AMBER Archive (2006)

Subject: RE: AMBER: Opteron vs. Xeon performance

From: Yong Duan (duan_at_ucdavis.edu)
Date: Sun May 14 2006 - 19:53:55 CDT

I think the most significant difference between Opteron and Xeon is the
power consumption (heat).

Dual-core/dual-CPU operons do not have much heat issue at 2.0GHz. I have a
bunch of these, have been quite pleased by the low power load (about 200W
per node).

Cooling a rackful of 3.2GHz dual-core Xeon is definitely a challenge with
each node taking 350-400W. If you have 40 of them on one rack, you'd need
4.5 tons of cooling power for each rack (assuming perfect cooling
efficiency). As for the electric bill, for each watt the computer uses, it
takes ~1.2-1.3W to cool. The total bill is about 1000W for each node if each
node needs ~400W to power up. This is not trivial if you happen to run a
bunch of them 24/7. You are talking about N x 24 x 365 Kwh per year, where N
is number of computers. This actually translates to about $1000 per node per
year, if you are in CA. In addition to the substantial cooling equipment
cost, your electric bill alone would be $100K per year for a 100-node
cluster. So, you probably do not want to install it in your garage even if
your company promises to give you the air-conditioning equipment for free
:). For opteron cluster, this cost is reduced by almost 50%. From this
perspective, Opteron appears to have a lot more room to go than Xeons.

yong

> The xeon is cranking out around 3.6 GHz these days; I am
> mostly seeing 2.2
> GHz opterons. I don't know where intel takes the xeon next,
> but it would
> seem that highspeed clock heating is an issue. On the other
> hand, the
> opteron already has heat issues down around 2 GHz, so I don't
> know how much
> further it is going.

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu