Return to BSD News archive
#! rnews 2498 sserve.cc.adfa.oz.au Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!simtel!news.sprintlink.net!howland.reston.ans.net!Germany.EU.net!nntp.gmd.de!news.rwth-aachen.de!news.rhrz.uni-bonn.de!RRZ.Uni-Koeln.DE!se From: se@MI.Uni-Koeln.DE (Stefan Esser) Newsgroups: comp.unix.bsd.netbsd.misc Subject: Re: Commerical OS's (was Re: aha0: DMA beyond end of ISA) Date: 18 May 1995 11:57:34 GMT Organization: Institute for Mathematics, University of Cologne, Germany Lines: 48 Message-ID: <3pfcne$ab6@news.rrz.uni-koeln.de> References: <MICHAELV.95May12222703@MindBender.HeadCandy.com> <1995May16.114607.16086@zippy.dct.ac.uk> <3pc8u8$66g@news.cloud9.net> <3pd7s5$6i9@shell1.best.com> NNTP-Posting-Host: fileserv1.mi.uni-koeln.de In article <3pd7s5$6i9@shell1.best.com>, rcarter@best.com (Russell Carter) writes: |> In article <3pc8u8$66g@news.cloud9.net>, |> Thor Lancelot Simon <tls@cloud9.net> wrote: |> >I am not happy with the performance I get from the NetBSD/FreeBSD ncr |> >driver, however. I plan to switch the '825 system to a BT956C as soon as |> >I can get my hands on one. The 53c8XX hardware has a reputation for good |> >performance, so I'm pretty sure this is a driver issue. |> The FreeBSD and NetBSD drivers are related, but perhaps not the same? Trust me, they ARE the same ! (And I've got to know, as you may remember :) |> I see better than 4 MB/s from most 5400 rpm drives, and better than |> 5.5 MB/s from 7200 rpm drives with the FreeBSD ncr 810 driver. I know of a |> system with 4 ncr 53c810 cards driving 4 5400rpm disks at an |> aggregate bandwidth > 12 MB/s using the same driver. That's pretty |> good performance, IMHO. Well, and I just checked the maximum command rate (i.e. used "dd if=/dev/rsd0a of=/dev/null bs=512" to do minimum size transfers). I got some 1600 transfers/s using my Quantum Atlas connected to a NCR 53c810 and driven by an 486/66. This makes the per command overhead less than 600 microseconds total (i.e. arbitration, send command, send a few messages, receive data, receive status, process status plus all system call and library overhead !) I'd like to know the number of 512 byte reads per second of other controllers. Use iostat to check for the number of transactions/s. (Read the RAW device, or you will get the buffer cache into the way !) Regards, -- Stefan Esser Internet: <se@ZPR.Uni-Koeln.DE> Zentrum fuer Paralleles Rechnen Tel: +49 221 4706017 Universitaet zu Koeln FAX: +49 221 4705160 Weyertal 80 50931 Koeln