From owner-freebsd-questions@FreeBSD.ORG Sat Feb 5 16:45:37 2005 Return-Path: Delivered-To: freebsd-questions@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 6F54C16A4CF for ; Sat, 5 Feb 2005 16:45:37 +0000 (GMT) Received: from imo-m25.mx.aol.com (imo-m25.mx.aol.com [64.12.137.6]) by mx1.FreeBSD.org (Postfix) with ESMTP id BC06D43D3F for ; Sat, 5 Feb 2005 16:45:36 +0000 (GMT) (envelope-from Freebsd9999@aol.com) Received: from Freebsd9999@aol.com by imo-m25.mx.aol.com (mail_out_v37_r3.8.) id n.9a.1f89d2ab (14374) for ; Sat, 5 Feb 2005 11:45:32 -0500 (EST) From: Freebsd9999@aol.com Message-ID: <9a.1f89d2ab.2f36522c@aol.com> Date: Sat, 5 Feb 2005 11:45:32 EST To: questions@freebsd.org MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit X-Mailer: 9.0 for Windows sub 5116 Subject: Re: Intel EMT64 Xeon vs AMD Opteron X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 05 Feb 2005 16:45:37 -0000 In a message dated 2/4/05 11:29:46 PM Eastern Standard Time, scottl@freebsd.org writes: > D'oh. One other thing. In the benchmarks I've seen, Opterons "Play > Nicer" with SMP because of the Hypertransport setup in some > applications. (IE, they don't fight over memory the way Xeons do). > Look for a motherboard that uses a "4+4" or "4+2" memory configuration > to take full advantage of this. (Differnt memory for each processor, > kinda) :With FreeBSD, it's a bit of a toss-up. There is no strong affinity :set or enforced between process memory and where the process is running. :Having some notion of affinity (i.e. NUMA support) would be a good :thing. Oh, and the 4+2 configurations are typically pretty poor, :regardless. And since i386 is clearly a better supported platform because of its must wider use, it make sense to choose the better supported platform in the case of a performance wash, at least until there is compelling evidence that an opteron platform is better in any way. Arguing architecture is completely and totally useless. The proof of the pudding is in the taste, and just because an architecture has the potential to be better doesn't mean that it plays out that way in real world implementations. And my dealer has a 3.0Ghz xeon at $130. less than opteron 246s, so where are you doing your shopping, if you think that they are "about the same"?