|This is the front of the Processing Element Bay. Each row could hold 9 PE. We built and wired the entire cabinet but only built 11 PE so we could field a row of 9 PE and have a few spares. A full up PEPE could control up to 8 of these PE bays for a total 288 PE. The large box at the end of each row was a power supply for that row. No more Element Bays or PE were ever built.|
|Processing Element Bay backpanel view. The power supplies were water cooled. The PEPE control console was in another cabinet. It sent control signals over the ribbon cables to each PE.|
This is the control console. It contained the interface to the CDC-7600 computer used for actual operation and to the B-1700 used for development and testing.
|Control Console with doors open.|
|Power supply. We had a fork-lift sort of device to help us remove the power supplies for maintenance.
It was a switching supply. The diodes in the supply were a real weak point. We had alot of failures of those diodes.
Cooled by chilled water.
|Control Console control panel. The LED display showed how many of each of the three units in each PE were online at any time. Notice the external clock BNC connector. PEPE was completely synchronous and could be operated with a clock speed of 10MHz down to a push button rate.|
|This is is the rear of the control panel.|
|This is the backpanel of the Control Console.|
|This is the interface logic.|
|Front of the PE bay. The top row contained 9 PE. It looks like the second row contained 2 PE.
The card on the far left was for clock distribution to the row.
|Backpanel of the Processing Element (PE) Bay.
Note the pattern of daisy chained cables. These cables distributed control signals from the Control Console to each of the 9 processors in each row for a total of 36 processors per bay. The Control Console could then control up to 8 PE bays for a grand total of 288 processors. IIRC we built 9 Processing Elements plus 2 sets of spares and one PE Bay.
|Typical PC board of PEPE.
This, I believe, was the Memory Board of a processing element.
|Here is a photo of a typical board showing some of the features. I believe this is the Arithmetic Unit (AU) of a processing element. The 24 pin IC were a 4-bit bit slice Arithmetic and Logic Unit (ALU). It appears there were 6 spare IC locations on the board.|
|PEPE Card/Element Tester. This box could test any of the boards in PEPE.|
Test and Maintanence Control Computer
This is the interface logic which connected PEPE to the B-1700 - the test and maintainence interface.
|Control Data 7600 - PEPE's main operational interface|
|System Development Corporaton hosted the BMDATC (Ballistic Missile Defense Advanced Technology Center) in Huntsville, AL.|
PEPE Match Book
|Click on the scan, above, for a pdf of the entire brochure. In it are photos of Bob Sidnam and Benny Sisk, the two people at SDC that we worked with. Also includes photos of Howard Welsh and Jerry Schweitzer (who worked with me at Burroughs).|
From the July 1977 Burroughs B-Line, an in-house magazine of the Federal and Special Systems Group.
|Date: Wed, 02 Jun 2004 21:32:13 -0400
From: John Sternbergh <johnastrapoint.com>
Subject: A Voice from the Past
I was searching for information about the Burroughs' PEPE system on the web, and I found your name and email addresses. It was a pleasant surprise.
I hope the years since we've seen each other (all 28 of them) have treated you well. Tacie and I now live in Durham, NC, and our daughters live near us. I'm currently employed by Blue Cross and Blue Shield of North Carolina, and I'm enjoying the challenge.
Please drop me a line if you're so inclined. It would be nice to hear how you're doing.
Great to hear from you. It certainly has been a long time. The last I remember, you were working for SMS across the street from the old Burroughs plant.
I still keep in touch with a few of the old crew; Chuck Yonko, Larry Watson and Jim Dix. Chuck and Larry both live in New York, near Binghampton and Jim still lives in PA. Chuck and Larry work for remnants of IBM which were sold to other companies. All three have been out here and I've seen them at various times over the years. I've written to Carl Semmellhack several times but he has never written back. Paulo MacCormick retired from Burroughs maybe 15 years ago and now lives in Miami, FL.
I've been in San Diego since leaving Burroughs in 1978. I worked for five years for National Semiconductor and then had my own company, Silogic Systems for almost 20 years. We did chip design. I got "bought" by one of my clients in 1999 and went to work for them for a few years. Last year, I left there and now I'm either retired or between jobs.
