Oscar, a thought for your next project for us

[DR]

This would be an all winter project.

It is called a Cray-1 or -2 depending on how you want to design it.

This should mesh perfectly with those of us having a PiDP-11 which was I guess from talking with one of the guys at the museum, something that their computer offloaded the output to so the little computer could do the busy work of printing or writing to disk.

They had something called a Data General Eclipse which looks a lot like the DEC stuff, maybe they copied the idea:

and speaking of drum storage:

Look at the size of the drive motor on this.  Those were fixed heads apparently.

I'm not sure we'd like all those wires, but a few would be nice as a facade and then let Simh take care of emulating the beast.

This was a fun visit to the Chippewa Valley Industry and Technology open house  on the 28th for Seymour Cray's 100th birthday.  I had no idea they had so many models.  The Serial #1 Cray-1 which went to Los Alamos Labs in Albuquerque

was there.  They said it had a period of operation of over 16 years and was very reliable.  Some disks that were about 4' across were mounted on the walls.  The various other Cray versions were open to look in and rap on the machined metal supports.  The cooling liquid which was sort of like freon was almost twice as heavy as water to compare a couple jugs of it to life.  So for those wondering about cooling their Pi-5s with a fan, I guess you could submerge the whole thing in this liquid.

Hope this was a fun diversion for the members here.

Dale

[Bob Darlington]

On Mon, Sep 29, 2025 at 11:22 AM DR <daleea...@gmail.com> wrote:

This would be an all winter project.

It is called a Cray-1 or -2 depending on how you want to design it.

<snip> 

I'm not sure we'd like all those wires, but a few would be nice as a facade and then let Simh take care of emulating the beast.

This was a fun visit to the Chippewa Valley Industry and Technology open house  on the 28th for Seymour Cray's 100th birthday.  I had no idea they had so many models.  The Serial #1 Cray-1 which went to Los Alamos Labs in Albuquerque

was there.  They said it had a period of operation of over 16 years and was very reliable.

<snip>

We (LANL -long before my time) had SN#1 for 6 months I believe, for shake down testing.  Also, we're in Los Alamos.  Sandia is in Albuquerque.   Some of the science that came out of it was the discovery of cosmic ray interactions with memory.   It went back to Chippewa Falls.  Shortly after, SN# 3 came back to the lab.  An inch or two taller to make room for whatever the equivalent is of ECC memory.   #3 is in our local museum today.   I have equipment in place in the exact spot #1 lived.  I'd like to hang a picture nearby and maybe mark the floor off in some tape so folks know it was there. 

-Bob

[Bob Darlington]

Yowsers, I can't get our own history right.   That should read serial #4 came back to us after #1 returned to Cray.   #4 is in our local museum.

-Bob

[Malcolm Ray]

On Mon, 2025-09-29 at 12:22 -0500, DR wrote:

They had something called a Data General Eclipse which looks a lot like the DEC stuff, maybe they copied the idea:

Data General was started by refugees from DEC, frustrated by the direction management was taking the company.

They were certainly competing with DEC in the mini market, first with the Nova, then the Eclipse. I believe they were

quite successful, but I recall the Eclipse being rather flaky at first.

For an insight into the company, read Tracy Kidder's book 'The Soul of a New Machine'. I read it decades ago, but

I remember it being unexpectedly gripping for a book about engineering. No doubt it took a few liberties.

[DR]

Not to hijack this forum but a brief diversion, if the moderators will
allow:

How did you get one Cray out and another in?  The one I saw had a LOT of
metal frame and stuff that wasn't made flimsily.

Did the computer get taken apart (but there were thousands of wires and
connections inside the circle) so did they get it jacked up and rolled
onto a pallet?  Did it go by super shock absorbed truck?  Was it guarded?

Just some off the top of my head questions that the mystique of the
machine brings up.

Oh, the implication was the 'up time' was very good.  Do you know how
often it went down and needed to be reloaded?  The Univac 1108 we had at
Madison was usually good from morning until night but I think I
restarted one of them a half-dozen times one Saturday.

Dale

(ever curious and very afraid that as time goes by, many details will be
lost as people are not asked or even here any more.)

[Bob Darlington]

Easy, just tip it over on it's side and roll it out.  ;-)

Honestly, I don't know what they did.  My assumption is that it was disassembled just like they would do after testing at Cray before packing and shipping.  

From what I remember hearing from folks that used it and installed it, it had no operating system.   LANL (LASL at the time) developed CTSS, the Cray Time Sharing System, but past that I'm fuzzy.  From a paper about the cosmic ray interactions, they detected 152 memory parity upsets.   In regard to uptime, it's worse than you can imagine:

"During the 25-week trial period in 1976, the Cray-1 averaged

5.2 h/day of computer operation, thus the total number of hours
the Cray-1 was run was 910 h"

See: First Record of Single-Event Upset on Ground, Cray-1 Computer at Los Alamos in 1976

https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=5658018

Apologies for the distraction everyone!

-Bob

[Three Jeeps]

IIRC, the DG Eclipse was the front-end preprocessor for the Cray - I think it was called the Maintenance/Control unit.  There was a Cray 2 (IIRC) at the University of Pittsburgh  Super computing center that was shared with CMU.  I never programmed on the Cray but I attended a briefing on how to develop code for the machine.  The Cray we had, had a VAX 780 (IIRC) as the front end that did vectorization for simulations that ran on the Cray.  I got put on a different effort so I never wrote the code to run on the Cray.

DG had a uphill battle with Nova sales.  The PDP 11s were so deeply entrenched that it was hard for DG to complete (IMHO). In the late 70s-early 80s, we had so many DEC PDP 11 at CMU that we joked they were being used as door stops.  Interestingly, the Nova hardware architecture was a  predecessor to RISC machines, however the use of microcode didn't make it 'true RISC.'  Ahhh, the good old days....

