YouTube: Amiga as Workstation in 1986

[tschak909]

After more than four years of trying to gather bits together, I have started a video series showing the viability #CommodoreAmiga as a workstation.

This set of videos are placed in the historical context of the Amiga roughly a year after launch, in the summer of 1986, eventually comparing context to other workstations of the time.

In this video, we show initial first impressions of capabilities using a base system and viewing the demonstration programs, while providing context on how each demonstration program worked.

The system at this point is an Amiga 1000, with 256K expansion, and two disk drives.

Since workstations were used primarily for scientific and engineering tasks, with the expectation of developing custom software for a given task, this is what the series will ultimately focus on.

#retrocomputing

[ Show youtube player ]

---

[ Show youtube player ]

It is early 1986, and we've received our development disks from #CommodoreAmiga. We set up our compiler disks, compile a couple of test programs, and run them.

Hard disks couldn't arrive soon enough.

#retrocomputing

[tschak909]

Amiga as a Workstation: Part 2: Adding RAM.

[ Show youtube player ]

We add more RAM to our #CommodoreAmiga using a schematic we found on Fred Fish disk #27. Once added, we add a couple of tools from the Amiga Macro assembler disk to allow AmigaDOS to see the RAM. We then copy our tools and code into the RAM: disk, and observe the massive speed increase.

[slaapliedje]

Awesome! I have wanted to do something similar with a series on alternative platforms and productivity software. I just need to figure out how I want to approach it.

[tschak909]

Amiga as a Workstation Part 3: Adding a Hard Disk

[ Show youtube player ]

In the previous part, we could use the extra RAM in our Amiga to vastly improve compile times by using the RAM disk. In this video, we add a Tecmar T-Card and T-Disk, which gives us 20 megabytes of storage, move our environment onto it, and test it.

[hammer]

If I'm a business customer, the demos shown are not ready for business out of the box.

I didn't see a "workstation" with ready "day job" applications.

[tschak909]

Originally Posted by hammer

If I'm a business customer, the demos shown are not ready for business out of the box.

I didn't see a "workstation" with ready "day job" applications.

I very much appreciate your snap judgement for an ongoing series, which does not, and will not focus on business uses at all, that's not what workstations were used for.

As this progresses, while I will be doing simpler demonstration application development, the point is to show how the system, as we build it up, was comparable to the 680x0 based workstations that dominated the field in the mid 1980s, and how the Amiga did handily do those applications. I am speaking as someone who was there doing it.

-Thom

[hammer]

Originally Posted by tschak909

I very much appreciate your snap judgement for an ongoing series, which does not, and will not focus on business uses at all, that's not what workstations were used for.

As this progresses, while I will be doing simpler demonstration application development, the point is to show how the system, as we build it up, was comparable to the 680x0 based workstations that dominated the field in the mid 1980s, and how the Amiga did handily do those applications.
-Thom

The Amiga 1000 didn't have Unix.

SunOS 3.0 was released in Feb 1986 with code based on BSD4.2 and Unix System V. Sun 3 hardware had Sun-3 MMU, 68020, 68881 in May 1985.

Pure "68000" workstation for 1986 is late.

https://ftpmirror.your.org/pub/misc/..._2.0_May85.pdf
Sun-3 has included ECC memory support since 1985! The encryption processor is AMD 8068. System integrity is a factor for workstations by the mid-1980s.

Sun-3 is ethernet capable e.g. AMD Ethernet Interface uses the 7990 chip or Intel Ethernet Interface uses the Intel 82586 chip.

The hardware features mentioned in the Sun-3 document are on my modern ASUS AM4/AM5-based PCs e.g. "workstation" ECC memory support, Ethernet Interface, PMMU, TPM encryption processor, IEEE-754 FPU and 'etc'.

Commodore progressively added features on the Amiga hardware platform, but it's missing ECC memory support e.g. A3000UX didn't have "workstation" ECC memory.

Local ECC-capable RAM and CPU accelerator could be added to A2000/A3000's local CPU slot. Big box A2000 is the minimum Amiga platform to add the missing hardware features from Sun-3.

Originally Posted by tschak909

I am speaking as someone who was there doing it.

I'm game for a pissing contest. My university has Sun Unix boxes and it's required to use Unix.

