stephenbrooks.orgForumMuon1GeneralTrying to Make a CPU performance Guide - Stephen help please?
Username: Password:
Search site:
Subscribe to thread via RSS
[TA]Overkiller
2004-01-30 05:50:33
Hello everyone reading this,

I am a member of team anandtech and due to a recent surge of participants i thought it would be nice to try and get a better understanding as to the relationship between CPU mhz & DPAD (also how an AMD system fairs as compared to an INtel one)....

here is the current discussion in our forums and our methods...

http://forums.anandtech.com/messageview.cfm?catid=39&threadid=1236189

Of course..if anyone has done this already, has the answers, is interested in helping us out, or simply sees a better way to go about this...let me know Smile

thanks!
-over
Stephen Brooks
2004-01-30 09:33:41
I've seen that you've found this thread, which is something I wrote for that purpose.

KB of results is a lot less reliable than Mpts, as the latter was designed to be a quantitative measure of how much calculation got done (including those rechecked results) - rather like FLOPS.  One thing you should bear in mind is that 4.33a has been optimised over 4.33, perhaps running 10-15% more efficiently, so be careful to figure out who is running the optimised one and who isn't.

I tend to use kpts/sec as a measure of rate for a single computer (1 Mpts = 1000 kpts).  See that the dark green text on this picture:

...quotes Mpts/sec for the whole project.

Right now I am using the "exchange rate" of 60 kpts/sec to 1 GHz.  Your computer's per-clock-cycle efficiency is measured in kpts/sec/GHz in that way, and the 60 I use was got from an average between Intel and AMD systems (it was like 45 for intel and 75 for AMD, something of that nature).  However, for 4.33a and above you will be getting more kpts/sec for the same GHz, so this conversion factor will go up, so the current 707GHz the green text is reading should really be scaled down a bit (it's equivalent to 707GHz running on the pre-optimised client, so really more like 500GHz on post-4.33a ones).

HB Pencils, also sold as "Moron's Choice" Graphite Cigars.
Stephen Brooks
2004-01-30 09:38:10
The more mathematical of you may notice that the the unit kpts/sec/GHz is equivalent to "particle-timesteps per million CPU cycles".

X Mpts/day = X/86.4 kpts/sec
X Mpts/hour = X/3.6 kpts/sec

HB Pencils, also sold as "Moron's Choice" Graphite Cigars.

[edit] I just had a look at the overall speed of the project and we're going steadily around 21`000 results per day (excluding one or two large dumps above that).  That's one every 4 seconds, a respectible rate which I think is going to come in handy when I try to crank up the number of parameters again (137 to 767) by adding a linear accelerator (v4.4).  The speed increase has been due to a combination of: the existing teams' diligence in recruiting new members; new teams such as Team China signing up; and the 20% or so boost the v4.33a optimisations gave.

[This message was edited by Stephen Brooks on 2004-Feb-03 at 16:06.]
Stephen Brooks
2004-02-03 08:10:52
Oops.  Oh yeah, I forgot to say you should turn the auto-download/merge option off in config.txt while you're running that benchmark program, or a load of results form everyone else will suddenly arrive in your file Eek

HB Pencils, also sold as "Moron's Choice" Graphite Cigars.
[TA]Overkiller
2004-02-03 23:06:56
Thanks for your great response stephen, we are continuing to look at our data and try to get some sort of efficiency guide up in the near future Smile

thanks for your contiinued support!
[TA]Overkiller
2004-02-04 15:52:53
Hey Stephen....weird results w/ your muon1bench program...

I hhave gotten some very weird muon bench scores on my P4C 2.8@3.2 [first 7 entries are when i was gaming]

Help?

