AMBER Archive (2005)

Subject: Re: AMBER: Interpreting Amber8 Benchmark Results

From: Robert Duke (rduke_at_email.unc.edu)
Date: Wed Aug 31 2005 - 16:37:53 CDT


No, not at all. Just in this particular domain. All the context switching
really screws up performance, and there are cache issues, etc. that make it
plain crazy to run an HPC cluster in multiprogramming mode. But
multiprogramming is still a great concept for a generic compute environment.
Regards - Bob

----- Original Message -----
From: "Bill Ross" <ross_at_cgl.ucsf.edu>
To: <amber_at_scripps.edu>
Sent: Wednesday, August 31, 2005 5:29 PM
Subject: RE: AMBER: Interpreting Amber8 Benchmark Results

>> Most large computer clusters bill on
>> walltime anyway so this really is the important value.
>
> I guess the days have passed of the single-threaded machine
> running multiple jobs at once.
>
> Bill
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>

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