[tschak909]

Originally Posted by hammer

The Amiga 1000 didn't have Unix. By 1986, 68K based Unix workstations have memory protection via PMMU. 68000 without a proprietary PMMU and Unix System V Release 2 in 1986 is not a workstation.

And yet, some of us used them as such, at that time, as they were substantially cheaper than the UNIX workstations, and could fit in discretionary budgets.

The videos I am making will systemically prove my point.

Dealing with people like you is exceptionally frustrating, especially because it's very clear you've never had to do engineering or scientific tasks like circuit design, or e.g. visualization of a heat map, and needing to do it on limited budgets. This was very common in places like NASA, which did use Amigas as workstations, a need which sprang out of needing to buy half a dozen systems for engineers to do things like circuit design; when the budget for a department had already been spent on a large HP 9000 system
that was meant to be shared amongst said engineers. These were things we did to get things done on time.

Yes, the Sun2 and Sun3 systems had more pixels, yes, they could have large disks, yes, they had memory protection, yes, they had ethernet but they were _SLOW_. Did you ever use one? Did you ever have to procure them? The Sun 3/50's that were standard issue for so many circuit designers were 12,500 for a diskless system, with a bw3 framebuffer. Circuit designers needed a cg3, price jump to more than 20,000 USD. Still diskless, by the way.

We even did work around the issue that there was no ethernet card yet by making a DMA card which talked directly to the HP 9000 system. It worked, until we could get an ethernet card (which was sometime around the fall of 1988, when we got AS225 and the A2065)

I can keep belaboring my points...

-Thom

[hammer]

Originally Posted by tschak909

And yet, some of us used them as such, at that time, as they were substantially cheaper than the UNIX workstations, and could fit in discretionary budgets.

The videos I am making will systemically prove my point.

Dealing with people like you is exceptionally frustrating, especially because it's very clear you've never had to do engineering or scientific tasks like circuit design, or e.g. visualization of a heat map, and needing to do it on limited budgets. This was very common in places like NASA, which did use Amigas as workstations, a need which sprang out of needing to buy half a dozen systems for engineers to do things like circuit design; when the budget for a department had already been spent on a large HP 9000 system
that was meant to be shared amongst said engineers. These were things we did to get things done on time.

Yes, the Sun2 and Sun3 systems had more pixels, yes, they could have large disks, yes, they had memory protection, yes, they had ethernet but they were _SLOW_. Did you ever use one? Did you ever have to procure them? The Sun 3/50's that were standard issue for so many circuit designers were 12,500 for a diskless system, with a bw3 framebuffer. Circuit designers needed a cg3, price jump to more than 20,000 USD. Still diskless, by the way.

We even did work around the issue that there was no ethernet card yet by making a DMA card which talked directly to the HP 9000 system. It worked, until we could get an ethernet card (which was sometime around the fall of 1988, when we got AS225 and the A2065)

I can keep belaboring my points...

-Thom

I updated my post, Sun 3 has ECC memory. My workstation background is based on SUN-based systems before using DEC Alpha.

[hammer]

Originally Posted by tschak909

Dealing with people like you is exceptionally frustrating, especially because it's very clear you've never had to do engineering or scientific tasks like circuit design, or e.g. visualization of a heat map, and needing to do it on limited budgets.

There you go again, personal attacks me. You argued for Amiga 1000 without IEEE FPU while Sun-3 has 68881! Are you nuts?

Engineering... I used AutoCAD R11 to R13 on Sydney WaterBoard's IBM PS/2 "32-bit" workstation and XCAD2000 in my home A3000. I added 387-33 FPU to my home 386DX-33 PC. IEEE FPU is a must-have.

Originally Posted by tschak909

This was very common in places like NASA, which did use Amigas as workstations, a need which sprang out of needing to buy half a dozen systems for engineers to do things like circuit design; when the budget for a department had already been spent on a large HP 9000 system
that was meant to be shared amongst said engineers. These were things we did to get things done on time.

NASA's Amigas usage from https://ntrs.nasa.gov/search?q=Amiga NASA listed Amiga's specific use cases.
The date range from 1989. Amigas were used in mostly visual/presentation use cases.

Originally Posted by tschak909

Yes, the Sun2 and Sun3 systems had more pixels, yes, they could have large disks, yes, they had memory protection, yes, they had ethernet but they were _SLOW_.

Better than zero. Sun-3 has ECC memory, not just memory protection. There are reasons for Linux's and FreeBSD's existence i.e. to drive down *nix cost.

