Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!news.ecn.uoknor.edu!paladin.american.edu!zombie.ncsc.mil!news.duke.edu!news.mathworks.com!newsfeed.internetmci.com!news.msfc.nasa.gov!elroy.jpl.nasa.gov!lll-winken.llnl.gov!fnnews.fnal.gov!gw1.att.com!nntphub.cb.att.com!not-for-mail From: dyson@inuxs.inh.att.com (John S. Dyson) Newsgroups: comp.unix.bsd.bsdi.misc Subject: Re: PentiumPro BSDI performace ? Date: 29 Dec 1995 16:17:45 GMT Organization: AT&T Lines: 27 Message-ID: <4c14b9$ec4@nntpb.cb.att.com> References: <4btp7t$nno@clarknet.clark.net> <4bv16t$3ml@web.ddp.state.me.us> <VIXIE.95Dec28232327@wisdom.vix.com> NNTP-Posting-Host: inuxs.inh.att.com In article <VIXIE.95Dec28232327@wisdom.vix.com>, Paul A Vixie <vixie@vix.com> wrote: >: I was wondering if the new PentiumPro 200 machines would acheive good >: performance under BSDI. The PentiumPro is a highly optimized chip and >: the 16-bit app performance is poor. How does this impact the BSDI os ? > >I just got a loaner Intel Orion with a P6-150. I'll benchmark it over the >weekend and report back what I find. I don't know if I have a compiler that >can take advantage of the P6 (or the P5, GCC 2.7.x notwithstanding.) > This info is second hand, but I believe accurate (I talk to David regularly.) David Greenman on the FreeBSD project has been benchmarking and tuning up the P6 processor about to be installed as wcarchive. He has noted a couple of interesting problems -- the most noteworthy is that the PCI to memory DMA performance appears to be limited to 4MBytes/sec. The problem is going to be fixed in a future version of the chipset (might already be out.) Just a caveat if you see the problem. It was manifest by slow (4Mbyte/sec) iozone perf, and missing UDP packets on a 100BaseT network. The problem is not noticeable on a 10BaseT network though. All in all, especially with the P6 improvements going into FreeBSD, the system screams and performs significantly better than a P5, but there are the above glitches. John dyson@freebsd.org