[Whit Turner]

I think Oscar has the Whirlwind Pi on his roadmap. Keeping along those lines, perhaps  a TX-2 would be nice :-)

Whit

--
You received this message because you are subscribed to the Google Groups "[PiDP-11]" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pidp-11+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pidp-11/88c4addc-eef7-4f74-baa3-e8a3682a6a75n%40googlegroups.com.

[Johnny Billquist]

The Nova was release about at the same time as the PDP-11 (actually
slightly before), and it was actually rather successful, leading to
several models.

Eventually it was followed by the Eclipse, which was also a 16 bit
machine. I can't say much about that.

'The Soul of a new machine' book describes the development of the
Eclipse MV, which was a 32-bit machine compatible with the Eclipse. That
one I had a little more exposure to, and I don't have any fond memories.
Quite a mess, if you ask me.
And with that machine, I think DG struggled in competition with the VAX.

And just as one more comment on Nova, and being a predecessor to RISC.
Nova is a very obvious follow-up/expansion of the PDP-8, which is even
more RISCy.

Johnny

On 2025-09-29 22:48, Three Jeeps wrote:
> IIRC, the DG Eclipse was the front-end preprocessor for the Cray - I
> think it was called the Maintenance/Control unit.  There was a Cray 2
> (IIRC) at the University of Pittsburgh  Super computing center that was
> shared with CMU.  I never programmed on the Cray but I attended a

> briefing on how to develop code for the machine.  The Cray we had,had a

> VAX 780 (IIRC) as the front end that did vectorization for simulations
> that ran on the Cray.  I got put on a different effort so I never wrote
> the code to run on the Cray.
>
> DG had a uphill battle with Nova sales.  The PDP 11s were so deeply
> entrenched that it was hard for DG to complete (IMHO). In the late 70s-
> early 80s, we had so many DEC PDP 11 at CMU that we joked they were
> being used as door stops.  Interestingly, the Nova hardware architecture
> was a predecessor to RISC machines, however the use of microcode didn't
> make it 'true RISC.'  Ahhh, the good old days....
> On Monday, September 29, 2025 at 2:51:11 PM UTC-4 Sheepless wrote:
>
> On Mon, 2025-09-29 at 12:22 -0500, DR wrote:
>>
>> They had something called a Data General Eclipse which looks a lot
>> like the DEC stuff, maybe they copied the idea:
>>
> Data General was started by refugees from DEC, frustrated by the
> direction management was taking the company.
> They were certainly competing with DEC in the mini market, first
> with the Nova, then the Eclipse. I believe they were
> quite successful, but I recall the Eclipse being rather flaky at first.
>
> For an insight into the company, read Tracy Kidder's book 'The Soul
> of a New Machine'. I read it decades ago, but
> I remember it being unexpectedly gripping for a book about
> engineering. No doubt it took a few liberties.
>
>
>

> --
> You received this message because you are subscribed to the Google
> Groups "[PiDP-11]" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to pidp-11+u...@googlegroups.com

> <mailto:pidp-11+u...@googlegroups.com>.

> To view this discussion visit https://groups.google.com/d/msgid/
> pidp-11/88c4addc-eef7-4f74-baa3-e8a3682a6a75n%40googlegroups.com

> <https://groups.google.com/d/msgid/pidp-11/88c4addc-eef7-4f74-baa3-
> e8a3682a6a75n%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: b...@softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol

[cr...@ruffspot.net]

Cray 1 frames were shipped from the factory as one piece and were typically moved using some type of roller lifting jacks or cranes as appropriate. There were pictures of two Cray 1 systems (serial 3 then 14) at NCAR being lifted into or out of the access hatch to the subterranean computer room area at the Mesa Lab. 

Sent with Spark

--
You received this message because you are subscribed to the Google Groups "[PiDP-11]" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pidp-11+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pidp-11/99cd205e-8349-49f7-8f34-3480b15537ee%40gmail.com.

[John Hudak]

and don't forget the plumbers needed to do the install......

You received this message because you are subscribed to a topic in the Google Groups "[PiDP-11]" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/pidp-11/khfOfEroAVs/unsubscribe.
To unsubscribe from this group and all its topics, send an email to pidp-11+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pidp-11/1e55aef0-6cc0-463c-94d7-07e903f3ea87%40Spark.

[DR]

These messages are recollections and opinions from people who walked the
walk and lived the life, things which in far too short a time will no
longer be able to be documented or shared as these folks loose a voice
or are gone  entirely.


Thank you to the moderators to allow them to be shared and spark others'
having memories of things that they hadn't thought of in some time or
since that happened.


The camaraderie is what helps make these groups and forums special.

Dale


PS, I can't imagine lowering a CRAY or any other computer into  a hole
in the ground...  I assume it was to prevent prying eyes?

[dave avery]

When I went to school at IIT the Univac 1108 was in the basement of the business school building under the building lobby.  There was a section of the lobby terrazzo floor that was removable and the computer parts were lowered in by a rolling gantry crane over the hole.  The biggest unit was the FastRand horizontal moving head drum memory 

Sent from Gmail Mobile

--
You received this message because you are subscribed to the Google Groups "[PiDP-11]" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pidp-11+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pidp-11/2439b338-81a8-4bb7-98c1-bb3d29dc35ff%40gmail.com.

[Michael J. Kupec]

On Sep 29, 2025, at 2:51 PM, Malcolm Ray <m...@apathetic.org.uk> wrote:

I believe they were

quite successful, but I recall the Eclipse being rather flaky at first

My older sister ran an Eclipse for a construction company n Philly called Glascow to handle purchasing and payroll. 

Daily SOP was:

7:00 AM Power down then reload all necessary programs. 

