Date: Sat, 3 Jun 1995 16:20:49 +1000 From: Bruce Evans <bde@zeta.org.au> To: henrich@crh.cl.msu.edu, rls@kilroy.id.net Cc: freebsd-hackers@FreeBSD.ORG, rgrimes@gndrsh.aac.dev.com Subject: Re: A performance mystery Message-ID: <199506030620.QAA04316@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
> Compaq Proliant 575 Micron P90 Powerstation >time make 8:55 9:39 >ram-speed 10.85 Mb/S, 7.08 Mb/S 10.61 Mb/S, 20.66 Mb/S >iozone 200 2486434/4150208 2446772/4203499 >iozone 16 8192 3360694/2677660 2711469/13256071 >The make time is wall time, ram-speed tests the raw memory bandwidth, defeating >cache almost 100%. iozone 200 tests raw read/write performance to disk, as it >far exceeds ram size, memory cache effect is insignificant. iozone 16 8192 >tests the memory speed with cache involved, because the test size is so small, >(as Rod pointed out) its essentially just a bunch of bcopy's. >As you can see by the ram-speed and iozone tests (and the fact that its a >Pentium-75!) the Compaq should be sorely beaten in build time over the Micron, >but yet its not. Hence the mystery. Apparently the Compaq has a much better (memory) cache and it gets used a lot for building the kernel. Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199506030620.QAA04316>