About 10 years ago, I made an attempt to find PEPE. I traced it from Huntsville, to Kwajalein Island to Auburn University in Alabama. I finally actually spoke with someone who was involved in the "junking" of PEPE. It was completely gone and he didn't know any more of what became of it.
I noticed that searching on the web, I find refrences to PEPE in "History of Computing" and other such sites. I guess we've crossed over the line into being "olde timers".
My current time is wasted by genealogy research.
I haven't been back to eastern PA since I left. I imagine it has changed a LOT!
I hope this email gets to you and that you are doing well in your home. San Diego will never be the same.
I'm putting together a web page about the PEPE computer and I was wondering if you have any photos of GreatValley Labs?
|Date: Sun, 16 Sep 2007 15:35:20 -0600
From: Alan Whiteman
Subject: Re: Burroughs GVL
Hum ... I don't think so - much was lost and discarded when I was packing up and leaving in those frantic last few weeks.
|Date: Wed, 21 Apr 2010 11:51:43 -0700 (PDT)
From: Nina Cornell <nina.cornellyahoo.com>
Read your site on PEPE. Wondered why you never mentioned Bell Labs (Dave Bergland) and later John Cornell, Project Manager at SDC in re the origins of the computer.
SDC bid and won the contract to build the 13 element processor as conceived by Bell Labs Whippany. A group from SDC, Santa Monica, Ca. headed by John A. Cornell relocated to Whippany and after a one year transition the group relocated to Huntsville.
Thanks for your email.
In my lowly position on the PEPE project, I didn't have much contact with the important people at SDC. We worked with Benny Sisk and Bob Sidnam.
I looked at the final report of Phase I of the program and I see the names: J.A. Cornell, G.J. Hanson, and R.G.Mueller. This document was dated 30 Jun 1973 shortly after Burroughs got involved and I got hired.
I would like to learn more about how the project was run during the early phases. We did our design work based on design speciifcations written by Honeywell.
If you have a chance to tell me more, I'll add it to the web page.
|From: Walter Wikiera <wjwikieramsn.com>
Date: Sat, 22 Sep 2012 07:33:47 -0400
This is Walt Wikiera retired. I happened to see your article on PEPE and it brought back good memories. This was one of the better programs in my career at Burroughs and working with you was a pleasure.
I presently live in Florida. I retired from Burroughs (Unisys) in 1990.
Best of Luck,
Thanks for your email. My web page keeps on attacting people from the old group at Burroughs.
Over the years, I've kept in touch with Bill Sullens, Larry Watson, and Chuck Yonko.
Two people that I would have liked to have heard from are Carl Semmelhaack and Herb White (my boss). Never heard from them.
PEPE was a good project, also one of the best projects that I worked on over the years. Sadly after the 1980's, companies weren't making large computers anymore. I spent most of my career designing ASIC's.
Would you like to write up your memories of the PEPE program? I could add that to my web page.
|From: Walter Wikiera
Subject: PEPE remembered
Date: Tue, 25 Sep 2012 09:55:03 -0400
Looking back on the PEPE organization I realized that we had some of the best talent a project could have. I did not remember that Don Faulis was part of that team until I saw him on the organization chart. He is one of my best friends and I have visited him several times in California. No one was a better mechanical engineer than Dick Stotler. Herb White was a meticulous and talented electrical engineer. As for Carl Semmelhaack, he was a brilliant software/system engineer, although if you didn't know him, he didn't come across as impressive. There were others that made the project as successful as it was: yourself, Bert Beckmann, Elmer Duckinfield, Harry Hess and many many others. Mark, it was a very talented team with much experience but it was adding new thinking that made it go. I would be remiss if I didn't add SDC personnel such as Bob Sidman to the PEPE team for it's success. I've worked on other good projects, but none more pleasurable than with the personnel on PEPE team.
|Date: Thu, 2 May 2013 16:22:02 -0500
Subject: PEPE at Auburn
From: Kyle Owen <kylevowengmail.com>
First of all, I have thoroughly enjoyed your website dedicated to PEPE. That was a very interesting time in computing history for sure, and your website really ties it all together very well. Thanks for the great pictures, too!
I see you have mentioned Auburn University as the final resting place of PEPE. I'm currently a to-be senior in computer engineering there, and as a vintage computer collector, I decided to do some digging on the big machine. Sure enough, several of the guys in the department remember PEPE quite fondly. I'll be trying to get in touch with a few of the guys that actually programmed it while it was partially working at the university perhaps over the summer.
