Author: Matt Taylor
Date: 18:41:22 04/22/03
Go up one level in this thread
On April 22, 2003 at 20:41:22, Aaron Gordon wrote: >On April 22, 2003 at 15:40:47, Tom Kerrigan wrote: > >>On April 21, 2003 at 19:41:22, Aaron Gordon wrote: >> >>>On April 21, 2003 at 17:41:17, Tom Kerrigan wrote: >>> >>>>On April 21, 2003 at 14:46:27, Aaron Gordon wrote: >>>> >>>>>>Not really. Processors have their strengths and weaknesses. I'm sure you could >>>>>>find some pathological code that runs much slower on the Athlon than the P4. >>>>> >>>>>If you find anything, let me know. So far I know of nothing. I however know >>>>>of *MANY* things that run pitiful on a p4. >>>> >>>>IIRC, a game of life thing I wrote a few months ago ran 2-3x slower on my AXP >>>>2000+ than on my 2.2GHz P4. Also I think the P4 has won pretty much every media >>>>encoding benchmark I've seen, often by large margins. >>> >>>Thats also bandwidth intensive, and it just so happens most pages run the bios >>>timings at the absolute lowest setting making the Athlons bandwidth horrible. >>>Pick ANY multimedia program, grab a P4 and lets benchmark against my system. >>>I've already done the benchmarks, as I said in a previous email.. properly >>>configured Athlons (non-gimped bioses) will beat a P4 any day. Any time you'd >>>like me to back it up let me know. :) >> >>Well, just to eliminate any sort of bias, let's go to SPECint-- >> >>gzip - P4 faster >>vpr - about even >>gcc - P4 significantly faster >>mcf - P4 WAY faster >>crafty - AXP faster >>parser - P4 faster >>eon - about even >>perlbmk - about even >>gap - P4 a lot faster >>vortex - about even >>bzip2 - about even >>twolf - about even >> >>I don't see the P4 doing horribly on any of these tests and it does quite well >>on several. >> >>-Tom > >I ran the benchmarks, here you go. > >P4-2.53GHz - Linux >File used for Testing: test.wav, 272,692,456 bytes > >Bzip2: bzip2 -9k test.wav >user 4m31.960s > > >Gzip: gzip -9 test.wav >user 1m45.430s > > >GCC (Crafty 19.1 compile time): make >user 0m21.840s > >Lame MP3 encoder: ./notlame -q 0 -V 0 --new-vbr test.wav bench.mp3 >2 minutes, 27 seconds. > >-------------------------------------------------------- >XP-2.5GHz - Linux >File used for Testing: test.wav, 272,692,456 bytes > >Bzip2: bzip2 -9k test.wav >2m8.690s > >Gzip: gzip -9 test.wav >0m22.770s > >GCC (Crafty 19.1 compile time): make >0m15.290s > >Lame MP3 encoder: ./notlame -q 0 -V 0 --new-vbr test.wav bench.mp3 >1 minute, 19 seconds. >-------------------------------------------------------- > >Now, to put these results in perspective... > > >In GCC it would take a P4-3618.09MHz to equal the XP-2507MHz. >In GCC it would take a XP-1755.12MHz to equal the P4-2533MHz. > >In Gzip it would take a P4-11728.33MHz to equal the XP-2507MHz. >In Gzip it would take a XP-539.77MHz to equal the P4-2533MHz. > >In Bzip2 it would take a P4-5445.94MHz to equal the XP-2507MHz. >In Bzip2 it would take a XP-1166.04MHz to equal the P4-2533MHz. > >In lame it would take a P4-4713.30MHz to equal the XP-2507MHz. >In lame it would take a XP-1347.29MHz to equal the P4-2533MHz. > >----------------------------------------------------------------- > >I'm sure you're thinking, "What the %(&#(*". This is why if you'd like, Tom, >you can telnet into my Athlon XP machine and run the exact same tests (or >any other tests you wish to run), as well on the Pentium 4 2533MHz machine. >You'll need to email Matt Taylor at para@tampabay.rr.com for account >information. Email me at agordon@newageoc.com or speedycpu@attbi.com and I'll >set you up with an account. I also have an AthlonXP 2500 with a very similar configuration. The only difference aside from on-board gizmos and the P4's video card is ram. The AthlonXP 2500 has 248 MB of ram, and the P4 2.53 GHz has 512 MB. The P4 uses an IBM 60GXP 60GB 7200 rpm disk, the kind that like to die every couple months. The AthlonXP runs off of a WD hdd, but it has the same model IBM disk in it as well. Both machines use factory settings. The last important difference is that the P4 sits idle 24/7 whereas the AthlonXP 2500 routes network traffic and hosts servers. As Aaron stated, I wouldn't mind letting people log in to bench either machine. You're welcome to find pathological code for Athlon. The P4 stalls on some of the simple ALU ops that the Athlon can execute at a rate of 3/cycle. (Read: bswap, adc/sbb, push, inc/dec, shift/rotate, setcc.) More complex instructions are worse. The mul/imul instructions have an execution latency of 5 cycles on Athlon, and the result of imul r32, r32 (most common) is available the next cycle. On P4 it is more like 5 cycles later. The P4 does execute the most basic 6 ALU ops (add, sub, and, or, xor, not, mov) very quickly, however. If you compare the absolute instruction execution latencies on P4 & Athlon, the P4 only wins for the basic ALU ops since it runs at a higher clock speed. There are only so many of these that you can issue before you have to branch. Athlon also has a much bigger I-cache, and it has better decoders. If we want to compare apples to apples, we might as well talk about Opteron since reported performance figures are about 33% higher than Athlon in 32-bit code on a 32-bit OS making Opteron roughly twice as fast as Xeon clock-for-clock. Looky looky, some SPEC benchmarks: http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_8796_8800~69690,00.html http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_8796_8800~69691,00.html As best I can tell, the 144 is 1.8 GHz (roughly 3.6 GHz P4). More benchmarks: http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_8796_8800,00.html -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.