Date: Thu, 4 Nov 1999 13:04:26 +0100 (CET) From: Remy Nonnenmacher <remy@synx.com> To: Joachim.Strombergson@emw.ericsson.se Cc: freebsd-smp@FreeBSD.ORG Subject: Re: Good SMP Motherboards Message-ID: <199911041204.NAA05362@gw0.boostworks.com> In-Reply-To: <3821378D.D532AB2@emw.ericsson.se>
next in thread | previous in thread | raw e-mail | index | archive | help
On 4 Nov, Joachim Strombergson wrote: > Tja! > > On 03-Nov-99 Bruce Albrecht wrote: >> I'm looking for a good SMP motherboard for $150 or less. Any >> recommendations? If I want also onboard SCSI, what's a good, cheap MB >> run? > > I don't know how much they cost in the US, but when I bought my ABIT BP6 > in June/July this summer I paid approx 160 USD for it. The board is a > PPGA370 board for Celerons. I use two Celery 366:s on it and get a rock > stable system @ 467MHz. I really like the extensive settings of the BIOS > that includes temp watch, individual voltage control and fan control. > > I haven't done any comparison with other systems, I just appreciate how > much faster it is compared to my old, P133 system. :-) > > BTW: Has aybody done any FreeBSD-SMP system rating? What would be good > benchmarks? > Here is some using the rc5des client software (latest version) : Abit BP6 2x466 Celerons, -CURRENT : 2.60 Mkey/s ( 100$/proc) C440GX+ 2xPIII Xeon 500, -STABLE : 2.73 MKey/s (10000$/proc) L440GX+ 2xPIII 500 , -CURRENT : 2.82 Mkey/s ( 260$/proc) No overclocking. The C and L 440 are pure Intel boxed rack servers (SC440 and LB440GX). All Mobos uses ECC-SDRAM, no EDO. Remarks: - We would have expected the Xeon machine to perform better than the PIII standard processor. In this matter, rc5des and dry say : No real difference. - CURRENT implements a kind of processor affinity that seems helpfull. - I can't upgrade the Xeon machine to CURRENT since it's a production machine but I hope to see improvements as soon as it will be. (Jordan, fork, fork !! ;) ). - Linear interpolation for a Celeron 500 would lead to 2.80 Mkey/s which is nearly what gave the PIII. If this is exact, the performance/price ratio is outstanding. - This is a pure core test. Do not expect same ratio on dayly operations. - David Malone have a really interesting graph that shows impact of cache size and RAM speed over a core-intensive process. It can be found at : http://www.maths.tcd.ie/~dwmalone/comp/perf.ps and shows the L1, L2 and RAM speed and size steps. Speculation: Intel build the same core every time. If the cache runs full speed, it's a Xeon. If only half of the cache runs at full speed, it's a PIII, if a quarter of the cache runs at full speed it's a celeron. If the cache runs at half speed, it's a PII. If nothing works, back to the fundry. Probably exagerated, but it's the basic idea. RN. IeM To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199911041204.NAA05362>