As you are definitely aware, it was in fact scrapped. All of the 7400-series ICs were desocketed and stuck in drawers, only to (rarely) be used again in some digital electronic labs and such. One of the other remnants is a Processing Element Bay badge stickered to a 19" rack. I'm sure the copper and what not from those power supplies were recycled, and apparently the hydraulic lift used to remove large parts from PEPE is still being used around the department, from what I understood.
I figured you may be interested in a picture. It's not much, but maybe it'll help tie up the bottom of your webpage or something.
Kyle Owen, W4GNU
Kyle noted that the rack cabinet to which the badge is attached was not part of PEPE.
If you are in Huntsville, you could try to find the old System Developments Corporation office (which I think was bought by Burroughs, which is now Unisys). Bob Sidnam and Benny Sisk were our two main contacts with that company. They would both be in their 70's and certainly retired. Jerry Fowler was one of our techs who stayed with PEPE when it was left in Huntsville and he might still be around somewhere. He would probably be in his 60's.
Besides the two main PEPE cabinets, there would have been a card tester cabinet and the B1700. The B1700 had a maintenance interface to PEPE. There was a separate interface from the CDC7600. The CDC7600 was a very large expensive machine. I doubt that it came with PEPE. The B1700 could have been used to do everything that the 7600 did, although through a much slower interface.
I've attached a file that shows the interconnection of the cabinets.
Good luck, Mark
Thanks for the added info, Mark! I'm loving all of this research. I've got some more research to do at Auburn, for sure. I'm going to get a dissertation from one doctorate that likely wrote about his experience with PEPE while it was at Auburn. I may see about getting that before I leave for Huntsville.
I've updated my album. Feel free to grab the images from the album and rehost them. For posterity's sake, I took a photo of the "RIP" sign on some real MECL. Only the most astute of observers would notice the difference, but you know, it's always the little things. :)
Dunstan Hall is fixing to be torn down, so I'm glad I got access to those rooms prior to the demolition. Turns out I also found some old stacks of punch cards filed away in a storage closet. Looks like mostly data, but there may be some more programs back there too. Probably not PEPE-related, though. There were several other machines in Dunstan that required punch cards, of course.
If you give me your address, I could send a few MECL parts your way if you'd be interested. I can only assume that the date codes around 1973 were from PEPE, and likely no later. I did come across a large quantity of parts from 1976, too, but I would say those could not have come out of the original PEPE.
Also, on an unrelated note...I checked out more of your website, and you sure do have a stunning MGB! About two years ago, I was dead-set on getting an MGB (I couldn't afford an MGA) but my dad really urged me to consider some Italian engineering instead. I bought a 1976 Fiat 124 Spider and have had a blast driving it, but find myself working on it perhaps more than driving it. Well, not quite, but I'll say I did spend a lot of time fixing the brakes, though it's still pulling hard left when braking. I think I may have a stuck caliper or something on the right, despite all four being "professionally" rebuilt by my shop. The Fiat's bumpers are certainly not as attractive as the early all-chrome stuff, but it's more tolerable than the later MGBs in my opinion.
Thanks for the new photos. The MECL chips with a date code of 7325 would be about right. I started at Burroughs in April of 1973 at the start of the PEPE project. We probably had the design ready to buy parts within a year, so parts made in 1973 were a good candidate.
I recall that all of the MECL chips were socketed so removing them would have been easy. Its funny that its been so long but I remember exactly the function of the 10164 in the photo (8 to 1 mux) and I probably still remember all of the other chips functions as well. We delivered PEPE in 1976 to Huntsville. It may be possible that SDC ordered more ICs to build up the maintanence spares for the trip to Kwajalein Island. I would like to have a few of the chips. They may be ICs that I actually touched!