Muon1Bench started.  Interval = 300.0 sec
Call 'muon1bench 900' or similar from DOS to change the interval.

uptime+0 Mpts+0.0 No estimate so far
uptime+2101 Mpts+303.3 Estimate 144.33 kpts/sec
uptime+2401 Mpts+316.1 Estimate 131.62 kpts/sec
uptime+3302 Mpts+480.0 Estimate 145.35 kpts/sec
uptime+5403 Mpts+783.1 Estimate 144.92 kpts/sec
uptime+7505 Mpts+1116.0 Estimate 148.70 kpts/sec
uptime+8405 Mpts+1389.5 Estimate 165.30 kpts/sec
uptime+17111 Mpts+2927.9 Estimate 171.10 kpts/sec
uptime+18612 Mpts+3224.7 Estimate 173.25 kpts/sec
uptime+20714 Mpts+3523.7 Estimate 170.11 kpts/sec
uptime+22815 Mpts+3827.7 Estimate 167.77 kpts/sec
uptime+24616 Mpts+4140.0 Estimate 168.18 kpts/sec
uptime+25817 Mpts+4444.0 Estimate 172.13 kpts/sec
uptime+26718 Mpts+4749.1 Estimate 177.75 kpts/sec
uptime+27919 Mpts+5049.8 Estimate 180.87 kpts/sec
uptime+28219 Mpts+5208.0 Estimate 184.55 kpts/sec
uptime+29420 Mpts+5526.6 Estimate 187.85 kpts/sec
uptime+30320 Mpts+5802.1 Estimate 191.36 kpts/sec
uptime+31221 Mpts+6092.2 Estimate 195.13 kpts/sec
uptime+32422 Mpts+6402.4 Estimate 197.47 kpts/sec
uptime+33322 Mpts+6705.6 Estimate 201.23 kpts/sec
uptime+34223 Mpts+7007.9 Estimate 204.77 kpts/sec
uptime+35424 Mpts+59055.6 Estimate 1667.10 kpts/sec
uptime+36324 Mpts+59359.1 Estimate 1634.12 kpts/sec
uptime+36624 Mpts+59382.2 Estimate 1621.36 kpts/sec
uptime+37525 Mpts+59684.4 Estimate 1590.50 kpts/sec
uptime+38426 Mpts+59987.1 Estimate 1561.10 kpts/sec
uptime+39326 Mpts+60226.6 Estimate 1531.44 kpts/sec
uptime+40227 Mpts+60519.9 Estimate 1504.45 kpts/sec
uptime+41428 Mpts+60822.4 Estimate 1468.14 kpts/sec
uptime+42328 Mpts+61152.0 Estimate 1444.69 kpts/sec
uptime+42628 Mpts+61166.4 Estimate 1434.86 kpts/sec
uptime+43529 Mpts+61501.5 Estimate 1412.87 kpts/sec
uptime+44730 Mpts+61814.0 Estimate 1381.93 kpts/sec
uptime+45630 Mpts+62092.4 Estimate 1360.75 kpts/sec
uptime+46531 Mpts+62335.0 Estimate 1339.63 kpts/sec
uptime+47131 Mpts+62538.4 Estimate 1326.88 kpts/sec
uptime+48032 Mpts+62775.0 Estimate 1306.93 kpts/sec
uptime+53135 Mpts+64288.3 Estimate 1209.88 kpts/sec
uptime+54036 Mpts+64591.3 Estimate 1195.33 kpts/sec
uptime+54336 Mpts+64647.5 Estimate 1189.76 kpts/sec
uptime+55237 Mpts+64960.4 Estimate 1176.02 kpts/sec
uptime+56438 Mpts+65266.4 Estimate 1156.42 kpts/sec
uptime+57338 Mpts+65574.1 Estimate 1143.63 kpts/sec
uptime+57939 Mpts+65723.6 Estimate 1134.35 kpts/sec
uptime+58839 Mpts+66025.6 Estimate 1122.13 kpts/sec
uptime+59740 Mpts+66330.7 Estimate 1110.32 kpts/sec
uptime+60040 Mpts+66351.2 Estimate 1105.11 kpts/sec
uptime+60340 Mpts+66487.5 Estimate 1101.87 kpts/sec
uptime+61541 Mpts+66804.1 Estimate 1085.51 kpts/sec
uptime+62442 Mpts+67108.9 Estimate 1074.74 kpts/sec
uptime+63643 Mpts+67411.6 Estimate 1059.21 kpts/sec
uptime+64543 Mpts+67714.1 Estimate 1049.12 kpts/sec
uptime+65444 Mpts+68019.3 Estimate 1039.35 kpts/sec
uptime+66344 Mpts+68263.9 Estimate 1028.93 kpts/sec
uptime+67545 Mpts+68565.8 Estimate 1015.10 kpts/sec
uptime+68446 Mpts+68878.3 Estimate 1006.31 kpts/sec
uptime+69346 Mpts+69180.2 Estimate 997.60 kpts/sec
uptime+70247 Mpts+69401.2 Estimate 987.95 kpts/sec
uptime+71148 Mpts+69705.8 Estimate 979.73 kpts/sec
uptime+71748 Mpts+69860.9 Estimate 973.69 kpts/sec
uptime+72949 Mpts+70180.5 Estimate 962.05 kpts/sec
uptime+73849 Mpts+70483.9 Estimate 954.42 kpts/sec
uptime+74750 Mpts+70751.4 Estimate 946.50 kpts/sec

