• Please review our updated Terms and Rules here

i860 based Risc Unix workstation - PCS Cadmus FX.1 aka Firebox - Unix V4 -- quite a rare system imo

W

Wildfire

Experienced Member
Joined
Jul 1, 2014
Messages
252
Location
Germany, near Göttingen
Hello Folks,

quite a while ago during the covid time a son of a father who i once had a short contact with about PCS Cadmus 68000 based Unix workstations contacted me.
He told me that his father passed away and they found some email conversation and wanted to ask if i have some interest in getting the Cadmusses.
I said yes and on a day we both had time i drove there to look after what they got there.

Ok, there has been quite some hardware (also SGIs and a few other Unix style things) but my goal was to get the 68000 based systems which up to then i had none of.
While looking after some spare parts in a storage room i noticed a peecee style mini tower with a sticker "Firebox i860 40MHz 16MB Ram" on it
and after a few seconds all alarm bells ringing in my head
told me that this is some rare item.
There also has been a spare unused 5 1/4 full height NOS scsi harddrive and i figured out that there also is a complete 2nd "Firebox" as a spare part !
Also there was an QIC 525 style tape with some writing on it like " UFS root dump backup".
So everything one needs has been there.

So i told them i would also like to have these system and spareparts and we could arrange us.

One of the boxes is 40 MHz, the other 30 MHz.

I cleaned the tower a little bit and started to investigate how this eventually should work.
The firebox seemed to only need a power connection the old style like the harddrive brick with 5v/12v.

Sadly the powersupply worked but none of the two units came up with some output at the serial console.

There may be a chance that i can contact some people who worked with these units and eventually get it up and running again.
I will try to image the internal harddrive as one next step and also the backup tape. UFS should without problems be readable with Linux dd and ddrescue.

So if here is someone who eventually has such a system up and running or has worked with these it would be great to hear from you !

2nd i wrote this thread for documentation of what i have to let the forum take part in seeing these pieces of at least a little bit of computer history.
Photos follow with the next post.
 
Pictures:
 

Attachments

  • 20210327_001849.jpg
    20210327_001849.jpg
    847.5 KB · Views: 46
  • 20230813_011227.jpg
    20230813_011227.jpg
    554.3 KB · Views: 43
  • 20210327_000208.jpg
    20210327_000208.jpg
    971.9 KB · Views: 37
  • 20210327_000231.jpg
    20210327_000231.jpg
    876.6 KB · Views: 41
  • 20210327_000849.jpg
    20210327_000849.jpg
    1.5 MB · Views: 41
  • 20210327_000734.jpg
    20210327_000734.jpg
    1.6 MB · Views: 41
  • 20210327_000616.jpg
    20210327_000616.jpg
    2 MB · Views: 38
  • 20210327_000424.jpg
    20210327_000424.jpg
    1.6 MB · Views: 36
  • 20210327_000059.jpg
    20210327_000059.jpg
    1.6 MB · Views: 47
Thinking about why the i860 failed as a general purpose CPU I vaguely wonder if more dynamic code generation techniques like JIT LLVM had been around back then if that huge gulf between its theoretical peak and more typical (and relatively lousy) performance might have been better.

… of course, that would have ballooned other system requirements like RAM, and maybe at that level of max performance the overhead would have just been too much, but it’s sort of interesting to ponder. Wonder if anyone bothers trying to write software for the i860 today.
 
I took the i860 as a half-hearted attempt at a new procressor--and performance was badly oversold. The "half-hearted" effort was what killed it. (I did buy the book and was moderately excited). There supposedly was a version of NT developed for it, but I've never seen the beast.

Itanium suffered a similar fate. I think if some enterprising engineer were to introduce another non-x86 CPU, he'd be furloughed to mind the coffee maker. Intel knows darned well that any silicon coming out had better execute x86 code. (XScale being the exception, but Intel didn't invent ARM).
 
I’ve seen reasonable arguments that if Intel had been serious the i960 would have been the better architecture to push for “mainstream“ applications but, yeah, I don’t know how serious they ever were. I would guess that even as early as the late 1980’s Intel’s reptile brain deep inside had grasped better than any other chipmaker (especially Motorola) how important backwards binary compatibility was in the consumer market and being better at that than anyone else was their roadmap to victory.
 
I worked at Sperry back in the 80s on an 1100 followon machine (Saturn was the code name. I seem to remember there were two others too, but I don't remember the code names). They were not backwards compatible and were eventually dropped. Even they knew you had to keep backwards compatibility.
 
Oh, neat! I didn't know there ever were proper commercial-ish applications for the thing; all the retrospective coverage makes it sound like something that was essentially stillborn.

Always have wondered what it'd be like to get a real talent with machine-language coding (some demoscene types, say) and see what they could do with an infamous VLIW flop like the i860. Puttering around with the Parallax Propeller certainly gives you an appreciation for the challenge of optimal instruction ordering, I must say - and that architecture is specifically designed for maximal flexibility in that regard, right down to its instruction format.
 
I believe it. Tricky enough doing that stuff with the conditional-execution and flag-preserve options that (I'd bet money) the Propeller got from the DG Nova by way of the Xerox Alto; having to do without...man.
 
The 6600 did have a certain architectural advantage over the i860--no condition codes. You had conditional branches on a (arithmetic) register being zero, nonzero, plus or negative (or a combination of same); each register had its content flags. So the arithmetic computation for a branch could be carried out far in advance of the branch. This also works since the 6600 at the time, had 8 60 bit arithmetic registers, in addition to 8 18-bit address and 8 18-bit index registers. And operations are three-address. Remember, this was all second-generation implementation---lotsa transistors.

ARM, of course, took a different approach with each instruction being conditionally executed on the state of the status register.

At any rate, the i860 architecture makes for a good read. I can recommend the Neal Margulis' "i860 Microprocessor Architecture" published by Osborne-McGraw-Hill (1990).
 
m_thompson said:
I have a quad-i860 board in one of my Sun workstations.
Click to expand...
It's not a Sun i860 workstation, it's Sun workstation that also happens to have an i860 board in it, right? Like it would have a graphics card or so other VME card (if you machine was, indeed, a VME machine)?
 
You must log in or register to reply here.
Back
Top