11:30 AM Everyone log off, power down then reload all necessary programs. 

4:30 PM Everyone log off for the day, power down then reload all necessary programs then make sure scheduled batch jobs for night were queued up. 

Deviate from above and bad things happened! 😵

Have a great Day!

Michael J Kupec 

Sent from my iPhone

[cr...@ruffspot.net]

The NCAR Mesa Lab computer room was originally smaller and located on the first basement level. When it was expanded it had to be located underground on the up hill side of the building. The roof of the expansion was covered e with soil and landscaped to match the vegetation. 

Sent with Spark

--
You received this message because you are subscribed to the Google Groups "[PiDP-11]" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pidp-11+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pidp-11/2439b338-81a8-4bb7-98c1-bb3d29dc35ff%40gmail.com.

[terri-...@glaver.org]

On Monday, September 29, 2025 at 2:51:11 PM UTC-4 Sheepless wrote:

Data General was started by refugees from DEC, frustrated by the direction management was taking the company.

In the beginning (for the purposes of this narrative) was the PDP-8. It had a number of limitations (mostly regarding memory size - with enough memory you can implement subroutines for any instructions you want that the CPU doesn't supply). But not if you have limited memory.

Within DEC, an idea for a 16-bit CPU called the "PDP-X" was floated and additional design work was done. The "-X" was because it wouldn't be assigned a family number unless a decision was made to take it to market. The people whose names appear on nearly all of the PDP-X memos are E. de Castro, L. Seligman, H. Burkhardt and B. Young.

The PDP-X was neither a "wide PDP-8" or anything like the later PDP-11. The project went from (approximately) June 1967 through February, 1968. I believe the 2/68 date is approximately when the project was cancelled.

Fed up that their machine wouldn't be built, at least E. de Castro and H. Burkhardt left and founded Data General in 1968. This was taken as an insult on two different levels by DEC - that people would be so unhappy that their project was cancelled that they would quit instead of meekly accepting being assigned to other projects, and that they would have the nerve to start their own company. DEC never learned from (or got over being offended by) this experience, as shown by Prism/Mica and Windows NT.

DG started out by deciding "What can we build?" and the answer was the Nova - a 16-bit machine that owed nothing to either the PDP-X or the (future) PDP-11. Some concepts were rather PDP-8-ish, but it wasn't a PDP-8 clone, either. It was simpler than the (future) PDP-11, which made it faster for quite a few use cases.

One of DG's first innovations was to use 15" square (roughly) PC boards, so that space wouldn't be taken up by interconnects between boards. PDP-8 systems had either many small boards with a very complex backplane, or had dedicated backplanes for specific peripheral controllers, with the latter persisting in the PDP-11. The Nova was released in 1969 (along with a lawsuit from DEC). This hardened the DG founder's idea that the industry was out to "get them", and in turn they played extreme hardball with other companies that DG felt were trying to "get them" (see: Keronix arson).

The PDP-11 was introduced by DEC in 1970.

They were certainly competing with DEC in the mini market, first with the Nova, then the Eclipse. I believe they were

quite successful, but I recall the Eclipse being rather flaky at first.

The Eclipse was late to market, but was a "bet the company" type thing with all of the DG's resources behind it. Most of the teething problems happened while the systems were still on the factory floor at DG. The Eclipse was a much more sophisticated system, while maintaining upward compatibility with the Nova. By the time SPC got one of the first Eclipse S/200 systems (the S/100 and S/200 were released at the same time, with the /200 being a chassis that held double the number of cards) in 1975, it was quite reliable (compared to other CPUs of the time).

For an insight into the company, read Tracy Kidder's book 'The Soul of a New Machine'. I read it decades ago, but

I remember it being unexpectedly gripping for a book about engineering. No doubt it took a few liberties.

The MV/8000 (which is what that system was released as) was a machine nobody (other than its designers) wanted or supported. The 32-bit follow-on CPU was supposed to be developed by the FHP (Fountainhead Project) in Research Triangle, NC. This demoralized engineers back at Xborough (Northborough, Southborough, Westborough Mass) who felt that they were relegated to "raking the coals" of the dying Nova and Eclipse product lines. Tom West sold the MV to upper management as a "What if FHP runs into trouble?" and to potential MV team members as "We're going to save the company - are you interested?" and provided a brick wall between those two sides. That was where his genius was, though he was no slouch as a hardware designer either.

It turns out that the FHP did fail, the MV/8000 saved the company (for a while at least). Failure of complex new systems was by no means limited to DG - HP had their own mess with the HP3000 ("November [1972] is a Happening"). But HP recalled all systems in the field, retrenched, and eventually released a reliable system that users wanted, although a bunch of the marketing promises quietly vanished between the original announcement and the shipment of the redesigned systems.

[justme1968]

there are some very interesting posts about the inner workings and history of the first cray machines in this https://www.modularcircuits.com/blog/articles/the-cray-files/ series of postings about the development of an emulator. 

[Ken Hansen]

Crays are fascinating machines, unlike more conventional computers of their time - while there's a 'cool' factor to them for sure, I'm not clear on how it would ever be anything more than a "blinkinlights" shelf decoration, they just aren't built for use in a conventional (sitting at a terminal, interacting, running end-user programs) AFAIK

Ken

On Sep 30, 2025, at 09:02, 'justme1968' via [PiDP-11] <pid...@googlegroups.com> wrote:



--
You received this message because you are subscribed to the Google Groups "[PiDP-11]" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pidp-11+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/pidp-11/1ba53c20-0f4a-4453-b4ff-295925d76d39n%40googlegroups.com.

