On Thu, 2 Jun 2005 19:45:44 +0000, Brian J. Beesley wrote:

>_Assuming there is adequate cooling_ you should change the memory first as it 
>is most probable to have a fault and is anyway the easiest to change.

There's plenty of cooling - the board etc. is not yet built into a case, but is 
running
on a workbench in an air-conditioned (=refrigerated) data-center.  
Room-temperature is 
kept at about 22C (current outside is 32C). 
The heat-transport away from the board is maintained by a separate fan. 

>The next thing to try is the power supply - replace with a quality component, 
>cheap PSUs can and sometimes do make systems unstable. I've never had 
>problems with Enermax PSUs, which aren't cheap but have the bonus of being 
>reasonably quiet.

Yeah, I've been there too - the setup is being supplied by a PSU from a 
Swisscase
box (the one intended for it in the end).  I've never had problems with them 
before.

>In my experience processors _never_ have intermittent problems unless they're 
>made to run too hot or are overclocked, and faults on main boards are rare 
>too. 

That's the kind of statement I was after - I was fairly certain about that 
myself,
but it's good to have it confirmed.  And as you can see from my later posts, it
was the memory in the end. 

What astounds me is - this time I bought two no-name memory-sticks from a known 
Swiss supplier
and they *both* end up being RMAed.  I admit I always buy no-name - the extra 
money in 
buying Kingston et al. is IMHO just not worth it.  
And I still don't think so - I operate a fairly large cluster and have 
occasionally bought large 
amounts of memory bulk. Sure, there's always a few duds, but a 100% of a 
shipment is highly unusual.

Thanks for chipping in anyway. 


/Per Jessen, Zurich


_______________________________________________
Prime mailing list
[email protected]
http://hogranch.com/mailman/listinfo/prime

Reply via email to