Date: Sat, 28 Jan 2006 14:40:39 +0100 From: "O. Hartmann" <ohartman@mail.uni-mainz.de> To: JoaoBR <joao@matik.com.br> Cc: kono@kth.se, Daniel Rock <freebsd@deadcafe.de>, mumag@nist.gov, freebsd-amd64@freebsd.org Subject: Re: dual vs single core opteron 100's Message-ID: <43DB7457.6050605@mail.uni-mainz.de> In-Reply-To: <200601280912.52907.joao@matik.com.br> References: <200601270957.44032.kono@kth.se> <43DA6906.9030508@deadcafe.de> <200601280912.52907.joao@matik.com.br>
next in thread | previous in thread | raw e-mail | index | archive | help
This is a multi-part message in MIME format. --------------060904080806080005020605 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit JoaoBR wrote: > On Friday 27 January 2006 16:40, Daniel Rock wrote: > >> Alexander Konovalenko schrieb: >> >>> I have upgraded my AMD64 Athlon 3000+ to dual core X2 4400+. Now I can >>> run two oommf tasks at the same time, and performance (I measure total >>> execution time of the task) is around 186% comparing with 100% when only >>> one task is running. This 7% degrade in performance per task is probably >>> due to concurrent data transferring CPU<->RAM. I am very satisfied with >>> X2 but just wonder if dual core Opteron gives better performance? Does >>> anybody run OOMMF on Opteron? >>> >> Besides the different CPUID, Athlon64 X2 (with 2x1MB Cache) and Dual-Core >> Opteron 1xx are exactly the same. >> >> > > Hi > I believe this is not exactly correct > > If I am not wrong there is actual only the X2-4800 and 4400 with 2x1MB cache, > all others are having max 512K We were talking about comparable Athlon64/Opterons for socket 939. > > > The Opterons are having 2MB of cache, not 2x1 > No, that's wrong. Dual-core 1XX Opterons do have two times 1MB 2nd level cache for each core, not a unified 2MB cache! > the manufacture technology of opterons and athlons-X2 are quiet different and > there other tiny "server" related points > As you can read in many technical reports and newsgroups, Opteron 1xx for s939 and Athlon64 of comparable cache size use identical core logic. The only difference in between is the Athlon64 has a unlocked clock multiplicator. > (for me doubtless) a server-MB for dual opteron is also faster then a US100 > 939 socket MB and probably much more stable > > IMO the opterons are faster when talking server, nevertheless the athlons are > much cheaper, not the processors but I can run real cheap MBs Whether a socket 939-MB built for hosting 1xx Opterons will accept a Athlon64 chip depends on BIOS - a simple recognition of the CPUID! Tests revealed that dual core Opteron 1xx for socket 939 and Athlon64-X1 are at same performance at the same speed (assuming the same cache size). There is no performance difference if both chips have the same technical specs. > > > when I do not need more than 3.5GB of RAM the performance difference is not > that much so the athlon-X2 are an attractive alternative for US500-1000+ less > > I have lot's of cache/gw servers and I am changing all to X2, the disk r/w > performance advantage with good memory chips is very big in comparism to > i386 P4 machines. > > I run also expensive perl tasks on this servers which gave my a bottleneck on > UP a machines, the X2 SMPs are managing this as perfect as my dual-otperon > server and for me the X2 is a very very good and cheap solution. > > I can not say anything for workstations but as long as you do not pass more > than 6-8MB/s traffic through the machine a X2 Athlon may do it as good as an > Opteron system but probably depends also on what you do particulary with this > machine. > > Joćo > > All right, talking about scientific calculations, I would prefer a real dual chip plus dual core Opteron 2xx system due to the fact, memory bandwith is much more efficient when using NUMA architecture. But in my opinion, comparin dual chip and dual core Opterons (and their Athlon64 counterparts) is comparing cherries and pineapples ... > > > > A mensagem foi scaneada pelo sistema de e-mail e pode ser considerada segura. > Service fornecido pelo Datacenter Matik https://datacenter.matik.com.br > _______________________________________________ > freebsd-amd64@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-amd64 > To unsubscribe, send any mail to "freebsd-amd64-unsubscribe@freebsd.org" > --------------060904080806080005020605--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?43DB7457.6050605>