I could make a lot of jokes about English vs. Italian engineering but I'll refrain from doing that ...... I have a cousin Italy who has an original Fiat 500. Talk about small, its about half the size of the newly introduced 500. And NO synchros in any gear! I love my MGB though. I tried to keep it running but the dual carbs were too much for me. I got it professional help a few years ago and now it runs great. Nice old technology - I can actually understand how it works. Your 124 is certainly beautiful.
|Date: Wed, 6 Nov 2013 12:30:28 -0600
Subject: More PEPE Stuff
From: Kyle Owen <kylevowengmail.com>
Sorry it's taken me so long to get back in touch with you. This past summer didn't turn up anything in Huntsville regarding PEPE, though I was told there may be a guy up there who is still alive that used PEPE in Huntsville and would be willing to chat. I may be back up there this winter and might be able to further investigate. I also tried looking around for the SDC building, but didn't find anything that resembled it. Do you recall where it would've been located, by chance?
I did find some more PEPE parts around Auburn, though. I will try to get a package sent your way soon. I think you'll be quite impressed to see what I've found: some sample backplane wiring, as well as some black/white twisted pair as seen connecting the front panel to the rest of the machine. I'll send you some MECL parts and some of the coaxial wirewrap wire for nostalgia's sake. I'll try to have that shipped in the next couple of weeks before exams.
Oh, I also did some digging around at the library and found a thesis written on active filters. Turns out this was all done on PEPE during its short running lifetime at Auburn. Apparently it never was quite up to snuff when it was here.
I also found what I suspect was the hydraulic lift used for removing and installing the power supplies. I'll send you a picture later; maybe you can verify that.
Within the past month or so, I started working with a professor of computer science to restore some old computers of his, including an original PDP-11/20, a PDP-8/E, a few MicroVAXen, etc. I started asking around the EE building to find some photographic history of computing, and sure enough, one professor had some photographs from no earlier than 1981 of the department's computers, including a PDP-11/40 system, the VAX-11/780, some Harris 24-bit mainframe, etc. Unfortunately, I saw no shots of PEPE, as I suppose it may have been trashed by then. I've scanned the pictures in if you'd like to see them: http://imgur.com/a/Kwnn9
I hope you and the MGB are doing well,
6 MECL 10K chips.
10164 8-line Multiplexer
10162 Binary to 1-8 Decoder (High)
10101 Quad 2–input OR/NOR gate
10102 Quad 2-Input NOR Gate
These all have date codes of 1973 which is when we started the design of PEPE.
One of many flat cables used to interconnect PEPE with other computers. In this case, the "IB" was the Interface Buffer that connected PEPE with its maintainance computer, the B1700.
A clock distribution cable. Every flip-flop in PEPE was clocked at the same time within a few nanoseconds.
I don't know exactly where this wire might have been used.
|From: Jim Petrosky <jmphotog1msn.com>
Subject: Great Valley Labs
Date: Mon, 23 Feb 2015 10:05:55 -0500
My name is Jim Petrosky and I too had worked in Great Valley, GVL-3 to be exact. There was a repair depot in the early '70's where we worked on DFSU's. Our "clean rooms" were inflatable clear plastic tents. At sometime in the earlier '70's, we were moved to Downingtown and had a real clean room built. Worked there until I left in 1978 to pursue professional photography which I still do today.
I found your PEPE site, http://www.silogic.com/PEPE/PEPE.html , and thought I'd say hi. Also wondering if you may have known some of the supervisors such as Ed Litwinic, Jim Shaughnessy (or maybe O'Shaughnessy, Atwell, Minarcik or Neal Doll. Spellings may not be correct but close enough.
I want to say you were in GVL-1. Yes??
James Michael Photography
Thanks for your email.
When I was on the PEPE Program, I worked in the building lowest down the small hillside. If I recall, the guard house was in front of the middle building and we were in the building on the left. I don't remember the building numbers.
I don't recognize any of the names you listed. Dick Stotler might be the only person that I worked with who ended up in Downingtown.
I'll add your email to the web page if that is ok with you.
|That would be GVL-3 where I had worked. GVL-1 was at the top of the hill.
Even though we never met back then, I enjoy the throwback of life at a simpler time. Not familiar with Dick Stotler.
Thanks for getting back to me. By all means, feel free to add my email.
PARALLEL ELEMENT PROCESSING ENSEMBLE
As studies of ballistic missile defense systems progressed, the postulated threats expanded greatly in terms of the number of objects arriving simultaneously and the sophistication of the penetration aids. This increase in threat influenced ABM design and especially increased the estimate of throughput needed for ABM data processors.
