From owner-freebsd-amd64@FreeBSD.ORG Wed Jan 25 17:06:49 2006 Return-Path: X-Original-To: freebsd-amd64@freebsd.org Delivered-To: freebsd-amd64@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 8826C16A41F for ; Wed, 25 Jan 2006 17:06:49 +0000 (GMT) (envelope-from antonfb@hesiod.org) Received: from paris.hesiod.org (cust.hesiod.org [66.219.69.76]) by mx1.FreeBSD.org (Postfix) with ESMTP id 88E6343D6B for ; Wed, 25 Jan 2006 17:06:47 +0000 (GMT) (envelope-from antonfb@hesiod.org) Received: from [192.168.1.9] (atlas.hesiod.org [192.168.1.9]) by paris.hesiod.org (8.13.3/8.13.3) with ESMTP id k0PH6kIJ098718 for ; Wed, 25 Jan 2006 09:06:46 -0800 (PST) (envelope-from antonfb@hesiod.org) Message-ID: <43D7B026.10303@hesiod.org> Date: Wed, 25 Jan 2006 09:06:46 -0800 From: Jeff Anton User-Agent: Mozilla/5.0 (X11; U; FreeBSD amd64; en-US; rv:1.7.12) Gecko/20060114 X-Accept-Language: en-us, en-gb, en, fr-ca, fr-fr, fr MIME-Version: 1.0 To: freebsd-amd64@freebsd.org Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: dual vs single core opteron 100's X-BeenThere: freebsd-amd64@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting FreeBSD to the AMD64 platform List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 25 Jan 2006 17:06:49 -0000 I just finished building an Opteron 165 on a Tyan K8E S2865AG2NRF. I'm confident that dual core was the right move for me. Now I have a couple of my own database and computational applications and I've added threading support to most of these applications. Matrix multiplication threads like a dream. Even database stuff which is just reading records and aggregating improved 25% realtime at a cost of 165% cpu time relative to 70% cpu time single threaded. (a lot more thread processing overhead with the db stuff to get computation done while another thread waits for disk I/O.) Just operationally, X seems more interactive, but moving windows seems sluggish I think because the mouse processing produces more intermediate move points giving a smooth but sluggish window move. Mozilla is using threads. W/FreeBSD 6.0 run 'top' then hit 'H' to see each thread. My Mozilla currently has five threads. I have not yet tried overclocking. I think the Opteron will be fine. When building the machine I ran for 15 minutes without the cpu fan due to bumping the fan cable. (A little concerned that the system ran at all) After 15 minutes of fanless operation the BIOS reported the CPU was 122 deg C. Hard to believe, but I burned my hand reattaching the fan power. For the record the issues I had with putting the system together were: 1. The 2865 MB came with a BIOS v 1.0 which does not recognize dual-core (reported unknown processor) flashed to v 3.01 2. Hard drives were delivered late so I did a test FreeBSD install with and old small (20Gbyte) disk. FreeBSD 6.0 seemed to install but then crashed with a divide by zero error. This is a reported problem with the ataraid driver I learned by searching this lists archives. However, I also tried moving a disk from a working FreeBSD 6.0/i386 installation and that also crashed with a divide by zero error at the same place. That this happened with both the amd64 kernel and the i386 kernel makes me think this is strictly a motherboard (nVidia 4) ataraid driver issue not a strictly amd64 issue. Maybe the bug report should be re-classified. I tried a disk with a working FreeBSD 5.4/i386 installation and that actually ran but did not find the onboard ethernet. 3. The INSTALL notes still have several references to using floppy disks even though there are no floppy disk images. 4. When my SATA disks came I tried the 3.0 Gb mode since the Tyan documentation claims the board supports that, but FreeBSD 6.0 reported SATA150 and crashed being unable to write to the disks. Changed disks to 1.5 Gb mode and they worked. (Actually I had one crash due to 'lost mount' which was a cable problem I think.) 5. The reported problem with X11 not working because of no 'sync on green' support is a minor problem at best. Lack of 'sync on green' is only an issue with component cableing not an SVGA cable. Mind you the onboard ATI Rage XL has only 8 Mbytes of SGRAM (Tyan documentation here is poor) so you can not run with much resolution using the onboard graphics. Jeff Anton