microsoft Windows/Intel x86 juggernaut was stumbling with their mass market Windows 3.11 replacement"? The way I remember this time is x86/Windows PCs exploding in popularity and Mac declining. But maybe that's a distorted view of a home rather than business user.
What was the target market of a desktop PC running AIX at the time?
37 comments
[ 2.7 ms ] story [ 90.4 ms ] threadhttp://www.corestore.org/Mvc-002s.jpg
http://www.corestore.org/powerstack.jpg
I believe the contemporary PowerPC port of Solaris 2.5.1 would also have a chance of working, as well as the OS/2 port. I think by 1996 there was a working Linux distribution with the nascent PPC port?
EDIT: A trusted friend in this space (part of the MorphOS/Amiga PowerPC community) told me that one would absolutely be able to run the PPC Solaris port. The proper reference machine was actually the RS/6000 (see: https://virtuallyfun.com/2020/05/28/powerpc-solaris-on-the-r...), which ran on the 601, so the PowerStack I and PowerStack II would be absolutely capable of running said Solaris port competently.
Very mysterious, and since it's little endian, it only runs on these PReP/CHRP machines with LE-supporting firmware, like Windows used. I assume it'd work, if Solaris has the device drivers...
Solaris was ported to the PowerPC architecture (PPC 601 only, I think) and required a PCI bus to run on, but it was a one-off act and more of a mere curiosity rather than a serious investment, and the Solaris PowerPC port was quickly abandonded. Solaris could run on a Motorola PowerStack though as the PowerStack was essentially a PC with a PPC 601/603 CPU and with standard PC components (e.g. you could install a USB PCI card and attach a USB keyboard/mouse but I do not think AIX 4 had readily available PCI USB drivers, not initially anyway).
My mention of the RS/6000 was mainly to point out something which Spijdar explicitly called out, which is that these PowerPC ports (which wouldn’t be on most PowerStacks anyway as everyone I know ran Unix on them) targeted the PrEP/CHRP platforms explicitly. So just having a PowerPC chip of the right family isn’t enough.
I am curious about this. RS/6000's had had a microchannel bus and architecture until mid late nineties – when IBM started the transition onto the PCI bus in their RS6K workstations in the earnest (but not the servers), and – if my memory serves me well – the first few generations of new, PCI-based RS6k workstations had some hardware compatibility issues. Solaris also required Open Firmware to access the device tree and to boot up, and PReP proved to be a problematic replacement.
> I’ve personally installed Solaris on a PPC ThinkPad, which is even more different from a PowerStack than an RS/6000 is.
The PowerStack was effectively a standard PC with standard PC components inside, but with a PowerPC CPU and it came in a fancy, futuristic looking case which were the only two redeeming qualities / discerning features about the PowerStack. There was nothing else that was special about the PowerStack, and that was the point Motorola, Bull, Be and (to a certain extent) Apple and IBM were trying to make: we also make standard PC's, just like everyone else out there, that have a faster and a better CPU (PowerPC) compared to Intel CPU's. But then Motorola could not deliver, and IBM was not all that interested, and Intel dropped the Pentium Pro. Sun's own Ultra's 5/20/20/60 were, essentially, the same thing: a standard PC with a fancy CPU enslosed.
I am happy to share my findings regarding what all needs to be done to install Solaris on this particular system family.
https://devblogs.microsoft.com/oldnewthing/20180806-00/?p=99...
https://stuff.mit.edu/afs/sipb/contrib/doc/specs/ic/cpu/powe...
https://stuff.mit.edu/afs/sipb/contrib/doc/specs/ic/cpu/powe...
My fun fact is all PowerPC processors except the 970 "G5" CPU can do little endian! And the Linux kernel is actually capable of running LE processes on a BE kernel [1]. There was a bug for a long time in Linux where the kernel signal handler assumed all 32 bit processes are BE, but after patching that, 32 bit "ppcle" executables work.
[1] Well, you can load mismatched endian processes, but syscalls and ioctls need data in the kernel endianness, so you would need a very weird libc and userland to actually use this
Today one’s best bet is a stable PPC Linux release or NetBSD.
Officially speaking, no, OS/2 PPC never supported any Motorola machines, only a limited selection of IBM machines. [0]
It probably could have been made to support Motorola PowerStack without too much work, maybe just a few extra drivers. But OS/2 PPC was essentially cancelled by IBM before it was even finished – it appears the "release" IBM made was simply a case of "we promised people we would release this, so let's just ship it unfinished so we can tell everyone we keep our word".
[0] https://www.os2museum.com/wp/os2-history/os2-warp-powerpc-ed...
TL;DR: While IBM developed OS/2 to work on PPC, they never explicitly promoted or provided direct support for OS/2 on Motorola’s PPC. I was sloppy.