And this for my Athy 1700+ @ 2.1 Ghz


Muon1Bench started.  Interval = 300.0 sec
Call 'muon1bench 900' or similar from DOS to change the interval.

uptime+0 Mpts+0.0 No estimate so far
uptime+1200 Mpts+303.1 Estimate 252.49 kpts/sec
uptime+2701 Mpts+605.2 Estimate 224.06 kpts/sec
uptime+3901 Mpts+906.4 Estimate 232.32 kpts/sec
uptime+4801 Mpts+1063.7 Estimate 221.52 kpts/sec
uptime+6302 Mpts+1368.5 Estimate 217.14 kpts/sec
uptime+7802 Mpts+1661.6 Estimate 212.95 kpts/sec
uptime+9003 Mpts+1963.8 Estimate 218.12 kpts/sec
uptime+9903 Mpts+2224.1 Estimate 224.57 kpts/sec
uptime+11404 Mpts+2528.4 Estimate 221.71 kpts/sec
uptime+12604 Mpts+2851.4 Estimate 226.22 kpts/sec
uptime+13805 Mpts+3091.2 Estimate 223.92 kpts/sec
uptime+15905 Mpts+3225.6 Estimate 202.79 kpts/sec
uptime+17106 Mpts+3520.1 Estimate 205.78 kpts/sec
uptime+18606 Mpts+3823.3 Estimate 205.48 kpts/sec
uptime+19807 Mpts+4125.5 Estimate 208.28 kpts/sec
uptime+20407 Mpts+4271.2 Estimate 209.29 kpts/sec
uptime+21908 Mpts+4576.2 Estimate 208.88 kpts/sec
uptime+23108 Mpts+4905.9 Estimate 212.30 kpts/sec
uptime+24308 Mpts+5206.8 Estimate 214.19 kpts/sec
uptime+25509 Mpts+5503.0 Estimate 215.72 kpts/sec
uptime+25809 Mpts+5512.6 Estimate 213.59 kpts/sec
uptime+27009 Mpts+5829.7 Estimate 215.84 kpts/sec
uptime+28210 Mpts+6135.2 Estimate 217.48 kpts/sec
uptime+29410 Mpts+6437.2 Estimate 218.87 kpts/sec
uptime+30911 Mpts+6741.1 Estimate 218.08 kpts/sec
uptime+32111 Mpts+7045.3 Estimate 219.40 kpts/sec
uptime+33012 Mpts+7260.0 Estimate 219.92 kpts/sec
uptime+33912 Mpts+71208.8 Estimate 2099.78 kpts/sec
uptime+35112 Mpts+71512.5 Estimate 2036.65 kpts/sec
uptime+36613 Mpts+71817.1 Estimate 1961.50 kpts/sec
uptime+37213 Mpts+71986.0 Estimate 1934.40 kpts/sec
uptime+38414 Mpts+72289.8 Estimate 1881.86 kpts/sec
uptime+39014 Mpts+72437.5 Estimate 1856.69 kpts/sec
uptime+40514 Mpts+72743.2 Estimate 1795.47 kpts/sec
uptime+41415 Mpts+72970.2 Estimate 1761.92 kpts/sec
uptime+42615 Mpts+73274.8 Estimate 1719.44 kpts/sec
uptime+43816 Mpts+73593.8 Estimate 1679.61 kpts/sec
uptime+45016 Mpts+73890.4 Estimate 1641.41 kpts/sec
