16 Aug
2024
16 Aug
'24
7:33 a.m.
On Aug 15, 2024, at 9:44 PM, ben via cctalk
<cctalk(a)classiccmp.org> wrote:
On 2024-08-15 7:39 p.m., cz via cctalk wrote:
Having the diagnostics available is critical for this. But yes, if you do then you're
right.
It takes me back to 1974, when our college 11/45 crashed and then failed hard early in the
reboot. Called DEC FS. While waiting for the tech to arrive, we (my classmate Josh Rosen
and I) ran CPU diagnostics. Turned up an MMU failure, failed bit in some CSR. Josh
grabbed the engineering drawings to find the data path in question, and pointed at one
chip, saying: "that one". I asked "why not one of the others?"
Answer: "because that's the most expensive chip in the path".
We told Jim Newport when he arrived what we found, including Josh's blaming a specific
chip. Jim said "we'll see". Soon after he grabbed his soldering iron and
spare chip box to replace -- yes indeed -- the specific chip Josh had pinpointed.
paul
Eh, it will go for what it goes. Try and keep in
mind how tough it was to keep a 1010 running in 1995. Then add 30 years to that.
The thought of tracking down a bad flip flop on a thousand flip chip boards really makes
me think "yow".
True, but back then things were designed to fixed and tested.