Author: Keith Evans
Date: 20:27:49 04/23/03
Go up one level in this thread
On April 23, 2003 at 22:08:41, Matt Taylor wrote: >On April 23, 2003 at 01:01:37, Keith Evans wrote: > >>On April 23, 2003 at 00:43:27, Tom Kerrigan wrote: >> >>>On April 22, 2003 at 22:09:16, Aaron Gordon wrote: >>> >>>>On April 22, 2003 at 21:20:15, Eugene Nalimov wrote: >>>> >>>>>Here are *official* results for Spec2k. Please notice that Athlon benchmarks >>>>>were submitted by AMD itself. >>>>> >>>>>http://www.spec.org/osg/cpu2000/results/res2003q1/cpu2000-20030224-01964.html >>>>>http://www.spec.org/osg/cpu2000/results/res2002q4/cpu2000-20021202-01875.html >>>>> >>>>>So: base Spec2k for P4/3.06 is 1099. For Athlon XP 3000+ score is 995. Higher is >>>>>better. >>>>> >>>>>Thanks, >>>>>Eugene >>>> >>>>I've done official testing for AMD using AMD's methods. This was when I was >>>>working on the optimized Quake 3 dlls. They had me disable everything in the >>>>bios. This means the test took a pretty large hit performance wise. Why? I asked >>>>AMD the same thing. They responded with, "Intel doesn't think it's fair, so if >>>>we set the bios timings to the fastest settings possible we'd have a large >>>>lawsuit on our hands and AMD doesn't need that". If you need confirmation of >>>>what I'm saying email me at speedycpu@attbi.com and I'll give you the contact >>>>information to the guy at AMD and he'll verify everything I've said. >>>> >>>>So, for a properly configured Athlon, my results are there and plain as day. >>>>Like I said, run them yourself on the same systems I ran them on. >>> >>>I don't see how that works. Intel has the "performance at all costs" reputation >>>for SPEC scores, even going so far as to use its committee clout to make >>>profile-directed optimizations allowed for base scores, and now you're saying >>>they use artificially slow memory timings? You can be sure that the competition >>>(Sun, IBM, HP, etc.) runs their memory as fast as possible--is Intel going to >>>sue them, too? Also, Intel submits slightly higher scores than Dell for the same >>>processors. Does Dell also run its memory slow? And what would the charge be for >>>this lawsuit, anyway? And besides, why do slow memory timings hurt AMD and not >>>Intel? >>> >>>It's one thing to suggest that some sites might be somewhat biased in Intel's >>>favor to get free stuff from them, but in this free-press society, not all sites >>>can be biased, or it would be a major coup for the one that does the exposee. >>>Besides, what benefit would aggressively anti-Intel sites (e.g., AMDZone) get >>>from biasing their reviews towards Intel, and their reviews are remarkably >>>similar to other sites' reviews. >>> >>>Suggesting that all hardware review sites are biased and that Intel, Dell, and >>>AMD are all part of a conspiracy to artificially lower their own SPEC scores... >>>did you forget your tin foil hat today? >>> >>>-Tom >> >>Well if they thought that said settings would produce unreliable behavior, then >>they might feel uncomfortable quoting performance under said conditions. Makes >>sense... > >The default settings are good enough for consumers but unreliable for SPEC? > >-Matt No the default settings are good enough for consumers, and apparently good enough that they use them for SPEC. The "optimized" settings may be operating components out of specification, so they could feel uncomfortable quoting results obtained with those in benchmarks. For example do those people who tweak BIOS settings related to DRAM know how to read a datasheet and verify that all of the parameters are being met? Tom was basically wondering why they might "hold back", and I offered a potential reason. I doubt that they would intentionally cripple their benchmarks. Believe me I've worked in the electronics industry for a while, and marketing people will do anything possible to quote good numbers. I remember back when Diamond was shipping overclocked graphics cards. By default they would be overclocked, and then we people had trouble they would call up and the tech support guys would tell them to edit a ".ini" file to fix the problem. I'm sure the cards that went to reviewers were carefully screened so they could be overclocked and produce good benchmarks.
This page took 0.01 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.