Author: Robert Hyatt
Date: 15:06:48 02/18/03
Go up one level in this thread
On February 18, 2003 at 16:43:58, Matthew Hull wrote: >On February 18, 2003 at 16:41:44, Robert Hyatt wrote: > >>On February 18, 2003 at 15:57:03, Matthew Hull wrote: >> >>>On February 18, 2003 at 15:35:37, Charles Worthington wrote: >>> >>>>On February 18, 2003 at 15:27:55, Drexel,Michael wrote: >>>> >>>>>On February 18, 2003 at 14:44:10, Charles Worthington wrote: >>>>> >>>>>>Actually the 3175 kNs was derived at from a mathmatical formula: 1200 kNs for >>>>>>single 3.06 cpu times 1.85 (For Duals) then that figure times 1.33 for >>>>>>hyperthreading. Until I have the system in hand to test it the math is all i >>>>>>have to go by. Why hasnt anyone tested the dual 2.8 Xeons by now? They have >>>>>>certainly been on the market long enough that someone here would have tested >>>>>>them other than Bob. >>>>> >>>>>Unfortunately Bob dont like to install Fritz 7 or Fritz 8. I assume he only >>>>>would have to send chessbase a mail and they would ship him all products for >>>>>free. >>>>>He is probably Winboard-Crafty Fan :) >>>>>CC enthusiasts in Germany generally prefer Dual AMD systems. So do I >>>> >>>>I can't say which system I prefer yet. I will soon have both in hand then I will >>>>conduct unbiased tests. The germans prefer AMD because they don't run the new >>>>Xeons. The price in Germany is a major factor. Price-wise I prefer the AMD as >>>>well but i was not looking for the best value i was looking for the best >>>>performance with cost not an object. >>> >>> >>>Talk about money being no object... >>> >>>Does anyone know what Crafty performance is like on the latest s390 hardware >>>under Linux? >>> >>>If I knew anyone at IBM, I'd quiz them as to how well Linux scales to a 16-way >>>Z900 or (8-way for that matter) running on the bare metal (as opposed to running >>>under VM). And if it does scale, then compile crafty on that and at least find >>>out the n-way speedup factor for more than 4 processors. >>> >>>Matt >> >> >>That's not a particularly good architecture for chess. I've run on several in >>the past, but >>they were pretty much dogs for chess. Lots of I/O throughput, but not good >>number- >>crunching platforms at all.. > > >What about memory? Do you know how they handle shared memory across n >processors? Even though the machines might be slow, would a test of n-way > 4 >be interesting? > >Thanks, >Matt They do OK there, but as I said, the goal for that machine family was really high I/O throughput, not high computational throughput. IE it was never intended to be any sort of supercomputer/number-cruncher machine at all. multiple processors are always interesting from a test perspective of course... > > >> >> >> >> >>> >>> >>>>All I am saying is that the AMD users have >>>>not actually owned a newer Xeon system so when they quote performance specs they >>>>have no idea what they are talking about. My Xeon system i ordered from Dell is >>>>the very first one ordered in the 3.06 platform with the new E7505 Chipset. No >>>>one can claim to know that the AMD is faster than this box until we see for sure >>>>by testing them. What will they say if the Xeon turns out to be significantly >>>>faster?
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.