[pbi...@gmail.com]

FWIW the first Eclipse model and Nova 4 (last Nova-line model) shared the same CPU hardware ... right down to the PCB.  That was because the Eclipse introduced bit-slice design and the differences between these two CPUs was merely microcode (the Eclipse used more -- both to support extension of the instruction set and to consolidate functions previously requiring separate boards).  Prior Nova-line models used '181-based ALUs and a mix of random logic increasingly concentrated into small ROMs and PALs.

The Nova 4 was not a planned product; the Nova 3 was intended to be the end of the Nova-line CPUs.  However customer demand (aka "sales resistance" to the introduction of the first Eclipse model) necessitated a temporary retreat ... fortunately quite possible given the bit-slice hardware design and the inherently upward-compatible instruction set design of the Eclipse.  if you wanted to add memory management on the Eclipse you paid for a different set of microcode ROMs; there was no change in the actual hardware other than plugging in more memory boards.

If one wanted to make a CPU design comparison between the original Nova and the PDP-8 it would be to the 8/S.  The first NOVA was nibble-oriented thus requiring 4 passes to produce a 16-bit result.  This is what allowed a single 15x15" PCB to hold a complete CPU.

If one wanted to make a system design comparison between the Nova-line computers and the PDP-11 family it would IMO revolve around the electrical design of the backplane.  The PDP-11 was rigorously a bus-based design -- initially the Unibus.  The Nova-line followed more a seat-of-your-pants design evolution; while there was a well-controlled I/O bus that enabled a wide range of I/O controllers to be shared across all Nova-line and then Eclipse computers ... the memory interface changed from model-to-model throughout the core-era.  It didn't really gel into a fixed design until the Nova 4/Eclipse era and the switch to an all-MOS memory environment.  DG constantly fiddled with the CPU-Memory interface to gain the best possible system performance at the lowest cost-point in the core-era.  In consequence the Nova - Super Nova - Nova 800/1200 - Nova 2 - Nova 3 generally require different memories ... an annoying property if you were "raised" on the PDP-11 family.

On Monday, September 29, 2025 at 8:06:11 PM UTC-4 terri-...@glaver.org wrote:

On Monday, September 29, 2025 at 2:51:11 PM UTC-4 Sheepless wrote:

Data General was started by refugees from DEC, frustrated by the direction management was taking the company.

...

The PDP-11 was introduced by DEC in 1970.

[terri-...@glaver.org]

On Tuesday, September 30, 2025 at 3:41:12 AM UTC-4 pbi...@gmail.com wrote:

FWIW the first Eclipse model and Nova 4 (last Nova-line model) shared the same CPU hardware ... right down to the PCB.  That was because the Eclipse introduced bit-slice design and the differences between these two CPUs was merely microcode (the Eclipse used more -- both to support extension of the instruction set and to consolidate functions previously requiring separate boards).  Prior Nova-line models used '181-based ALUs and a mix of random logic increasingly concentrated into small ROMs and PALs.

I do need to respond to this. The early Eclipse models through the S/230 were all '181 based. The closest match to the Eclipse S/200 would be the Nova 3, also '181 based, although the boards were very different.

The first common Nova / Eclipse implementation was the Nova 4 and the Eclipse S/140, which were 2901-family based.

The Nova 4 was not a planned product; the Nova 3 was intended to be the end of the Nova-line CPUs.  However customer demand (aka "sales resistance" to the introduction of the first Eclipse model) necessitated a temporary retreat ... fortunately quite possible given the bit-slice hardware design and the inherently upward-compatible instruction set design of the Eclipse.  if you wanted to add memory management on the Eclipse you paid for a different set of microcode ROMs; there was no change in the actual hardware other than plugging in more memory boards.

Per my comment above, the Nova 3 and Eclipse S/200 were contemporaries. The Eclipse line continued to evolve up to the S/230 with '181 ALUs. The S/140 came out at a time where there had been no new Nova processors for some time, and rather than force people to pay for CPU features they didn't need (this was long before the era of "pay to enable functionality already in the hardware" had trickled down to 16-bit minis, although IBM had been doing it for years on 370-class CPUs), the lower-cost Nova 4 was created. The Nova 4 and the S/140 were also the first "soft" front panel in discrete logic systems. While DG did have a MicroNova CPU on a chipset (later a chip), DG didn't continue to evolve it the way DEC did (where it took over the entire then-current PDP-11 product line). The MicroNova also had (by necessity) a soft front panel.

DG continued to evolve the 16-bit Eclipse line, with the last  CPU model I know of being the Eclipse S/280. The later Eclipse models were targeted at the AOS operating environment, while earlier Eclipse systems (and many Nova configs that had hard drives) ran RDOS. AOS was extended to the MV/8000 and later 32-bit models as AOS/VS.

Along the way, there were two "odd birds" - the C/350 (which confusingly came out before the S/230) which was intended as a commercial machine. I believe it mostly ran RDOS with Business BASIC (the S-series Eclipse models ran Extended BASIC). There was also the M600, which I believe was an AOS-only machine. During one visit to DG to drop off my code updates*, I asked what the rationale for the M600 was, and the answer was "darned if we know".

If one wanted to make a CPU design comparison between the original Nova and the PDP-8 it would be to the 8/S.  The first NOVA was nibble-oriented thus requiring 4 passes to produce a 16-bit result.  This is what allowed a single 15x15" PCB to hold a complete CPU.

The 8/S was a bit-serial machine made out of lots of smaller modules, with a complex backplane unique to that model. The early Novas were indeed nibble-serial, but the backplane established by them continued with minor changes through at least the Eclipse S/230 era. Backplane changes were needed to support MOS memory and Eclipse memory management. 