In response, a new concept of architecture for the ABM data processor was suggested.14
Because a large part of the processing associated with radar tracking and discrimination required that the same set of algorithms be repeatedly applied to each object, parallel elements might carry on the processing. In 1964, research began on a content-addressable memory invented by Lee and Pauli of Bell Laboratories. This memory offered an approach to the needed parallel processing, and a follow-on development program supported by the Advanced Ballistic Missile Defense Agency (ABMDA) led to the Parallel Element Processing Ensemble (PEPE) concept.
PEPE was a programmable, special-purpose computing machine that augmented conventional sequential computing in ABM data processing. Processing capacity was largely independent of traffic because an independent parallel element was assigned to each object in track. Each parallel element was, in fact, a small digital computer, with an arithmetic unit and memory. In addition, each contained a special-purpose input unit called a "correlator," which associated radar replies with the appropriate track by simultaneously comparing each radar reply with predicted track positions. Most of the control circuitry was in an ensemble control unit, which was connected in turn to a more conventional "host" sequential digital computer. The host computer stored instructions for the parallel ensemble, sequenced through them, and passed them to the ensemble control unit. The host computer also did the processing that could be most effeciently handled by a sequential computer.15
By the mid-1960s, a study was under way to adapt PEPE to ABM. The intent was to realistically assess feasibility and cost factors. Several studies were launched, primarily in the areas of software development and testing.
Using readily available components, a processor with 16 elements was built with integrated circuits and tested with an IBM 360/65 as a host computer. This "IC Model" of PEPE was used in the demonstration tests discussed below. A study which showed the feasibility of using more advanced large-scale integrated circuits in PEPE was completed toward the end of the development project.16
Since the job to be done by parallel processing elements would be done the same way by a sequential computer, similar programming methods could be used. A parallel version of FORTRAN, P-FOR, became PEPE's basic programming language. P-FOR was supported by a compiler and an assembler to convert programs into machine code. Also, programs could be written for input to the assembler using PAL, the Parallel Assembly Language.
The language and software system were available well before any hardware so that programs could be tested by simulation. The P44 precompiler converted each operation on parallel data in a P-FOR program into a DO loop on an array in a standard FORTRAN program. The FORTRAN program could be readily tested on any machine with FORTRAN capability.
In addition to P44, which tested P-FOR programs at the source level, the Parallel ABM System Simulation (PASS) simulated operation at the machine level. PASS Tests I to IV, each testing a broader system, were planned. PASS I and II demonstrated PEPE's capability for basic ABM processing and were completed. PASS III and IV were replaced by tests defined by ABMDA, as noted below.
To identify problems and evaluate the advantages of PEPE, several specific applications were studied:17
• SAFEGUARD. Routines planned for SAFEGUARD as developed in NIKE-X simulations were converted to parallel form in PASS I and 11.18
• VIRADE. As discussed previously under Defense of Strategic Forces, the VIRADE concept added the problems of changing sites to the basic ABM problems.19
• ABMDA defined tests. For a final evaluation of PEPE as part of the Bell Laboratories development program, ABMDA defined two systems: Zero Order Software (ZOS) and Preliminary Hardsite Defense (PHSD). These replaced PASS III and IV. The final PHSD demonstration was against a threat defined by General Research Corporation and transmitted to Bell Laboratories by data link from Santa Barbara, California in interrupted real time. The PEPE system used in this test was the 16-element IC model supplemented by sequential simulation, and it achieved essentially all the test objectives.20
• The ability of PEPE to carry a large, constantly growing portion of SAFEGUARD data processing was established. The threat level defined for the current SAFEGUARD System did not make PEPE cost effective. Its cost effectiveness would have to be established for a given threat, for a given ABM system, and with the current state of the processor art considered.
• The feasibility of increasing system capability by removing processing from a sequential computer and assigning it to a parallel processor was established.
• The high level language, P-FOR, was found to be a powerful tool in rapidly programming a complex system.
14. Ballistic Missile Defense, Advanced Development Program, Advanced Data Processing, Vol 1, Bell Laboratories, September 30, 1969.
The first of three annual reports; introduces the Parallel Element Processing Ensemble (PEPE) concept and details its architecture and software and initial simulations.
