Author: Matt Taylor
Date: 22:50:56 02/18/03
Go up one level in this thread
On February 18, 2003 at 20:33:29, Charles Worthington wrote: >On February 18, 2003 at 14:29:41, Charles Worthington wrote: > >>On February 18, 2003 at 14:22:43, Joachim Rang wrote: >> >>>On February 18, 2003 at 14:02:37, GuyHaworth wrote: >>> >>>> >>>>There was a 'Chess Duel' last year which featured Fritz and Shredder, Intel and >>>>AMD afai-remember. >>>> >>>>Deductions could be made about Fritz-v-Shredder and Intel-v-AMD, at least in the >>>>hardware configured. >>>> >>>>Maybe someone here still has the URL to the results. >>>> >>>>g >>> >>>http://www.heise.de/ct/schachduell/default.shtml >>> >>>unfortunately all is in german :-( >>> >>>well the conclusion was, that AMD MP2000+ is about equal to Xeon 2.4 Ghz. >>>So I expect, that AMD MP2600+ is only slightly worse, than Xeon 3.06 Ghz >>>(perhaps Hyperthreading will give the Xeons an edge). >> >> >>Yes I also feel that the hyperthreading is a factor. Maybe up to a 33% factor. >>But thats why I'm testing so that we can all know whether to take that into >>account when deciding on a purchase. I expect the Xeon to outperform the MP2600+ >>on two threads alone. On 4 threads I am pretty sure it will far exceed the 2600. > > > >Below is a link to an article which fully explains hyperthreading technology and >it's benefits. After reading the article even a novice like myself can see how >chess programs will benefit from the technology significantly. There is far more >involved here than merely getting an extra thread or two for your program. >Hyperthreading also fills in the data "holes" as the processed information >passes through the Front Side Bus. This results in up to a 50% increase in the >amount of work being done by the cpu's. The technology is not a "hoax" designed >to merely create virtual processors to "trick" your program. Since Chess engines >are resource hungry and tend to use 100% of the available cpu processing power, >it only makes sense to put the front side bus to work and make it earn it's >keep. Especially if it is a 533FSB chipset we are talking about. It is clear >from this article that hyperthreading is the single most important advancement >in CPU Technology in recent years. Yes, the Intel Processor's are expensive; That's debatable. There are -many- other advances in the past 5-6 years that were important. If I may date as far back as 1995, advances in Intel architecture (not always unique or novel) include SIMD, branch prediction improvements, data prefetching, superscalar architecture, etc. By the way, Intel's claim is 30-40% performance increase from HT. Yes >the AMD gives more performance for the dollar; but no the AMD will not perform >better overall than the New generation Xeon processors utilizing hyperthreading >technology and anyone who fully understands the technology, or owns both >systems, would not claim otherwise. Intel spends more money in research and >development per year than AMD's entire annual budget so naturally their >processors are going to be more expensive. While AMD has been striving to design >processors that simply faster, Intel has been looking at ways to make them >faster AND more efficient and their research has now come to fruition. Below is >a link to a detailed technical artice which explains the technology for any who >are interested. it was a very fascinating article and I highly recommend it: >http://arstechnica.com/paedia/h/hyperthreading/hyperthreading-1.html The Athlon has changed very little since its inception 5-6 years ago. As for making a processor more efficient, that is wrong. It is trivial to look up thermal dissipation data for AMD and Intel processors. Prior to HT it was quite obvious which processor had better performance/Watt. I think it is still very much one-sided, but it is more difficult to say. By the way, you must also consider AMD's strategy as well as Intel's before you hop on the Intel train. Hyperthreading fills in gaps between memory accesses with useful work. The x86-64 strives to minimize those gaps. Memory latency on my dual-Athlon is about 133-ns. On x86-64, AMD docs say it will be within 80-ns. Somebody (I can't remember who) claimed that latency could be as low as 50-ns in certain configurations. Furthermore AMD is striving to get 64-bit computing on the desktop. That will have a profound effect on bitboard engine efficiency. -Matt
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.