Consistency bugs on IBM 370/158 AP/MP

[James Dow Allen]

Thirty-five years ago, interesting bug syndromes surfaced on some IBM 158
APs and MPs. At first IBM might have thought they had software bugs -- the
problems were coincident with its 1977 introduction of MVS/SE and VM370/SE.
But they probably observed early on that all the failing machines had
NatSemi dynamic add-on memory installed.

I wrote up an explanation of the underlying hardware bug, going into
details and tangents for 4000 words. I mentioned it a decade ago in this ng
and got one or two appreciative responses. Perhaps there are 2 or 3 other
nerds who want to read about this memory bug in excruciating detail.
Here's the URL:
http://james.fabpedigree.com/bug22.htm

James Dow Allen

[Jon Elson]

Post by James Dow Allen
Thirty-five years ago, interesting bug syndromes surfaced on some IBM 158
APs and MPs. At first IBM might have thought they had software bugs --
the problems were coincident with its 1977 introduction of MVS/SE and
VM370/SE. But they probably observed early on that all the failing
machines had NatSemi dynamic add-on memory installed.
I wrote up an explanation of the underlying hardware bug, going into
details and tangents for 4000 words. I mentioned it a decade ago in this ng
and got one or two appreciative responses. Perhaps there are 2 or 3 other
nerds who want to read about this memory bug in excruciating detail.
http://james.fabpedigree.com/bug22.htm

Interesting! A friend of mine bought a pair of IBM System 370/145's and
moved them to his house. These had AMS/Nat Semi 1 MB memories, but they
were static, I think 16 Kbit chips. I bought one of the memory boxes,
and was going to build a bit-slice computer and repurpose the Nat Semi
memory as the main memory for that. Well, I got the bit-slice machine
working, but the effort in creating all the necessary microcode to
emulate any useful architecture put a stop to it. I eventually sold the
memory boards to a chip broker. I did end up putting a MicroVAX-II
in the cabinet.

Being static RAM, they did work fine in the 370. It was pretty weird
how NS hacked the memory into the CPU, with external power supplies
feeding additional voltages into the 370's backplane boards, and a bunch
of cards with 10K ECL on them to translate from MST levels to ECL.

Jon

[Peter Flass]

Post by Jon Elson
Interesting! A friend of mine bought a pair of IBM System 370/145's and
moved them to his house.

Wow! What kind of rewiring did he have to do?

[Jon Elson]

Post by Peter Flass

Post by Jon Elson
Interesting! A friend of mine bought a pair of IBM System 370/145's and
moved them to his house.

Wow! What kind of rewiring did he have to do?

Hopeless! The 145 had a 17 KVA GE motor generator set in the back,
and a whole bunch of gear to soft-start the motor. We built a massive
L-C phase converter and actually got the M-G set to spin, but he only
had a 60 A 240 V service, and when we closed the delta relay, it
always tripped the breaker. The nominal running current for a 20 Hp motor
would be 66 A per phase. Phase converters draw a HUGE amount of
reactive current, so it would likely take a 120 A service to even THINK
of getting that MG set online. I think it would be totally insane to
try to run a 145 in a home! Assuming that it doesn't draw the full
load the MG is capable of, it probably still was about 15 KW of DC
load. And, of course, he ONLY had the CPU and memory cabinet.
Not a hell of a lot you can do from the console typewriter!

So, it all went to scrap! He could have sold the boards out of it
(when he first got it, at least) to a broker for real $$.
I have a box of tri-lead from one of the machines.

Jon