Subject: Re: Trouble booting on Multia
To: Ross Harvey <ross@teraflop.com>
From: None <jack@jupiter.com>
List: port-alpha
Date: 04/21/1998 15:19:12
> > Here at Jupiter we use Multias with proprietary software as X terminals.
> > We have had a large number fail with a number of different symptoms.  We 
> > finally tracked all of the problems down to one of 6 Resistor Packs on the 
> > bottom of the PCB.  The RPs develop shorts between adjacent pins.  By 
> > replacing these RPs we cleaned up our entire bone pile.
> > 
> > Jack
> 
> Wow. If this is indeed the usual problem you have really helped a lot of
> people. This report does fit with an earlier successful multia repair that
> was done by cleaning the boards.
> 
> I suppose I should amend my earlier axppci33 advice by pointing out that
> the multia has the only frame buffer netbsd alpha has ever supported.
> Although there do seem to be temporary problems with the code, we can
> fix that up for a multia revival... :-)
> 
> Ross

Based on what I've heard already, I guess I'd better get specific.
The RPs in question are all 4.7K pullups.  The numbers on the silkscreen are:

E221, E222, E226, E227, E228, and E233

The parts that we have seen that are bad, and it is bad parts, are all marked

   ---------------
  | 4820P-002     |
  | -472          |
  | xB 9435 B     |
   ---------------

Remember that there are several different models of Mulitas.  Ours are 166MHz 
ones, with a PCB number of 5022554-01 D1.

Based on what we've seen, DEC got a bad batch of the 4.7K pullup RPs that 
develop shorts between adjacent pins over time.  We now routinely replace all 
these RPs.  BTW, we also replace E215 with a 74ABT623.  This part buffers the 
ECC bits from DRAM and the ABT part seems to work much better at temperature.
Once this is done, the Multias seem to be as reliable as anything else.

Jack