Subject: Re: Multia repair
To: None <>
From: Ross Harvey <>
List: port-alpha
Date: 06/08/1998 13:44:31
We repaired two multia's here at Avalon and I can confirm that the Jupiter
fix works just fine. We used the (CMOS) 74FCT623 because it is easier to
get than the ABT part. (ABT is really top-of-the-line but the FCT will run
even cooler.) Even without the Jupiter info, it's fairly obvious from a
quick visual that the F623 is burning up.

We replaced the resistor networks as well. The 20-lead is really hard to buy
so we used digi-key 16-lead networks and soldered 4 1206 ceramic chip resistors
to the remaining lands. (You have to orient two of them on their narrow sides.)
We bussed the other terminals together with 30 AWG (or was it 36?) magnet
wire and connected it to the common pin 20. (Land 20, resistor network 16.)

Note, however, that the odds against both the networks and the F623 failing
at the same time are astronomical. I would suggest trying just the F623
replacement, since we _know_ it needs to be done.

BTW, it is _not_ OK to run the alpha without airflow, however, up to a point
the negative consequences are primarily a reduction in the CPU's lifetime.
This might never be noticed. If you have a thermometer or thermocouple
that works around 65C you can monitor it will need some
thermal grease. Measure the temp between the studs at the base of the heat
sink and don't let it go much over 65C. (The "pain" test is a good one,
BTW...if you can hold your fingers on it, it's fine, if you can't, it
might be, might not be...)