[6987] in Release_7.7_team
Re: 32 bit dialups vs. 64 bit clusters...
daemon@ATHENA.MIT.EDU (Alex T Prengel)
Thu Sep 30 19:58:18 2010
From: Alex T Prengel <alexp@MIT.EDU>
To: Greg Hudson <ghudson@mit.edu>
Cc: Jonathon Weiss <jweiss@mit.edu>,
"release-team@MIT.EDU" <release-team@mit.edu>
In-Reply-To: <1285882880.20521.1083.camel@ray>
Content-Type: text/plain; charset="ISO8859-1"
Date: Thu, 30 Sep 2010 19:58:07 -0400
Message-ID: <1285891087.3142.9.camel@dit>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
On Thu, 2010-09-30 at 17:41 -0400, Greg Hudson wrote:
> It may be too late to change course on this, but:
>
> * Back in the day, when people suggested we move to 64-bit, I was never
> able to get a specific reason why it was desirable. Is there a specific
> reason now?
>
> * If you go to ubuntu.com and click "Download Ubuntu", the 64-bit
> desktop install CD is "Not recommended for daily desktop usage." I
> don't really know why, but when I noticed that yesterday it made me
> wonder about the decision to go 64-bit in Athena clusters.
>
> I don't have a personal stake in this; I just wanted to express some
> skepticism that it's better to be leading-edge than trailing-edge on
> this front.
Thanks for the input. I think at this point we are committed to 64; I'm
personally aware of a moderate number of vocal requests that we go 64
bit, particularly one from an MTL administrator (mbm@mtl) who felt
strongly that it would make a large difference in performance for an
application he maintains for the department (Cadence). I don't recall
anyone on release-team expressing any opinion that we should stay with
32. I think heavy Matlab users prefer 64 over 32 for performance
reasons, though I don't know of quantitative metrics to back this up.
Alex