 

If one wanted to make a system design comparison between the Nova-line computers and the PDP-11 family it would IMO revolve around the electrical design of the backplane.  The PDP-11 was rigorously a bus-based design -- initially the Unibus.  The Nova-line followed more a seat-of-your-pants design evolution; while there was a well-controlled I/O bus that enabled a wide range of I/O controllers to be shared across all Nova-line and then Eclipse computers ... the memory interface changed from model-to-model throughout the core-era.  It didn't really gel into a fixed design until the Nova 4/Eclipse era and the switch to an all-MOS memory environment.  DG constantly fiddled with the CPU-Memory interface to gain the best possible system performance at the lowest cost-point in the core-era.  In consequence the Nova - Super Nova - Nova 800/1200 - Nova 2 - Nova 3 generally require different memories ... an annoying property if you were "raised" on the PDP-11 family.

If you were changing CPUs, you changed memory on either the PDP-11 or Nova/Eclipse. I did a in-the-box upgrade from core to MOS in SPC's S/200, and then did my own MMU (and later CPU) re-implementations** to end up with something that outperformed an S/230.

The problem with connecting peripherals from their connectors to the actual controller was that each controller came with a generic "paddleboard" connector that was stacked on the rear of the chassis. These paddleboards had many, many identical white wire-wrap wires that just had small number "flags" on them. To adapt a generic paddleboard to a specific peripheral controller required wire-wrapping each of the paddleboard wires to a specific pin (which varied by controller model) in a backplane slot. If you needed to shuffle controllers in the slots, you'd cut and unwrap the cut ends, toss the old paddleboard, and wrap in a new one.

By the time of the Nova 4 / S/140, this had gotten ridiculous enough that DG adopted a (mostly) standard backplane interface, and the paddleboards were gone, normally replaced by D-series connectors for peripherals. DG had also come out with a pre-wired "comm chassis" for their ALM-series multiplexors, which replaced much more primitive designs like the (IIRC) 4064. I also added a comm chassis to what we began to call the "Frankenclipse". 

* By this point, I'd submitted a bug report (called an STR) reporting something utterly bizarre: "RDOS Extended BASIC can be sysgen'd for any 32 terminals, as long they're the first 32"***. That resulted in my STR coming back, wrapped around a copy of the source code for XBASIC, with a bunch of other people's STRs in the box, along with a "If you could look at these too, we'd appreciate it".

A subsequent STR complained about the lack of RDOS device drivers for the 2MB fixed head-per-track 6064  "AOS paging disk" that I'd purchased, along with the 6061 removable pack disk drive (roughly the same size and capacity drive as the DEC RP06). That STR came back with the source to RDOS inside a wrapper made out of my STR, along with the usual other people's STRs.

** Having created a screamer of a system with the stock CPU (the AOS paging disk held RDOS and its overlays and XBASIC and its overlays, making disk I/O essentially instantaneous - 1 revolution of the paging disk) and moving user files off to the Zebra meant there was no system disk contention by user programs.

I'd made a bit of a nuisance of myself, and also having helped DG clear some STR backlogs, I turned my attention to the CPU and MMU. By this point, DG was pretty much "sell those weirdos at SPC anything they want, but unless it's a complete peripheral subsystem, tell them we won't support it". I ended up with some DG prototyping boards and went wild. Every once in a while we'd have something like a head crash on the Zebra and DG would come out. 

Most of the regular service guys were used to me and would just fix the drive, but occasionally I'd have to break in somebody new who wanted to run CPU diagnostics first. They'd invariably start with CPU diagnostics for the S/200 (which would fail) and they'd say "Your CPU's broken!". I'd say "No, you need to run the S/230 diagnostics and patch the timing loops at these locations". They'd do a double-take, look at the CPU that said "S/200" and had incandescent lamps instead of the S/230's LEDs and start head-scratching. I'd say "Let's run a couple passes of EMORT L and call it a day." and that almost always worked.

*** The reason for this was that we were running 32 terminals of XBASIC in the foreground partition and another 32 in the background partition during the day. A homemade batch queuing system let users edit and submit batch jobs in the other languages supported by RDOS. In the evening we'd take BG XBASIC down and run through the user-submitted batch jobs, spooling the output to files in the user's XBASIC login directory for them to review the next morning. This took full advantage of XBASIC allowing user-written assembler routines to be incorporated into the XBASIC interpreter.

When the SPC computer center moved to its own building in 1983, new DEC PDP-11s were waiting to be uncrated and installed. We announced the impending retirement of the Frankenclipse and had a mass revolt by our users. Even after we went to VAXes in 1986, the Frankenclipse happily hummed away in its small corner of the computer room.

Finally, it's long-postponed shutoff date came in 1988. SPC gave everything - hardware, software, schematics of my custom CPU and MMU, etc. to someone who swore he'd preserve it. Unfortunately, the guy who took it literally dropped off the face of the earth within 6 months - the point where his wife had to have him declared legally dead. I helped her sell off his collection of radio and computer gear that had been in their house, but there was no sign of the Frankenclipse or its components. Presumably it languished in a storage unit until being auctioned off as "unknown contents" - a crime against history if there ever was one. The Frankenclipse now lives on only in my memory and the memory of maybe two other people who played some part in creating it with me, other than a few documents I saved:

Attached find a copy of the support escalation booklet as well as part of a quote DG tried to give me for maintenance on the CPU before I re-designed it. Note the number of core and MOS memory boards in the same CPU (an impossibility),and the missing or made-up serial numbers for most of the system. 

[pbi...@gmail.com]

"Poking the bear" works!  Technical specifics regarding the evolution of the Nova/Eclipse line of 16-bit CPUs is hard to come by, particularly so on the Eclipse-era.  Thank you for amplifying and filling in various points; much appreciated :-}.  Now if I could find someone with a DG JR SBC who could make copies of the four "core" microcode ROMs for me -- mine is missing one, alas.  So, effectively a wall hanging as matters currently stand.  Shameful!

