heat, Transmeta chips and Sys Admin costs (was RE: disadvantages of a linux cluster)

[graham.mullier at syngenta.com]

I hadn't noticed the paper (so thanks for the repost) but don't agree with
everything in the intro - specifically the point about RISC chip complexity
and power draw/heat output:

The transmeta chips are clearly low-consumption, low heat output chips (a
Good Thing), while P4, IA64 are hot, very hot (not so good). However, the
MIPS chips are nowhere near as hot as the Intel chips - see the recent
publicity about the 16cpu/4U compute brick for the SGI Origin 3900 ["The
R14000A only consumes 17 watts of power." -
http://slashdot.org/comments.pl?sid=44763&cid=4643891]. What's the position
with other RISC chips (HP, IBM)?

The other point I'm not happy about is the extrapolation from 9 months of
trouble-free operation to a zero downtime cost vs other options. Makes it
looks too much like they've decided what the answer should be and they're
working back from that to define the itemised costs. Perhaps I'm being
unfair but 9 months seems like a very small sample when compared to the data
that will have gone in to the other options.

However, it is very good to see this, as I've been interested in the TM
chips for a while, so seeing a paper about this sort of very high density
config is useful.

Graham


Graham Mullier
Chemoinformatics Team Leader,
Chemistry Design Group,
Syngenta, Bracknell, RG42 6EY, UK.
direct line: +44 (0) 1344 414163
mailto:Graham.Mullier at syngenta.com



-----Original Message-----
From: M A Suzen [mailto:mehmet.suzen at bristol.ac.uk]
Sent: 12 November 2002 20:16
To: Sam Daniel; beowulf at beowulf.org
Subject: Re: disadvantages of a linux cluster


Hi,
Any body notice System Administration cost in table 5.


--On 12 November 2002 11:47am -0400 Sam Daniel <samhdaniel at earthlink.net> 
wrote:

> http://public.lanl.gov/feng/icpp02.pdf


_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf