stephenbrooks.orgForumMuon1GeneralGHz Benchmarking
Username: Password:
Search site:
Subscribe to thread via RSS
Stephen Brooks
2002-07-09 08:31:33
I was thinking it would be interesting to have a project-average-GHz report on the stats page like there was for the version 3.x project.  However I need a few people to be able to tell me how many results their v4.1x programs get in a certain period of time.

The three figures I need are R (number of results calculated), T (time taken to calculate them) and S (speed of computer in GHz).  It may also be useful if you say whether you've got an Athlon or P4 or whatever.

I'll first calculate a "GHz-seconds per result" figure for each person, which will be X=S*T/(R+0.5) where the 0.5 is because the program might have been part-way through a unit when the run finished, and on average half-way.  Then I'll average all those figures together to get an overall indicator of the number of seconds at 1GHz it takes to do a muon unit.  Alternatively I could evaluate X=sum{S*T}/sum{R+0.5} with sums over all people, which would be a bit more accurate I think. 

Anyway, once I've got that I'll look at the number of results returned in the last (e.g.) 7 days and work out

(results returned in that time period)*X/(seconds in 7 days).

...to give the average speed in GHz.


"As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline"

[This message was edited by Stephen Brooks on 2002-Jul-16 at 12:16.]
Pascal
2002-07-09 09:32:50
Hi Stephen,

on my first computer (Athlon Thunderbird C (1,2 GC/s, 256 MB DDR-RAM, Via Apollo 266 Pro)the client version 4.1 needed 16 minutes for one work unit (i.e. 20936 particles).

On my second PC (Intel Pentium 3, 600 MC/s, 256 MB SDRAM, Intel BX Chipset) the client - version 3.11 - needed 27 minutes for one work unit (i.e. 20936 particles).

This week I'll have a look at version 4.12 on my PCs.  Perhaps later or tomorrow I'll tell you more.

___________________________
1: Athlon TB-C, 1.2 GC/s, 256 MB DDR-RAM, Erazor x², ADSL-Flatrate, NIC Intel, Win 98 SE Mainboard MSI-6380 Rev.  1
2: Pentium III, 600 MC/s, 256 MB RAM, NIC Intel, Win 98 SE
Frank
2002-07-09 11:23:49
Hi Stephen,

on my computer (Dual Athlon MP (1,6 GHz (1900+), 1024 MB DDR-RAM, ASUS A7M266) the client version 4.1 needed app.11 minutes for two work units .
(2xMuon1_background.exe )


There is something wrong with my results:

13 Frank von Siebenthal 6538 Units 93`814`359 Particles?  136'879'568 is correct.

Perhaps you can correct this.

Greetings from Switzerland

My System:
DualAthlon MP 1900+, 1024 MB DDR-RAM, Gforce3 Ti500, ADSL-Flatrate, Win XP Pro, Mainboard ASUS A7M266D, Promise Raid 280GB (CPU's & GPU Watercooled cool)
Pascal
2002-07-09 11:34:03
Hi Frank,

where ecatly do you come from?  Soon I want to start a job in Switzerland.  Tell me your location and perhaps your email address.  Thx.

___________________________
1: Athlon TB-C, 1.2 GC/s, 256 MB DDR-RAM, Erazor x², ADSL-Flatrate, NIC Intel, Win 98 SE Mainboard MSI-6380 Rev.  1
2: Pentium III, 600 MC/s, 256 MB RAM, NIC Intel, Win 98 SE
gogomaus
2002-07-09 13:32:48
I´m using the famous ALDI-PC edition autumn 2000.
Intel P-III 900 MHz; 128 MB SDRAM; OS WinMe.
Overnight measurements of muon1v411 background.exe lead to 28 minutes per result run.  (v412 seems to operate at same rate since yesterday).
Looks like I have chances for getting the "slow moving award"... mad

As I stop muon-calculations while manualsending, enjoying PC-games or during longer absences, calculations from last 7 days results coming in, will lead to lower "effective" speed.
Stephen Brooks
2002-07-09 14:25:33
I've removed the rot from Frank's results so they should be OK next tim I upload the stats.

gogomaus, can you tell me the total number of units you did overnight and the total amount of time?  That way I get a more accurate figure.

Current figures are about 1100 GC/result (GC=gigacycles, as per Pascal's sig).  gogomaus is an outlier at 1512 GC/result whereas Pascal's 2nd machine got the lowest with 972 GC/result.  I'd like more information on overnight runs though - i.e. total time and total number of results.


"As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline"
kruemi [Swissteam.NET]
2002-07-10 01:05:16
quote:
Originally posted by Frank:

Greetings from Switzerland



Wow, one more from switzerland... But arn't you in the swissteam?
Just visit http://www.swissteam.net and join us. 

bye

kruemi

--
Don't read this.  And neither this!
masel
2002-07-10 09:21:17
On my PC i need 13-14 min per Workunit(20936 Particles).
PC: XP 1600+ (1.4 GHz), Win 98SE, 512 MB DDR-Ram, K7S5a
gogomaus
2002-07-10 14:52:31
I did those calculations repeatedly in the past, but unfortunately that sheet with my notes has been abused by my daughter meanwhile; she definitely needed it much more than me...
Depending on parameter sets I had minimum record at 14 minutes and maximum of 34 minutes (if my memory is working o.k.)
But I have a perfect protocol of last night : Starting at 00:10 (my PC clock) background-v412 ran for 37 unit (time stamp of explorer, so no 0.5 units have to be added) at 18:31. My poor brain would calculate 29.5 minutes per unit.
I just manualsent them in.
Those were computed without any other task running on the PC, config set to L.
I will try another day with priority Normal/High.
kruemi [Swissteam.NET]
2002-07-10 23:11:35
Ok: using the timestams and the real-calculating-time (from the taskmanager) my muon1_background made 14 Units in 13h50min (timestamp of file, no 0.5 units needed to add)
It's a PentiumII 450MHz, about 1 unit per hour

bye

kruemi

--
Don't read this.  And neither this!
Stephen Brooks
2002-07-11 09:14:14
A pattern is emerging

I've just realised that Pascal's 600MHz machine figure was quoted with v3.11, which is substantially faster than 4.1x because it is calculating a different thing.  With that removed, the Athlon results in gigacycles/unit are

Pascal's 1st machine: 1152
Frank's dual processors: 1056
masel's result: 1134

The remaining valid P-III results for 4.x are

gogomaus run #1: 1512
gogomaus run #2: 1607
kruemi's benchmark: 1601 (actually a P-II)

So the Athlons appear to get about 40% more numerical calculation out of each clockcycle than the pentiums!  (1100 vs. 1550ish Gcycles needed)


"As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline"
gogomaus
2002-07-11 11:59:31
And here is another confirmation from todays measurement.
V412-background with config = N(ormal) crunched 45 units in 20h 36min.
If I´m right, this would mean 27.5 minutes per unit or in your GC-world :
1236*60/45*0.9 = 1483 GC per unit.

Next shopping time I should prefer Athlon processor !?!
pben
2002-07-11 15:42:48
I did 70 trials in 8hours 38 on a Dual Athlon 1.2Ghz, (two copies of the backgound progam).

I am have major problems with Win2k right now, either the new video card, the warm weather, or the phase of the moon.  Good thing I got SuSE 8.0 installed, so I know it is Bill G fault and not my hardware.  confused
Mikeydz
2002-07-15 17:38:34
17 hours, 50 minutes
73 results
Athlon 1.4gh
Pascal
2002-07-18 08:40:53
Client V4.13 background on 1st machine (Athlon TB-C, 1,2 GC/s, FSB 266 GC/s, 256 MB DDR-RAM)

41 results in 630 min

that is 15,4 min per result

___________________________
1: Athlon TB-C, 1.2 GC/s, 256 MB DDR-RAM, Erazor x², ADSL-Flatrate, NIC Intel, Win 98 SE Mainboard MSI-6380 Rev.  1
2: Pentium III, 600 MC/s, 256 MB RAM, NIC Intel, Win 98 SE
HETTATLONGUN
2002-07-19 10:20:27
R (results calculated, 12 hour period): 36.

T (time taken to calculate): 12 hours.

S (speed of processor): 900MHz Athlon.

AMD Athlon 900MHz (200MHz FS (Socket A), 512MB PC133 SDRAM, Windows ME.

I am using version 4.13.

I started the muon_1 background program at 18 July 2002, 21:05 hrs and ended the program at 19 July 2002, 09:05 Hours.  Only application running was muon 1.

A total of 36 results were produced.
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io

Site has had 16975863 accesses.