Author: Mark Young
Date: 11:05:27 07/10/98
Go up one level in this thread
On July 10, 1998 at 00:35:52, Robert Hyatt wrote: >On July 09, 1998 at 21:56:06, Mark Young wrote: > >>On July 09, 1998 at 21:32:53, Robert Hyatt wrote: >> >>>On July 09, 1998 at 20:34:27, Mark Young wrote: >>> >>>>On July 09, 1998 at 18:52:58, Robert Hyatt wrote: >>>> >>>>>On July 09, 1998 at 18:37:15, Ed Schröder wrote: >>>>> >>>>>>I just received some information from Rob at Kryotech who is the hardware >>>>>>sponsor for the Rebel-Anand match. >>>>>> >>>>>>On the BT-16 position Rebel9 with 28 Mb hash reported 19,434,138 nodes >>>>>>after a full 9 ply search. >>>>>> >>>>>>On a PII-266 Rebel9 needed 4:03 to do the 9 ply search. >>>>>> >>>>>>The Kryotech AMD-450 only needs 1:45 (!!) >>>>>> >>>>>>It's a beast this machine, no doubt.... >>>>>> >>>>>>When I get the machine next week I will do the bench mark test and >>>>>>put the results on the Rebel Bench Mark List. >>>>>> >>>>>>- Ed - >>>>> >>>>> >>>>>If you would (a) dump DOS, (b) go to WinNT, you could do a parallel search >>>>>and run on one of the quad 450mhz pentium II (Xeon) that have just been >>>>>announced. >>>> >>>>>The Xeon will toast the AMD clock for clock, >>>> >>>>The Xeon core is still the well-known 'Deschutes' core, used in the Pentium II >>>>as well as in the Celeron processor. And the benchmarks I have seen show no >>>>performance advantage over the Deschutes in a one-chip configuration. So I would >>>>think that the AMD chip is still the chip for ED to use until he does a parallel >>>>search with rebel. But you are right the PII and the Xeon is faster clock for >>>>clock running most programs, but not Rebel in dos. >>>> >>>>>and 4 of 'em would >>>>>be one fast machine. And there will be 8 processor versions by September. >>>>> >>>>>But, alas, not for dos nor win95/98. >>> >>> >>>the core hasn't changed, that's correct. However Xeon behaves just like the >>>original P6 chip, where the cache now operates at full cpu clock speed, rather >>>than being able to deliver data to the cpu only on every other clock cycle. >>> >>>My first benchmark on a PII/300 showed 1.41 times faster than a P6/200. That >>>missing .09 (20%) is caused by the 1/2 speed cache on the PII. If you take this >>>up to the PII/400, you begin to see a significant performance loss when compared >>>to a P6/200, factoring in the 2x faster clock not giving anywhere near 2x the >>>cpu performance. Xeon will. The 450 should clock in just as you'd expect with >>>crafty, running 2.25X faster. You aren't going to run an AMD at 450 yet, which >>>means that the Xeon will "toast" it pretty thoroughly. And when you factor in >>>quad processors, it's no contest, and AMD hasn't been able to make a multi- >>>processor specification that anyone is willing to implement... >> >>The benchmarks I have seen this did not happen. The faster Cache did not help >>the Xeon chip over the P II when running just a single program. Now when the >>chip was made to multi-task thats when the chip had about a 3% to 25% boost over >>the P II running at the same clock speed. If this was not done the Xeon and the >>P II benched the same. > > >Somehow our benchmarks are different. First, I don't see how the Xeon is better >at context-switching than a normal PII.. that is independent of cache >completely. however, I have benched my P6/200 vs PII/300's and get 1.41 every >time I try, using crafty. The first benchmark data I got on a Xeon (source I >can't reveal) was exactly 2.25X faster than what I am getting on my P6/200. >This makes sense as crafty has no MMX code whatsoever, so that both processors >are using the same core technology and relative cache speeds. But note that I >am a real 32-bit application here with no known-to-be-bad stuff tucked away to >hurt performance. > >For comparison, the AMD K6 seems a perfect match for the P6/200 when the clocks >are matched... But the Xeon is clocked faster.. I may not understand the data I have seen correctly( it would not be the first time :)). You can see what I have seen at http://www2.tomshardware.com/xeon.html There is also a very good paper by Intel you can download for anyone who want to know more about multi-processing on the Xeon chips at the above web site.
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.