I'm amazed to hear anyone actually ran OS/2 PPC (beyond its own development team, and a handful of ISVs IBM recruited to port software to it.)
The most common OS for all PowerStacks I have seen (quite a few) was a Motorola-specific version of AIX 4.1.3/4.
Of course, there was also a Linux port, but the machines were too expensive (compared to a regularPC) for a typical Linux use case.
The interest in NT/Solaris/etc on these machines as an install option is a more contemporary phenomenon, I think, and comes from these PowerPC ports not really getting traction and therefore being an extremely rare OS on an already rare hardware platform.
Was this the same ARC on Alpha gear? I got a PDW tower for cheap on ebay a while back, got it to boot... A few times before the I/O board crapped out, and I remember one of the two firmwares was called ARC. (the other being SMRC I think? For OSF/1 and VMS and anything else)
The regular Alpha firmware was SRM (or SRM console), this was used to boot OSF/1 (aka Digital Unix or Tru64) and VMS. Linux was able to boot both from SRM as well as ARC console. IIRC you had to reconfigure some other stuff when switching between the firmware versions, since the PALcode (think of low-level privileged machine code similar to SBI on RISC-V in M mode today) was different between SRM and ARC.
Both PPC and Alpha were rather open platforms back then, the PPC due to the IBM/Motorola/Apple alliance, Motorola's famously good databooks and the PowerOpen (PrEP/CHRP) standards, the Alpha also due to DEC's history in providing excellent documentation and especially John "Maddog" Hall's (https://en.wikipedia.org/wiki/Jon_Hall_(programmer)) great efforts to support the open source OS communities while he was head of the Unix group at DEC. I still have a TurboChannel Alpha machine (DEC3000) he sent me back then.
Good old times sigh. Now all you can hope for is a badly machine-translated and incomplete databook for an obscure Chinese RISC-V SoC...
Some more tidbits in case anyone's interested.
This web page by Michael Kraemer has more information on NT for the PowerStacks (there's also a lot of other information about 90s computer gear): https://web-docs.gsi.de/~kraemer/COLLECTION/MOTOROLA/wntps.h...
And this DEC manual contains more information on Alpha firmware than you ever wanted to know... https://manx-docs.org/collections/antonio/dec/MDS-2000-01/cd...
See https://download.majix.org/dec/palcode_dsgn_gde.pdf for PALcode details. Example PALcode can be found at http://bitsavers.org/pdf/dec/alpha/palcode.zip - it's interesting to read.
First of all, thanks for the info in this thread - the context is great and helps fill in a bunch of the story that the site glosses over.
Secondly, I’ve been teaching my nephew and niece Smalltalk on the Raspberry Pi thanks to Crosstalk. So I can finally thank you for bringing that port into the world.
You're very welcome - great to see this information is still valuable to somebody out there. As a student in the '90s, it was fascinating being able to play around with all sorts of exotic hardware and from time to time grab some of it from the university's scrap heap. One was able to learn so much...
I still try to keep some of the old machines running. It becomes harder to find spare parts, but due to great projects by the open source hardware and software community (such as the various SCSI disk emulators and also system emulators such as qemu), some of it has also become much easier. Now I just need more time...
And I'm really happy to see that you are using Crosstalk for Smalltalk's original purpose! It's amazing to see the response to my little side project - high time to update it a bit (and add a RISC-V port). I hope your nephew and niece like it!
I'm stuck with exploring the odd workstations I could get before the last reasonably priced ones dried up, and my modern POWER9 workstation. Things are better in a lot of ways now, but there's something magical about computers that power-on to Forth prompts and have ring binders full of detailed pin-outs and board schematics...
It originated on MIPS, iirc, and SGI MIPS machines used a related firmware.
On Alpha, there was AlphaBIOS firmware (first as an option that you needed to reflash, then as single mostly integrated package with SRM) which fulfilled the ARC standard combined with BIOS-like setup program.
Meanwhile my actual computer was either a Mac IIe or a 133 MHz Pentium Packard Bell beige box. In retrospect, this was probably a benefit - I really focused on learning how to do low level systems programming to make my code performant (I was really into demo scene).
Still, my interest in RISC and PowerPC did eventually pay off when my first startup company sold a million handheld PowerPC-based portable media players in China in the mid-2000s. I got to meet a lot of the original PowerPC engineers and team members and, as Freescale was our official partner, was able to tour the production facilities. My unhealthy obsession somehow eventually paid off - and now I can play with these machines in my free time.
We were a Sun shop mostly until Windows 2000 came out and then I left when I saw the death of Unix on the horizon.
I didn't have to touch Windows at work until 2020.
What was the target market of a desktop PC running AIX at the time?
was code-named Chicago, and they pre-leaked data about it for years in the press while struggling with Apple Computer for lots of markets