On Tuesday, September 30, 2025 at 5:12:20 AM UTC-4 terri-...@glaver.org wrote:

On Tuesday, September 30, 2025 at 3:41:12 AM UTC-4 pbi...@gmail.com wrote:

FWIW the first Eclipse model and Nova 4 (last Nova-line model) shared the same CPU hardware ... right down to the PCB.  ...

[DR]

Indeed.

I don't think a Teletype was ever connect to one to play Hamarabi!

The mystique about big, fast, rare and doing jobs like modeling airflow
in the atmosphere or inner workings of an atomic reaction all added to it.

I think now about the reports coming out of the CERN accelerator and the
millions of collisions per second being tracked and analyzed wondering
how many CRAY equivalents it would take to keep up with that information
flow.

[Lawrence Fisher (RealTimeCat)]

Redhat had a  podcast a few years back called "Command Line Heroes."

https://www.redhat.com/en/command-line-heroes/season-4 

Season 4 covered some of the history and rivalry between Data General and Digital. The Soul of a New Machine was used as a major reference for that podcast.

Season 1 gave some interesting info on "OS Wars".

[Bert Driehuis]

The Cray at CERN wasn't getting the utilisation they had hoped for. When I was visiting a friend who worked there, I was offered access to it. Few of my problems lend themselves to parallelization, and FORTRAN did not appeal to me, so I politely declined the opportunity. People looked at me like I had just turned down a winning lottery ticket.

I had access to a VAX/VMS cluster at one job and a nice PDP11/23 that I could practically use as my Personal Computer at another. Both had a C compiler, so the Cray offered nothing of value to me.

The trip down memory lane the PiDP11 project has unlocked is incredible!

--
You received this message because you are subscribed to the Google Groups "[PiDP-11]" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pidp-11+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/pidp-11/21f06a40-0a30-420b-b562-adfb42d92868%40gmail.com.

[DR]

I'm not trying to hijack the forum, honestly, and hope I am clearly
labeling subjects so those without interest will be able to easily skip.


I found this message from one of  the engineers who worked there, and he
gave this explanation of how a UPS of sorts worked back then.  Thank
goodness for battery backup now!


"

This is a easy one to answer and so simple that it really makes one stop and think about how simple solutions are many times overlooked!
The Cray system ran on 460V three phase 400 Cycle power, that allowed our power supplies to be roughly one seventh the size they would have had to be if we ran on 60 cycle. To generate the power we used three, seven hundred fifty horse power motor generators, two on line and the third running on the pony motor without the exciter turned on. The generators spun at 6,000 RPM and were wired for four poles so we got the 400Hz power. Now how did we do UPS? The rotors in the motor-generators weighted about 7,200 pounds and we had three of them spinning at 6,000 rpm which gives you a total of 21,600 pounds of metal moving right along! We used a intelligent power monitor on the 480V 50/60Hz primary power that was used to power the motor-generators and the disk drives. If we seen a phase drop the system immediately issued a instruction to the Cray operating system to unload memory to the drive field and shut down the system. Cooling would continue as the control system for cooling was all mechanical using expansion valves controlled by local gas pressure sensors; the compressors for the refrigeration were designed with liquid Freon reserves that would more than take away any latent heat. [The latent heat problem was there every time you shut down the computer even in normal use] The disk drives were protected from head crash several ways, first of all the rotating mass was very large, each disk had a stack of eighteen, eighteen inch disk platters that turned at thirty six hundred RPM and they were spun with a direct drive six HP servo motor that had a power supply with a large capacitor bank that could carry through for a period of time. There were four sets of heads, each head set had thirty six heads, the head sets were located around the disk pack at ninety degree points and the moved in and out in a linear movement with heads located at one hundred eighty degrees from each other so as to balance the forces and reduce vibration. This arrangement allowed us to write a full cylinder with every ninety degrees of rotation and our disk system used gather-scatter so we wrote to every drive on the system at the same time! The power supplies that drove the head moving coils also were equipped with grossly over sized filter cap banks to supply power as the unit powered down and extended the heads to the parking track near the center of each disk where rotational velocity would be the least. Once the heads were parked the servo motor windings on the drive pack were shorted and the pack came to a rotational stop very quickly, there was also a shipping brake that was built into the servo drive that was applied when the rotational speed dropped down to hold the disk pack from spinning when the drive was not powered up.

Our UPS had no batteries, just ten and a three quarter tons of iron spinning at 6,000 RPM
"

[sh...@oneil.me.uk]

You might be entertained to find these still exist - a flywheel UPS, or flywheel energy storage.  Besides low maintenance cost (you go through bearings instead of batteries!), one of the main things that makes them unique is that they recover a full "charge" in seconds - they don't have to recharge shelves of batteries, they just need to come back up to speed.

[DR]

I learn something everytime I read these posts.

[terri-...@glaver.org]

Many of the smaller IBM 370 models used the 3046 motor-generator set to produce 400Hz for the CPU. I caught one of my operators starting and stopping it repeatedly (via the 370's Power On/Off buttons) because on startup it made a sound like firing the phasers on Classic Star Trek.

Many commercial airplanes also used 400Hz for more efficient power distribution (but without the heavy, bulky converters - It turns out that jet engines have all of the necessary parts and are spinning anyway - hopefully, at least). 

[oscarv]

More replicas (or computer history capsules, I start to speak about these things in conceptual terms more and more - not good):

Yes! 

A Cray replica has become much more interesting as there are now good simulators with the proper OSes running on them. IIRC, you can even install a Linux on the Cray Y-MP.

I am not going to do it, but I am trying to talk a professional designer into doing it. One day she will give in, and with her skill set, it will be a real design piece.

An IBM 360 and 370 is high on my wish list too, and there too I hope that some people will step into the breach. Rich Cornwell already indicated that he'd be up for it. I need to organise myself a bit here (brain fog puts a brake on everything I do), but the first step is to send some original IBM parts to the switch manufacturer we use. Their R&D guy is happy to help with making replicas of its switches, dials and buttons.

