Stretch

From: Gary Oliver <go_at_ao.com>
Date: Fri Oct 27 12:52:03 2000

While a student at Oregon State University (way back around early to mid 70s)
we managed to obtain one of the beasts (well most of it anyway.) At the
time we were running a CDC 3300 (memory starved) and the hardware group didn't
want to pay CDC the 1e6 bucks necessary to fully populate the memory of the
3300.

The HW manager (an efficient scrounger if there ever was) found this 7030
(unfortunately I don't recall where it came from - rumor was LANL but I'm
not sure.) He concluded that we could hang the 6 memory boxes (36 x 32k
or 72 x 16k depending on "interface") onto the 3300.

Well that didn't pan out - but a group I was involved in *did* benefit from
this choice. We were running a small, home brew computer (Nebula) which
at the time used glass delay lines as memory (slow) and only had 4k x 35
bits of memory. We grabbed one of the memories and built a controller
and connected it to the Nebula. Worked great and boosted the speed of the
machine all the way to a whopping 33 usec per word time (it *was* 100 usec
word time!)

The memory interface was actually rather nice - totally asynchronous if I
recall, so the interface was mostly an issue of converting the parallel
stretch memory (with negative logic levels) to the serial bus (did I mention
that Nebula was a serial computer?) with TTL/DTL levels. Basically a giant
shift register. DMA was a little bit hairier, but it all fit in a single
19 inch rack card frame with about 20 44-pin "vector" cards. Mostly TTL
with some special negative-bus level converters on a few cards. I still
have a copy of the paper the designer wrote as a class project.

We were able to accomplish this partly because we had another computer with
a HUGE motor generator and the stretch memory needs a bunch of supplies
which we borrowed from this other machine. A power sequencer was built that
brought up the major bus voltages of the stretch memory *slowly* (motor
driven variac with a relay logic controller.) We built the relay logic from
(appropriately enough) IBM relays we glommed from an old accounting
machine card reader (sorry for the reckless disregard for antique equipment,
but it was the 70s...)

Fortunately, IBM was quite helpful with hardware docs for the machine. What
didn't come with the "donation" of the hardware was obtained from IBM
directly. We had it all (about 200 pounds of prints if I recall.) I just
wish it had been salvaged when the lab was torn down. I heard about the
discard to late. OSU had a (bad) habit of discarding ALL paper when a
piece of equipment was sold at surplus. I've had countless discussions
with the surplus management at the time about this and today they are
a little bit better in this regard.

The stretch was a fantastic piece of hardware for it's time. We learned
a lot about computers architecture tearing apart the memory design.

Pity I don't have pictures. It was a glorious mess of home-brew stuff.

-Gary

At 09:10 10/27/00 -0400, you wrote:
>While having a conversation with an old IBM buddy of mine we fondly
>remembered our days with the IBM 7030 (Stretch) computer. It was used to
>prototype the 360 before it came out. It was a real monster with only 8
>ever built (and sold). Does anybody else remember this monster?
Received on Fri Oct 27 2000 - 12:52:03 BST