Greetings,

The single K7 boards may well have problems (I haven't tested those as extensively), 
however, the Tiger, and the newer Thunder revision (the one with a heatsink glued to 
the northbridge) seem capable of being quite solid once the bum boards are screened 
out. They are very touchy about their power. I'm rejecting >20% of the power supplies 
I get due to instability which develops after several hours at 100% CPU. Anything less 
stressful than that and they seem fine. It makes testing hard. Just to keep things 
nice and confusing, the P.S. problem looks a lot like a flaky mainboard. Once I get a 
good supply in the system, and keep it in a cool room with good airflow, they work 
great.

The airflow is as important as the temperature. A rack of these things will form a 
'bubble' of hot air around them even in a perfectly cool room unless they get direct 
airflow. A proper datacenter environment is more than adequate.

G'day,
sjames

Quoting Ronald G Minnich <[EMAIL PROTECTED]>:

> lotsa people having trouble with K7s nowadays.
> 
> I don't buy the altitude argument. I buy the "K7 mainboards are poorly
> engineered" argument, personally, given my trouble with K7s on four
> different motherboards at this point. I also have a friend with a
> 32-node
> cluster with conventional BIOS and K7SEM mainboards -- it has frequent
> strange memory system problems that look like "flaky mainboard".
> 
> PIII looks better all the time.
> 
> ron
> 
> ---------- Forwarded message ----------
> Date: Sun, 24 Mar 2002 09:50:25 +0800
> Subject: Re: bummer!
> 
> I met another guy today with K7's that crash when doing heavy numerical
> computations (like mine).  His a Tyan boards. Also Duals.  His
> motherboard
> is the newer "tiger" board and mine is the previous (which was new when
> I
> bought it) "thunder" motherboard.  his are 1.6Ghz mine are 1.2Ghz.
> 
> I think its the altitude.  lower air flow plus lower thermal transport.
> I keep mine in a refrigerated room so its not the input air
> temperature.
> The output temperature is quite warm.
> 
> -- 
> 
> 
> 



----------------------------steven james, director of research, linux labs
LinuxBIOS Cluster Solutions                   230 peachtree st nw ste 2705
High-Speed Colocation, Hosting,                        atlanta.ga.us 30303
Linux Hardware, Development & Support             http://www.linuxlabs.com
* Visit us at SuperComputing 2002, Booth 1441 *   office/fax 404.577.7747/3
--------------------------------------------------------------------------

Reply via email to