Return to BSD News archive
Newsgroups: comp.unix.bsd.freebsd.misc Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!simtel!lll-winken.llnl.gov!fnnews.fnal.gov!gw1.att.com!nntpa!not-for-mail From: dyson@inuxs.inh.att.com (John S. Dyson) Subject: Re: IDE diskIO Message-ID: <DDB6CE.ABM@nntpa.cb.att.com> Sender: news@nntpa.cb.att.com (Netnews Administration) Nntp-Posting-Host: inuxs.inh.att.com Organization: AT&T References: <40l85l$bfb@trauma.rn.com> Date: Mon, 14 Aug 1995 16:12:13 GMT Lines: 43 In article <40l85l$bfb@trauma.rn.com>, Larry Snyder <larry@rn.com> wrote: >I'm running 2.0.5R with a WD Caviar 1.2 gig IDE drive >connected to my PCI based EIDE controller. The WD drive >spins at 4500 RPMs and has 64k of cache. > >I was considering replacing the Caviar 1.2 gig with a >pair of EIDE Segate 850 meg drives - each one spins at >5400 RPMs and has 256 of cache. I was wondering if I >would see a significant increase in diskIO with the pair >(or even one) of the Seagate drives verses the WD >(due to the limiting factor with the WD being A, the cache >and B, the RPM's). > >Any comments? If you are connected through a PCI interface or a another fast EIDE interface, I think that you will normally see the performance associated with that interface as set-up by the bios. Except for initialization and ignoring DMA and ATAPI -- programming EIDE vs IDE is the same. We also support 32Bit and Multi-block I/O as config options. I have done measurements and have found that on an EIDE (high speed) interface there is a speed increase and FreeBSD does take advantage of it. (I easily get 2.6MB+ and have measured around 2.8MB through the filesystem with reasonable kernel overhead on PCI interfaced Caviar drives.) I only get about 2.2MB/sec on my Caviars on an ISA IDE interface however. > >Maybe I should consider Linux since it supports the PIO and >faster EIDE transfer rates? > Well, the best way to compare is to measure the filesystem throughput and the kernel overhead and look at the results. The ATA-2 spec contains lots of features that drives can optionally support. FreeBSD does appear to get at least part of the performance associated with the higher transfer rates (since I guess the bios sets up some of the interface speed parameters) The I/O requests (except ATAPI and DMA) are identical between the high transfer rate modes and the normal ones. John dyson@freebsd.org