Big box Amigas had the potential.

Originally Posted by tschak909

Did you ever use one?

It's a requirement for my university subjects.

Originally Posted by tschak909

Did you ever have to procure them?

My SUN box usage is federal government-funded.

[Thomas Richter]

From 1990/91 I remember that we had some HP/UX workstations, based on 68020 and 68030, and they were barely usable with their 4MB or 8MB RAM. Working on the text console was probably fine, but as soon as you started X, the system was continuosly swapping, and starting an emacs slowed it down to a crawl. Thus, you had to bring a lot of patience to use the system. It was not quite practical to use this system as multi-user system, though one of the systems was exclusively reserved to only run the X window manager to make a things a bit faster.

What was seriously missing on the Amiga end was booting from harddisk (that arrived with 1.3, even though 1.2 should do it, but could not do it due to defects), and a network stack (which did not arrive until a lot later, and only as an add-on to the systems).

[Citation]

I'm definitely enjoying the series, so thank you Thom!

[fxgogo]

Just found your video, looking forward to the series.

[aeberbach]

Thom I am loving these videos, remembering the time I first sat in front of a 1-floppy 512K A1000 and tried to make a compiler work. Keep it up!

@hammer, just try to enjoy the videos...

[pixie]

It did give us a glimpse of the future, despite having or not fpu, today we use the computer more inline with what is shown in this series then dos, that's a given! Great series

[tschak909]

Originally Posted by Thomas Richter

From 1990/91 I remember that we had some HP/UX workstations, based on 68020 and 68030, and they were barely usable with their 4MB or 8MB RAM. Working on the text console was probably fine, but as soon as you started X, the system was continuosly swapping, and starting an emacs slowed it down to a crawl. Thus, you had to bring a lot of patience to use the system. It was not quite practical to use this system as multi-user system, though one of the systems was exclusively reserved to only run the X window manager to make a things a bit faster.

What was seriously missing on the Amiga end was booting from harddisk (that arrived with 1.3, even though 1.2 should do it, but could not do it due to defects), and a network stack (which did not arrive until a lot later, and only as an add-on to the systems).

As I am showing here, I was using a hard disk with 1.1.

-Thom

[Thomas Richter]

Originally Posted by tschak909

As I am showing here, I was using a hard disk with 1.1.

-Thom

Is that really the case? From what I see from your (the installed) startup-sequence, it is first booting off the floppy disk, calling the t-disk setup program, which then redirects booting to the harddisk. This is also how Kick 1.2 booted from HD - except that it is there done manually by redirecting all assigns as part of the startup-sequence of the workbench floppy. Autoconf and bootpoint booting arrived much later in the Kickstart. IOWs, try to boot with the floppy drive empty - I would be astonished if that works....

...the only way how a vendor could possibly do that is by heavily patching the system, that is, disabling the kickstart WOM early, booting from a ROM in the "host adapter", then depending on fixed ROM addresses in 1.1 (or 1.2), and then continuing the boot. In such a case, this host adapter would stop working with 1.2 (and again with 1.3). There were a couple such "custom hacks" in the day, but Kickstart had no possibility to detect booting hardware in 1.1, and while it should have had in 1.2, it did not work before 1.3.

[tschak909]

Yes, before we had full auto-boot in 1.3, we had to bootstrap off of the floppy and pull stakes to re-assign to the hard disk. That's just how it worked.

While AmigaDOS from the beginning could attach block devices from external code, there was no mount utility, so there were external programs which called exec and dos.library to present the device to the system. Mount would be added in 1.2, as would BindDrivers, the expansion library, etc, to make this easier.

It is also worth noting that the transition from 1.0 to 1.2 happened in the space of one year. It honestly was nuts, thinking about it now, but there was a tremendous amount of development happening in both first and third party camps.

[Locutus]

Just for balance and even though I really like UNIX, it's not correct to say that in 1986 a workstation implied UNIX.

DEC was selling significant amount of VAX/VMS based workstations and Apollo had Aegis aka DomainOS, which was at that timepoint not a UNIX either.

Both of them significant players in the workstation market at the time, 86 is probably the last moment to still release a non-UNIX-oid workstation platform though.

[pandy71]

@tschak909 - i share your perspective - Amiga was first machine to define new computing class - i call it personal workstation or perhaps home workstation - HW and OS provided functionality not existing earlier in this money range.