Return to BSD News archive
Xref: sserve alt.comp.periphs.mainboard.asus:149 comp.os.386bsd.questions:16825
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!bunyip.cc.uq.oz.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!nctuccca.edu.tw!news.cc.nctu.edu.tw!news!gate!lim
From: lim@gate.sinica.edu.tw (Carmay Lim)
Newsgroups: alt.comp.periphs.mainboard.asus,comp.os.386bsd.questions
Subject: Re: Problems with PVI-486AP4, NCR53c810 and FBSD 2.0
Followup-To: alt.comp.periphs.mainboard.asus,comp.os.386bsd.questions
Date: 28 Feb 1995 10:02:44 GMT
Organization: Computing Center, Academia Sinica
Lines: 46
Message-ID: <3iusc4$kvb@gate.sinica.edu.tw>
References: <3iab71$5qu@gate.sinica.edu.tw> <SEB.95Feb24161311@scotch.eua.ericsson.se> <3is55v$eq7@gate.sinica.edu.tw> <3it2ta$n8c@deep.rsoft.bc.ca>
NNTP-Posting-Host: lim%@gate.sinica.edu.tw
X-Newsreader: TIN [version 1.2 PL2]
Curt Sampson (a09878@giant.rsoft.bc.ca) wrote:
:
: Is this using iozone or raw transfers? I can get 4 MB/sec if I dd from
: /dev/sd0d to /dev/null,
Ugh, don't remind about dd... I somehow accidentally did a 'dd
if=/dev/sd0d of=/dev/sd0b' or something, and overwrote the really
important parts of my FreeBSD slice. :( I'm getting really good at
installing FreeBSD though. ;-)
: but iozone only gives me 800-900 K/sec. I'm
: using a Quantum Lightning 350, which I gather is a pretty fast drive.
I have a Lightning 540 in the other PC, but it is on an ISA
controller (giving me about 800K/sec with iozone). I ran iozone on the
Empire drive while the system was in single-user mode. The reported
throughput was just under 4000000 bytes per second (okay, not quite 4
megabytes per second). Running it now under X with a make running at
the same time:
> # iozone 50 65536
>
> IOZONE: Performance Test of Sequential File I/O -- V2.01 (10/21/94)
> By Bill Norcott
[...]
> Writing the 50 Megabyte file, 'iozone.tmp'...15.046875 seconds
> Reading the file...14.421875 seconds
>
> IOZONE performance measurements:
> 3484364 bytes/second for writing the file
> 3635366 bytes/second for reading the file
Using a more realistic 1K block, I get:
> IOZONE performance measurements:
> 1955956 bytes/second for writing the file
> 2634819 bytes/second for reading the file
I think I'm running into FreeBSD's I/O throughput limit on my
system here (2000 to 2500 operations per second) rather than the
bandwidth over the SCSI bus or the speed of the drive.
: I'm running NetBSD 1.0 with what I gather is the latest patch version
: of the NCR driver (V2 pl16 95/02/21).
The ncr.c Stefan sent me is pl18 (95/02/23). pci.c is pl3 (95/02/25).