Personally, I want to spend a couple of months 'polishing up' the four existing PiDP's, as I find that gives me perhaps more satisfaction. It helps keep them alive by creating more activity around them. And then the Whirlwind is the one I must/need/want to do with Guy Fedorkow.

Not everyone will know the story, but I think it's relevant now, because things really start to happen:

Three years ago I handed over the kit-making to my friends in Panama, Otto and Jose. So there is actually a little micro-factory that can be used as a resource for doing everything after the design of a replica has finished. That makes it very feasible for others to do replicas too - just let them do the production, according to your ideas/plans/design. In that mode, I did the PiDP-1 with Angelo Papenhoff. And Juergen Mueller brought in his Enigma and an upcoming LGP-30. But more could be done, plenty of space in that Panama warehouse we now have for this. 

The projects budget got close to zero with the PiDP-10 for a while, but that is coming back. And with Otto joining the effort, we have not only a budget again but also expertise in metal-parts making etc. 

In fact, this week I managed to tempt Henk Gooijen (pdp-11.nl) into a new project of his.  Just to whet your appetite: Think of these peripherals, powder-coated sheet metal, 

and built into a proper PiDP-11 rack (there's a youtube movie of last week):

https://obsolescence.dev/rack11.html

This WILL happen! Henk's prototype, neatly racked with the PiDP-11. Maybe 2026, maybe 2027, but it will. And not cost the world either.

So, the downside for me is that I have this stupid health issue of mine, and it limits me. But handing over production to Otto and Jose has opened up a much bigger capability. Otto's hobby is to deal with all sorts of small manufacturers (steel, aluminum, etc) and talk them into doing custom parts. Even though we're clearly not serious-volume clients for them, so much more is possible than I thought. He's working with his son on his own replica project, but he prefers not to talk about it until he's got something to show... but next yer, he will :-)

Why all this talk? If you feel tempted to design something in the computer history space (pre microprocessor era), this Panama micro-factory can really help out. If you guide Jose and Otto into how to make it, they will take care of the logistics and everything else. And this is not very commercial, so the cost is kept low.

Anyway - I've been away from the PiDP-11 Google Group for too long as 110% of my clear-brain hours went into the PiDP-1 project, but it is maybe a good moment to bring this all up. 'Obsolescence Guaranteed' has grown to include a whole bunch of people by now.

Kind regards,

Oscar.

[DR]

Oscar:

First, best wishes for  your health and recovery.  I have a wee bit of
fogginess also but my major problem is physical arthritis and strength
and fatigue, so I know of the frustrations.

The awareness of what one used to do only amplifies the frustration. 
Completely different as far as cause, but it reminds me a little of
Flowers For Algernon, if you've read that story.


IBM:  I think that scientists and engineers might not have as much
interest in making a tool such as a computer look pretty, but the
marketers and those wanting their business or university to have a
science fiction like computer on display no doubt had some slight
influence.  I worked at the Univ of Wisconsin Computing Center on Dayton
Street in Madison, Wisconsin and recall that the whole second floor
surrounding the computer floor looked down on the 1108, the tape drives,
the printers, the CalComp plotter, the Burroughs B5500 and a few DEC
PDPs and teletypes.  I would look down on that before I was hired to
work there an once working there, look up to see people just watching
the gizmo work.

The 360s were scattered over other installations.  I know the urge I had
to just push those fantastic IBM lighted buttons.  They were a piece of
art, I thought.  I  never have found any of them on eBay or for sale so
I hope your wizards at making switches can capture the solid, opaque
switches but certainly far more the back-illuminated ones.  I found two
IBM cell washers/separators for sale about 100 miles away and when they
weren't moving the owner let me have them for a few bucks.  I took the
switches out and a friend recycled the rest, which had a lot of weight
(over 400# each) and some interesting machined parts, pumps and motors
in them.  I finally have a few of the switches I lusted after for so
many decades.  Like all things, they aren't perfect but a fond reminder
of the 360's panel.


Be well, and know that at least here you have a supporter of your effort
to not only maintain history but allow quite a few people to recreate
the mystique and feeling of awe that these machines carried when they
were 'hot stuff' back then!

Dale

[terri-...@glaver.org]

On Wednesday, October 1, 2025 at 5:12:04 PM UTC-4 oscarv wrote:

An IBM 360 and 370 is high on my wish list too, and there too I hope that some people will step into the breach. Rich Cornwell already indicated that he'd be up for it. I need to organise myself a bit here (brain fog puts a brake on everything I do), but the first step is to send some original IBM parts to the switch manufacturer we use. Their R&D guy is happy to help with making replicas of its switches, dials and buttons.

As I think I mentioned the last time a "What future PiDP systems would you like to see?" came up, the IBM 360 and 370  systems mostly displayed the internal machine state for a) system programmers (both IBM and customer) to glance at to get a rough feel for what the system was doing and how healthy it was, and more importantly b) for service people to figure out what was wrong with a CPU when it started misbehaving. Also c) the lights looked darned impressive to people looking at the hardware - a blank-faced machine would likely have triggered a "We paid $500,000 (in 1970's dollars) for *THIS*?!?!

I don't know of any cycle-exact 360/370 simulators (which doesn't mean there aren't any), but even that would get you very little in the way of realistic front panel lights - basically address and data. You have some very boring front panels (basically power on/off, IMPL, IPL, interrupt and a bunch of check lights) on the later "soft console" 370s, and some rather extreme front panels (the 91/95/195 CPUs being a perfect example - just about every important logic signal appeared on the front panel). Then you have transitional machines like the 138 that had both the traditional lights+switches and soft console on the same system.