uptime+45316 Mpts+73905.2 Estimate 1630.86 kpts/sec
uptime+46517 Mpts+74218.2 Estimate 1595.51 kpts/sec
uptime+47717 Mpts+74537.0 Estimate 1562.05 kpts/sec
uptime+49218 Mpts+74840.0 Estimate 1520.58 kpts/sec
uptime+50418 Mpts+75143.0 Estimate 1490.39 kpts/sec
uptime+51618 Mpts+75444.7 Estimate 1461.57 kpts/sec
uptime+52819 Mpts+75758.4 Estimate 1434.29 kpts/sec
uptime+54319 Mpts+76062.4 Estimate 1400.27 kpts/sec
uptime+55520 Mpts+76366.3 Estimate 1375.47 kpts/sec
uptime+56720 Mpts+76656.7 Estimate 1351.47 kpts/sec
uptime+57921 Mpts+76961.1 Estimate 1328.72 kpts/sec
uptime+59121 Mpts+77264.8 Estimate 1306.88 kpts/sec
uptime+60622 Mpts+77571.5 Estimate 1279.59 kpts/sec
uptime+61822 Mpts+77873.1 Estimate 1259.62 kpts/sec
uptime+63023 Mpts+78177.2 Estimate 1240.45 kpts/sec
uptime+64223 Mpts+78491.5 Estimate 1222.16 kpts/sec
uptime+64523 Mpts+78509.7 Estimate 1216.76 kpts/sec
uptime+65724 Mpts+78811.9 Estimate 1199.13 kpts/sec
uptime+66024 Mpts+78919.0 Estimate 1195.30 kpts/sec
uptime+67224 Mpts+79222.1 Estimate 1178.47 kpts/sec
uptime+68725 Mpts+79522.0 Estimate 1157.10 kpts/sec
uptime+69925 Mpts+79828.6 Estimate 1141.62 kpts/sec
uptime+71126 Mpts+80130.9 Estimate 1126.60 kpts/sec
uptime+72326 Mpts+80384.8 Estimate 1111.42 kpts/sec
uptime+73526 Mpts+80679.8 Estimate 1097.28 kpts/sec
uptime+74727 Mpts+80967.3 Estimate 1083.50 kpts/sec
uptime+75927 Mpts+81271.8 Estimate 1070.38 kpts/sec
uptime+77428 Mpts+81624.9 Estimate 1054.20 kpts/sec
uptime+78628 Mpts+81904.7 Estimate 1041.66 kpts/sec
uptime+80129 Mpts+82193.0 Estimate 1025.75 kpts/sec
uptime+80729 Mpts+82352.0 Estimate 1020.10 kpts/sec
uptime+81329 No change in file size

odd?
Stephen Brooks
2004-02-05 00:45:47
What did I say two posts ago?  Roll Eyes
quote:
Originally posted by Stephen Brooks:
Oops.  Oh yeah, I forgot to say you should turn the auto-download/merge option off in config.txt while you're running that benchmark program, or a load of results form everyone else will suddenly arrive in your file Eek


HB Pencils, also sold as "Moron's Choice" Graphite Cigars.
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io

Site has had 17135048 accesses.