From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from lists.gentoo.org ([140.105.134.102] helo=robin.gentoo.org) by nuthatch.gentoo.org with esmtp (Exim 4.67) (envelope-from ) id 1IGGQX-0006Vm-6m for garchives@archives.gentoo.org; Wed, 01 Aug 2007 15:47:57 +0000 Received: from robin.gentoo.org (localhost [127.0.0.1]) by robin.gentoo.org (8.14.0/8.14.0) with SMTP id l71FihKK015621; Wed, 1 Aug 2007 15:44:43 GMT Received: from relay.sgi.com (netops-testserver-4-out.sgi.com [192.48.171.29]) by robin.gentoo.org (8.14.0/8.14.0) with ESMTP id l71Figtl015564 for ; Wed, 1 Aug 2007 15:44:42 GMT Received: from cthulhu.engr.sgi.com (cthulhu.engr.sgi.com [192.26.80.2]) by netops-testserver-4.corp.sgi.com (Postfix) with ESMTP id 480DD61B59 for ; Wed, 1 Aug 2007 08:44:41 -0700 (PDT) Received: from conejo.engr.sgi.com (conejo.engr.sgi.com [150.166.8.69]) by cthulhu.engr.sgi.com (8.12.10/8.12.10/SuSE Linux 0.7) with ESMTP id l71FifLo002131 for ; Wed, 1 Aug 2007 08:44:41 -0700 Received: from conejo.engr.sgi.com (localhost [127.0.0.1]) by conejo.engr.sgi.com (SGI-8.12.11.20060308/8.12.11) with ESMTP id l71FieRc474143 for ; Wed, 1 Aug 2007 08:44:40 -0700 (PDT) Received: (from rsanders@localhost) by conejo.engr.sgi.com (SGI-8.12.11.20060308/8.12.11/Submit) id l71Fieo5477011 for gentoo-amd64@lists.gentoo.org; Wed, 1 Aug 2007 08:44:40 -0700 (PDT) Date: Wed, 1 Aug 2007 08:44:40 -0700 From: Bob Sanders To: gentoo-amd64@lists.gentoo.org Subject: Re: [gentoo-amd64] Is there any difference with 4 core? Message-ID: <20070801154440.GB474447@sgi.com> Mail-Followup-To: gentoo-amd64@lists.gentoo.org References: <46B0A3D5.6080203@singnet.com.sg> Precedence: bulk List-Post: List-Help: List-Unsubscribe: List-Subscribe: List-Id: Gentoo Linux mail X-BeenThere: gentoo-amd64@gentoo.org Reply-to: gentoo-amd64@lists.gentoo.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <46B0A3D5.6080203@singnet.com.sg> Organization: SGI, Mountain View, California, U.S.A. User-Agent: Mutt/1.5.11 X-Archives-Salt: 7af68581-09e8-4301-a5a3-b3faed03e877 X-Archives-Hash: 1cb4fbda5a739ad882585d3928684ddb P.V.Anthony, mused, then expounded: > Hi, > > Need some more advice. > > Now the question is, is it better to go with 2 core or 4 core? > > The reason for this question is, that I heard there is a diminishing > return with more cores. Not sure if this is true with kernel 2.6.21 and > running at 64bit. > As far as the kernel is concerned, there are no issues. I've run on a 1024 shared memory ccNUMA system and regularly started up and killed 1024 jobs withiout problems. > The server needed to built is for the following apps. > > Hardware. > 1. Tyan Tank GT20 (B5191) > 2. 2 x Sata drives > 3. Software raid 1 > 4. 4Gb ram ecc > 5. Intel Core2Duo E6420 or Intel Core2Quad E6600 > The Intel platform does not scale all that well. Plus it has limited I/O abilities. You will find a lot of issues with high i/o bandwidth simultanous applications. That said, it compiles like a screaming banshee. We currenly run ongoing testing on a small cluster of something like the above - 16 blades with quad core cpus (128P), running diskless over dual Infiniband links to lead and login nodes, dual core 1U boxes with hardware raid. > Apps. > 1. Gentoo linux 64bit > 2. Apache 2 > 3. MySql > 4. Postgres > 5. Qmail > 6. Pure-ftpd > 7. Mod_perl > 8. php > 9. ruby > > Will all the instances of the apps be shared among the cores? > Good question. Without something like cpu sets, the sharing won't be equal amoung all cpus. > Please share the comments. > > I would really like to save some money. If the 2 core can do the job > there is quite a bit of savings buying just the 2 core. > Regardless of Intel platform, FSB saturaton will ocur due to the ancient bus. Additionally, unless you run 4-dimms, max memory bandwidth will not be achieved. Typically for the Intel 5000 chipset the motherboards have 8 DIMM slots. I see the one in the tank only has 6. Intel's northbridge came split the data up over 4 dimms to maximize bandwidth. It's important to performance. Additionally, there are other power savings issues to note - Woodcrest (dual core) does not support EIST (enhanced speed step) Clovertwon (quad core) does support EIST Memory should be run in performance mode. Allowing memory to be set into dynamic power mode (power savings) will decrease the max available bandwidth. Of the power saving software, powernowd allows better tuning and response to load changes. Much better than cpufreqd. > The other option is just to go with Pentium D with 2 core. That one is a > real saver. > And it's a crappy cpu. Don't waste you're money. Any AMD cpu running at half it's speed will eat it alive. Bob - -- gentoo-amd64@gentoo.org mailing list