• Please review our updated Terms and Rules here

Burroughs B6700 Memory...

rvaneynd

New Member
Joined
Dec 14, 2011
Messages
4
Location
A, A
Hi,

Have been given a memory module of a Burroughs B6700 (see photos attached).
Does anybody know how much memory this is?

I believe it is a 50Kb module, but I am not sure.

It says DataProducts Assy 716303-1 it is 3 Wire Core Memory.

IMAG0034.jpgIMAG0035.jpg

Thanks,
Raoul
 
Hi Raoul,

A few details which might help to decipher the board:

B6700 had 52-bit words (48-bits + 3 tag bits + 1 parity bit) and a maximum addressing capacity of 1MiW (1 048 576 locations, 6Mbytes) via 20-bits of address which was split into 14-bits to select a word (16KW range) and 6 bits to select a memory module (64 modules).

Do you have any more information about the machine this memory module came from? was the B6700 recently scrapped and if so where was it located? are there any pictures of the original machine?

cheers.
 
Thank you for the response.

I do not have much more information except for.
The B6700 was scrapped over 30 years ago, I think.

The modules came out of the B6700 that was installed at Technical University in Eindhoven, Holland.
My dad used to study and work there and when they scrapped it, he took the boards out and framed them as they make for some great Technology Art.

Now he has passed them on to me and I am just curious about the memory capacity.

Raoul.
 
Burroughs B6700 is from 1971/1972, although it was based on the B6500 which was 1968/1969. It had segment-based virtual memory rather than page-based virtual memory. Segments could be code or data (down to array level granularity).
 
My guess, assuming it is a single core plane, based on what look to be address decoders on the edges, this may be 1KW of memory, that is 1024 x 48 bits, so a total of 48Kbits or 6Kbytes, which would match your "...a 50Kb module" statement.

If you can place the board on an image scanner and get a close up picture we might be able to identify the chips involved and other features which might be more conclusive.
 
Last edited:
Planar Core

Planar Core

My guess, assuming it is a single core plane, based on what look to be address decoders on the edges, this may be 1KW of memory, that is 1024 x 48 bits, so a total of 48Kbits or 6Kbytes, which would match your "...a 50Kb module" statement.

If you can place the board on an image scanner and get a close up picture we might be able to identify the chips involved and other features which might be more conclusive.

It's been a long time... like 30 years since I saw one, but that looks like half a a "planar core" memory module from a B6700. It's the sense amp part, the actual core was on a second board bolted to those standoffs and connected by all of those pins on the edges. This was the last core memory Burroughs used before going to chip memory. Each one of these assemblies was 20 bits of a system word, and 3 boards made a complete 60 bit word. The extra bits were used to correct single bit errors on the fly, and detect multiple bit errors. I'm thinking they were 32K x 20 bits each. Anyway, the 6700 could have 6MB of main memory. Using this stuff, that took 4 cabinets about 4 ft wide, maybe 6 tall, and 2 deep. ;-)

They weren't supposed to be field repairable, but I did fix one that had a bad sense amp.
 
oops... I missed the second picture. I'm pretty sure that's a planar core module.
 
It's been a long time, but what little I can recall from my RFQ work (know the competition), wasn't the B6700 used as a frontend to the BSP?

I remember hearing about the Burroughs Scientific Processor back then, but I think it used a 7700 or 7800 as manager. I wonder if any of them were actually deployed? I know I never saw one... but that only covers maybe a half dozen 6700s or 7700s.
 
I don't know how many BSPs were actually deployed, but supercomputers were never a quantity seller. We had a couple of bids out against BSP proposals, but neither we (CDC) nor Burroughs won. The TI ASC was another competitor back then--I don't know if they ever sold any. Cray picked up the lion's share of sales right about then (early 70s).

It was an awkward time for the market--the world was moving to semiconductor memory and at the same time, moving toward ICs from discretes--and ICs were getting faster very quickly. Manufacturers were trying to jump on the bandwagon as fast as they could (it took a lot of time to design a new machine) and a lot of things turned out badly--I recall the Honeywell completely redesigned one of their older models to use ECL IC technology and found that the result didn't run any faster, but did eat more power.
 
I don't know how many BSPs were actually deployed, but supercomputers were never a quantity seller. We had a couple of bids out against BSP proposals, but neither we (CDC) nor Burroughs won. The TI ASC was another competitor back then--I don't know if they ever sold any. Cray picked up the lion's share of sales right about then (early 70s).

The good old days. ;-) Back when I was a student at Michigan State, the computer lab had a CDC 6500 and a 3600, at least I think those were the model numbers. Not that they would let any of us undergrads near it, just the observation room to see the blinkenlights.
 
I don't recall much about the 3000-series, only that they were 24- and 48-bit machines related somewhat to the 1604; most of my work was in 6000/Cyber 70 before I went onto the STAR-100 project. The 6500 was pretty popular with universities; I think Purdue also had one as well as Colorado. Basically two 6400s sharing memory and PPs. There was also a 6700--6600 and 6400 CPUs in the same box. I'd heard that a 6500 cost substantially less than a 6600, but never verified that.