These systems are more complex and at the same time more primitive than the existing PiDP-8, -10, -11 and even the -1. The only resource the service tech had to fix a balky CPU were the front panel lights and switches, the ALDs (carts of giant blue binders with "ASCII art" circuit diagrams of each board), and (if the machine was healthy enough and the hardware supported it) a special diskette with microdiagnostics instead of 370 microcode.

Even with those resources, 'my' 370/138 was down for a week as higher and higher levels of service escalations happened in IBM, culminating with the designer of the component in question (the IFA) flying in to diagnose the problem. And the 138 is a relatively simple lights-and-switches CPU. Refer to the attached picture. The silver knob on the right side of the CPU cabinet rolls a sheet of plastic film across a single row of indicators. The film has different legends for each possible position of the knob, and the lights in the row change accordingly.

The PiDP-11 "cheats" the uADDR display setting as SIMH doesn't run the original 11/70 microcode. That's a minor cheat as the front panel normally isn't displaying that info anyway - it's normally displaying address / data and machine mode (16/18/22, U/S/K, I/D, etc.) which is available from the simulator as it needs to keep track of those in order to function. The PDP-1/-8/-10/-11 all had "programmer" front panels which were intended mainly for users to interact with the system and see the results. The same would be true of the Altair/IMSAI type systems. On the 360/370, the machine was so expensive (either in up-front purchase or leasing) that letting someone sit in front of it and single-step the system would be a very expensive last resort. When the IBM service person came out and had to resort to that, they used their "CE key" to have the time charged to the service account and not the customer account. [Yes, I know the hour meter only ran when the CPU was executing instructions, but anyone who would make that distinction already understands my point]. You can see the two separate time meters on the front of the CPU above the telephone.

Probably the closest minicomputer analog of the 360 / 370 front panel would be the "Classic" HP 3000. It had a pair of '2001-style monoliths' with lots and lots of LEDs. The artwork on the prototype I saw was pretty primitive, being stenciled paint covered with varnish. Those were so rare even "back in the day" (HP recalled every single system) that I can't find a good picture of the trimmed-back production monoliths, although this one is the closest:  https://www.hpmuseum.net/images/3000_1972-PromoPhoto-24.jpg

If I had my choice of future PiDP system, I'd like a Data General Eclipse S/200 for sentimental reasons. There probably isn't enough demand to actually create one, though. But it has a mini / micro type programmer's front panel and not the maintenance front panel you see on IBM systems, so it would certainly be do-able.



   

[Adam Thornton]

If you can get rcornwell to do the simulator for you...the 360/195 has an amazing front panel.  But it was mostly for service engineers and doesn't help during regular system operations much.

I'm more interested in a software standpoint in one of the 370s because I grew up as a systems programmer on a VM/ESA system and I love VM (VM/370 is what I run on Hercules at home, and I just don't care for MVS and predecessors very much), but the 370s didn't have nearly as lovely front panels as the late 360s.

Adam

--

You received this message because you are subscribed to the Google Groups "[PiDP-11]" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pidp-11+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/pidp-11/650efd5b-5434-4224-ad3a-e7114bf34afcn%40googlegroups.com.

[oscarv]

On Thursday, October 2, 2025 at 5:23:57 AM UTC+2 terri-...@glaver.org wrote:

I don't know of any cycle-exact 360/370 simulators (which doesn't mean there aren't any)

Richard Cornwell did one! See  https://github.com/rcornwell.

But yes - the IBM front panels mostly are about diagnostics.

If I had my choice of future PiDP system, I'd like a Data General Eclipse

I had planned one after the PiDP/8, I even got a real front panel to work from. But at the time, I could not find any OS or other software for it. And gave the panel away - to a good cause, I forget who but he had restored a whole Eclipse and was missing the front panel. But I think, indeed, it might be too much of a niche. The PDP-12 is in that category as well, and my personal choice would be a Linc. And (from my native Holland) an Electrologica X8. The first machine with a stack pointer, and Edsger Dijkstra wrote its software. So many things left to do!

Kind regards,

Oscar.

[oscarv]

The dilemma to solve for a 370 replica is that you probably want to have it with a dual heart. Rich Cornwell's simulators for authenticity, but then also a boot mode with a modified Hercules - because every IBM person I meet (limited sample count I have to admit) is used to Hercules. You'd end up with proper replica behaviour (Rich's simulators) and another mode with a non-period correct MVS/Hercules: running on an older 370 that never could do MVS.  A bit like the PiDP-10 which does run TOPS-20, even though a KA10 never could. But Rich wrote the KL10 simulator and was kind enough to let it drive the KA10 front panel :-)

Kind regards,

Oscar.

[terri-...@glaver.org]

On Thursday, October 2, 2025 at 12:23:47 PM UTC-4 oscarv wrote:

I had planned one after the PiDP/8, I even got a real front panel to work from. But at the time, I could not find any OS or other software for it. And gave the panel away - to a good cause, I forget who but he had restored a whole Eclipse and was missing the front panel. But I think, indeed, it might be too much of a niche. 

The LSSM has a DG Eclipse-style lights-and-switches front panel. I forget what system it is from. I'll have to pull it from the warehouse the next time I'm there.

The software is almost all out there, courtesy of Bruce at Wild Hare: http://www.novasareforever.org then select Archives / Software / DG Nova and Eclipse 

[DR]

Oscar:

I have attached a picture of one of the coveted IBM switches from the era of high-design and snazzy looks for computers.

This one looks more opaque, which some of them were, but this one is lighted and more translucent.  I wish I had powered it up to show the look it gives, but I'm sure you know exactly what the appearance is from having been around them before.

Anyway, I'm hoping  you can find a switch guy to reproduce the fixed indicator lights and the push to do something switches which were also illuminated.  They were amazing works of art, in my opinion.  Dale