15. IEEE COMP CON '72 Digest:
PEPE Computer Architecture, B. A. Crane, M. J. Gilmartin, J. H. Huttenhoff, P. T. Rux and R. R. Shively
These papers present a brief overview of the characteristics, PHSD test implementation, and performance of PEPE in a general distribution publication. It was presented as an example of "Innovative Architecture, " the theme of the Conference.
16. Ballistic Missile Defense, Advanced Development Program, Advanced Data Processing. (U), Vol 2, Bell Laboratories, September 30, 1971. (SECRET)
Presents an overview of the PEPE IC Model, details and results of PEPE application studies to Ballistic Missile Defense, ZOS, PHSD and Off-loading, and results of hardware implementation studies.
17. Ballistic Missile Defense, Advance Development Program, Advanced Data Processing (U), Vol 2, Bell Laboratories, September 30, 1969. (SECRET)
Presents the results of studies of the Application of PEPE to SAFEGUARD, VIRADE, and to Coherent Waveform Processing.
18. Ballistic Missile Defense, Advanced Development Program, Advanced Data Processing, Vol 1, Parts 1 and 2, Bell Laboratories, September 30, 1970.
This second annual report presents a detailed description of the Integrated Circuit (IC) PEPE model brassboard hardware and its support software and the PASS II evaluation studies.
19. Ballistic Missile Defense, Advanced Development Program, Advanced Data Processing (U), Vol 2, Bell Laboratories, September 30, 1970. (SECRET)
Presents results of GPSS simulation studies of Ballistic Missile Data Processing and design alternatives and a study of the application of PEPE to SPRINT missile guidance.
20. Ballistic Missile Defense, Advanced Development Program, Vol 1, Advanced Data Processing, Bell Laboratories, September 30, 1971.
The third annual and final report of PEPE studies at Bell Laboratories. Presents an overview of the PEPE system, hardware and software, and the principal final year studies and demonstrations, ZOS, PHSD, and offloading. Includes an introduction to proposed LSI implementation.
|From: "Jamshed Mulla" <jmullacomcast.net>
Date sent: Sun, 10 Sep 2017 18:06:08 -0400
I stumbled on your page about PEPE and was really impressed with all the details you have recorded.
I did not work on PEPE itself, but my contact with it came in later years (1976-1978). I worked on my Ph.D. in the late '70s at the University of Michigan where my advisor had a grant from RADC (Rome Air Development Center). I'm not exactly sure how we got connected with SDC and the PEPE project, but in 1976 I started working on my thesis on parallel computing and wanted to look into expanding on the scope of SIMD machines like PEPE.
I actually did visit the PEPE installation at the BMDATC in Huntsville with my advisor. In fact, I remember the exact date (Aug 15, 1977) because when we were flying home and connecting flights in Memphis the next day, we heard about Elvis's death! :)
My thesis proposed a MIMD version of PEPE where the PE's would have more processing power and actual capability to store parts of the application locally, so that each PE could execute multiple instructions in parallel, thus reducing overall processing time. A lot of the thesis dealt with calculating the expected speed improvements of MIMD over SIMD. As I recall, I also proposed a programming language for such a MIMD machine. A fellow student who was working on his Master's degree built a prototype of the machine using (I believe) 8 Motorola 6800 processor boards.
If you are interested, a brief paper on my thesis is on-line here <https://deepblue.lib.umich.edu/bitstream/handle/2027.42/5687/bac5077.0001.001.pdf?sequence=5> . I used the term "associative processing" as I presented the concept as a natural extension of associative memories which were somewhat the rage in the day. :)
Thanks for your email about PEPE.
I looked at your paper. I also recall associative memory being a hot topic back then. I don't think it was ever appreciated as being as a really different way to approach computer programming.
If its OK with you, may I add your email to the PEPE web page?
(My Elvis is dead story: I was still working for Burroughs. We were back from Huntsville and working on the next computer. We were meeting the customer (Lincoln Labs at MIT). Our
team just met for breakfest before heading to the Labs and one of other members told us of Elvis's death.)
Absolutely! No problem with adding my e-mail to the web page. I am also going to forward your link to my college colleague who worked on the physical implementation of the associate processor. His name is Dave McCubbrey. He has worked on parallel processing systems, mainly for image recognition, for most of his career. He is now president of a company called Pixel Machines.
email : email@example.com