All you needed in addition to the box was an MG set (208V-3-phase-400Hz), the usual HVAC for peripherals and a chilled water supply for the CPUs. Aren't you glad that you're not collecting those? :)
 
... Each one of these assemblies was 20 bits of a system word, and 3 boards made a complete 60 bit word. The extra bits were used to correct single bit errors on the fly, and detect multiple bit errors. I'm thinking they were 32K x 20 bits each.
I've never heard of it described this way, do you have a reference for the memory layout you describe which suggests that the B6700 had 60-bit memory words?

I am using this reference where it states:

Code:
B6700 Reference Manual ([url]http://bitsavers.org/pdf/burroughs/B6500_6700/1058633_B6700_RefMan_May72.pdf[/url]), 
page 1-8 in the section on "Memory Words":

Each memory word contains 48 information bits, three control bits, and a parity bit.
for a total of 52-bits per addressable word.

I also read in the Burroughs B6700 Handbook Volume I Hardware, in section 8 (Interface Cables) which details the memory subsystem I see just 52-bus lines described connecting the CPU to the memory subsystem.
 
I've never heard of it described this way, do you have a reference for the memory layout you describe which suggests that the B6700 had 60-bit memory words?

I am using this reference where it states:

Code:
B6700 Reference Manual ([url]http://bitsavers.org/pdf/burroughs/B6500_6700/1058633_B6700_RefMan_May72.pdf[/url]), 
page 1-8 in the section on "Memory Words":

Each memory word contains 48 information bits, three control bits, and a parity bit.
for a total of 52-bits per addressable word.

I also read in the Burroughs B6700 Handbook Volume I Hardware, in section 8 (Interface Cables) which details the memory subsystem I see just 52-bus lines described connecting the CPU to the memory subsystem.

I don't have any of my old reference books, I'm just going on my own memory from back then. I was a field service engineer on the 6700. The system did have a 52 bit word - 48 data bits, 3 tag and one parity. For this particular memory system, there were actually 60 bits per word. The other 8 bits never left the memory controller. The memory controller generated an 8-bit code in hardware that was determined by the other 52 bits. On reading, it was compared and could determine if there was a single bit error or multiple bit error. If it was single, the error bit number could be determined and flipped. Kinda neat, unless that part of the controller failed.
 
All you needed in addition to the box was an MG set (208V-3-phase-400Hz), the usual HVAC for peripherals and a chilled water supply for the CPUs. Aren't you glad that you're not collecting those? :)

I seem to recall the B6700 used enough power for several suburban homes. The AC power mod ran on three phase, although I don't remember the voltage. Each of the mainframe boxes had a 400A or 600A logic supply with the odd voltages of 4.75 and -2v. Handy way to heat one's house. ;-)
 
I seem to recall the B6700 used enough power for several suburban homes. The AC power mod ran on three phase, although I don't remember the voltage. Each of the mainframe boxes had a 400A or 600A logic supply with the odd voltages of 4.75 and -2v. Handy way to heat one's house. ;-)

That's pretty close to the truth. In January, 1974, I had to make a trip from Sunnyvale to St. Paul/Arden Hills. This was the height of the Arab oil embargo. Everything was cold--including my room at the Ramada. At CDC, there was ice on the inside of the windows of the offices. People were wearing three sweaters at work.

I holed up with my work in the STAR machine room at ADL between a couple SBUs. Nice and toasty. The next evening, I brought a pillow and slept on the floor. It must have been the only place in the Twin Cities that was a comfy 74 degrees. :)
 
I seem to recall the B6700 used enough power for several suburban homes. ...Handy way to heat one's house. ;-)
I can believe it; when I worked with Burroughs they had one for a while in the DP room where I would occasionally use its smaller sibling, a 2700, and it was a biggie all right.

The CPU & memory cabinets alone took one entire wall of a fairly large room, and it had the most impressive panel of blinkenlights I've ever seen; it was in house for a couple of months being tested and set up for the Metro Toronto police department, and when idle it displayed the MTP crest on its panel of lights which ISTR was about 3 feet square; those bulbs alone probably would have heated one or two small rooms...

UTAS B6700 front panel.jpg

As a matter of fact that's probably yours in the picture, Nigel.
 
Last edited:
Each of the mainframe boxes had a 400A or 600A logic supply with the odd voltages of 4.75 and -2v. Handy way to heat one's house. ;-)
The B6700 hardware manual lists several other voltages too: for the core-memory system there are +1V, -50V, +63V and +37V. In another part of the CPU/DCP system there are also +12V and -12V - the power supplies must have been quite complex to supply so many different voltages.

Burroughs appear to have their own logic integrated circuits listed as CTuL (Complimentary Transistor Micro Logic), they were positive logic, "true" was +2.5V and "false" was -0.5V


Jim: do you happen to remember what was special about the B6748 "restricted model"? it is not listed in the reference manual I have but I found another reference to it in Doran's book.
 
Back
Top