From mboxrd@z Thu Jan 1 00:00:00 1970 From: Richard Scobie Subject: Re: Resets on sil3124 & sil3726 PMP Date: Tue, 04 Sep 2007 09:21:59 +1200 Message-ID: <46DC7AF7.1010304@sauce.co.nz> References: <46DC62E4.8050901@sauce.co.nz> <20070903203441.GF23498@curie-int.orbis-terrarum.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from smtp.sauce.co.nz ([210.48.49.72]:55140 "EHLO smtp.sauce.co.nz" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754838AbXICVR4 (ORCPT ); Mon, 3 Sep 2007 17:17:56 -0400 Received: from [192.168.4.111] (unknown [192.168.4.111]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by smtp22.sauce.co.nz (Postfix) with ESMTP id 1A70F288263 for ; Tue, 4 Sep 2007 09:17:53 +1200 (NZST) In-Reply-To: <20070903203441.GF23498@curie-int.orbis-terrarum.net> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: linux-ide@vger.kernel.org Robin H. Johnson wrote: > The single PMP numbers they have (under "Addonics ADSA3GPX8-4EM Striped RAID Set > Performance Comparison") has Write=211MB/sec Read=231MB/sec. True. I wonder if the bus spec of 3Gb/s is somewhat optimistic in the real world - a bit like saying one can get 132MB/s from a 32bit 33MHz PCI bus. The absolute ideal throughput for 1.5 Gb/s would be 187MB/s, so the throughput figures above mean the link must be at 3Gb/s, it's just not inpressive, given that the 5 drives can do at least 300MB/s and the theoretical maximum for 3Gb/s is 375MB/s. In another test at the same site, a Marvell 88SX7042 based card gets slightly better throughput - 196.8 write 251 read: http://www.amug.org/amug-web/html/amug/reviews/articles/sonnet/e4p/ Looking at the graphs towards the end of the article "Seagate 160GB 7200.9 Port Multiplier Write/Read Performance", it looks as though better performance may be had by distributing disks across multipole PM ports, rather than all on one - see 5 drives on one PM v 4 drives, each on it's own PM. There are quite a number of tests of many PM cards and enclosures here: http://www.amug.org/amug-web/html/amug/reviews/reviews.html Regards, Richard