Date: 25 Oct 2000 09:44:43 -0400 From: Allan Strand <stranda@cofc.edu> To: Tim McMillen <timcm@umich.edu> Cc: freebsd-questions@freebsd.org Subject: Re: Choosing a fairly high-speed compute server Message-ID: <868zrdt4qc.fsf@linum.cofc.edu> In-Reply-To: Tim McMillen's message of "Mon, 23 Oct 2000 16:33:34 -0400 (EDT)" References: <Pine.SOL.4.10.10010231615480.20641-100000@stargate.gpcc.itd.umich.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
Thanks for the pointers. I think it has some parallizability, but frankly I am daunted by the networking hassles of creating a cluster. I already spend plenty of time maintaining my toy network of a few boxes. Right now, I am leaning towards a couple of cheap hi-speed boxes performing completely separate runs of the simulation independently. a. Tim McMillen <timcm@umich.edu> writes: > Depending on how parralizable your process is, you might want to look into > creating a cluster. You may be able to do it more cheaply. > See http://acme.ecn.purdue.edu They did get lucky with part > availability. > Also for price / performance the AMD Duron beats the P3 easily. > See www.cpuscorecard.com/cpu_latest.htm > www.tomshardware.com/cpu/00q4/001017/athlon-09.html > if that link doesn't work look for the cpu guide. > or just search Tom's hardware for Duron. The FP performance kills the P3, > while most other things are a little behind. > > You could create 2 or more boxes using Duron Chips and use them as a > cluster. The one drawback is that as of august there were no chipsets > for the Duron that could handle ECC memory. Don't know if that was a > requirement for you. The ECC supporting chipsets were on the way > supposedly. Hope that helps > Tim > > > On 23 Oct 2000, Allan Strand wrote: > > > Hi all, > > > > I have written a pretty complex simulation in c++. It seems to work > > pretty well at the moment, but it is too slow for my liking. I am in > > the process of optimizing the code, but there is a point where that > > will not help speed things up. The program does lots of memory > > allocation/deallocation and that seems to be it's speed downfall. > > > > So, I need a new box and I wanted to solicit input on some choices. > > First of all, I really plan to have this box dedicated to > > computational problems only, so I don't really need soundboards, fancy > > monitors, modems, etc. All that is required is a NIC, >= 256M ram, > > and high cpu+cache+bus performance. Of course I want to minimize > > price. The program could probably be threaded at a gross level, so > > SMP might ultimately be an option (even if not threaded?), but I was > > thinking of starting with a 800Mhz PIII machine. Does this rambling > > make sense? Can FBSD utilize the PIII features? Is there a > > hi-performance computation in FBSD FAQ? > > > > TIA > > > > a. > > -- > > Allan Strand > > > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > > with "unsubscribe freebsd-questions" in the body of the message > > > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-questions" in the body of the message -- Allan Strand, Biology http://linum.cofc.edu College of Charleston Ph. (843) 953-8085 Charleston, SC 29424 Fax (843) 953-5453 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?868zrdt4qc.fsf>