time to complete one unit

[bailey]

I am running a celeron 2 566 fc-pga at 850mhz, it took 16 hrs to do one unit, is that about what it should be for what I have ?

[Phil B]

Not sure guy. But here is the log on a p2 350 running @392


Date Time Work Unit Name Process Percent Spike Spike Gaussian Gaussian Gaussian
Done Done Time-hr Done Power Score Power Fit Score
9/11/00 4:27:45 AM 16jn00ab.5630.20690.298580.184 10.311 100.00% 170.303 0.629 1.439 9.026 0.159
9/11/00 2:33:05 PM 20jn00aa.1997.28704.134656.155 9.940 100.00% 180.700 0.655 1.663 7.630 0.218



Guess that's about a ten hour avg. Since that did not format so well, I'll just post the times on a P3600 @702-6.504
5.837
6.073
I use a utility called task manager 2000 to make sure unneeded programs are not running in the background.
This is on V2.04
Sorry I can't give more results but I just restarted doing seti.

[Edited by Phil B on 09-11-2000 at 08:00 PM]

[bailey]

it's ok , I posted before I read the other threads, I have downloaded the newer version and started a another unit, it is running in the backgroung right now and seams as fast as the other version was just by itself, my cpu temp went up to 41 deg centgrade so it's a long way from 90 deg which is the max for my cpu, I'll see what this one does.
thanks.
bob

[HAL9000]

I was hitting around 10 to 12 (ver 2.03)hours with my 533 Celeron, screen blank after 5 min, running all the time. On version 2.7 it's been in as little as 6 hours, but sometimes still takes up to 12.

[Jomama]

I average about 10 hours on a P3 450 running all the time and blanking the screen, the MFLOP @ about 40 and eff. < 10
This is when the darn spool32.exe doesnt show up, then it eats 98 % of the CPU time. I have to keep an eye on it. I am experimenting trying to find what is calling it up. Often as now, the file is not loaded after a boot up. Other times it just shows up.
I also use Task2000, it is great. spool32.exe is being called by MSGSRV.exe, whatever the heck that is. I have removed anything that may be a message server and have never had a fax program on here. (Because you know, that sending a fax goes to the dame dialog box as a print file.)
Anyway, Seti is a great tool for optimizing your OS. Also it can drive you crazy. http://www.ngi.it/forum/smiles/rolling_eyes.gif

[bailey]

ok, big improvement, 4hr 53 min with screen blank and 2.7

[Jomama]

4 hours! Bailey you are the man! I hope you are in our group. (Alf hunters of PC Mech).
Are you using WIN98?
Are you using your comp for anythng else or does it run all day long? Ive noticed that IE slows it down
Your CPU cycles per MFLOP and MFLOPs per sec numbers would be very interesting. Do you have SETISPY loaded?

[bailey]

yes I'm in the group, and as far as the rest of your questions, I have no idea of what your talking about, I run seti at bed time and the first two were still running when I left for work in the morning, but did the blank thing when I got home from work and it did one before bed time, I will have another one done before I wake up in the morning.
running a celeron 566 fc-pga,850mhz, asus p3b-f,128 meg.

[bailey]

just dl setispy and installed it will let you know the numbers tomarrow, thanks

[Phil B]

Congrats bailey! I knew that celly was gonna crunch better than that 16 hour deal.

[bailey]

ok heres the numbers.
171.93megaflops/second
4.82 cpu cycles/flop
total time 4.45 hrs

not bad
yes running w98

[Edited by bailey on 09-13-2000 at 07:47 AM]

[Phil B]

Gotta love V2.7. Seti Spy estimating 4.5-4.8 hr per w/u. Will post later (tomorrow) with avg times. This is on a p3@702

V2.7 cut at least thirty min. off the avg. for a stock p2350.
Think I'm gonna change the rest of em over to 2.7 since it looks like a win-win situation.


bailey- no, not bad at all! :) :) :)

[Byte 2.0]

We are all kicking major butt. I am going to assume that everyone has seen the results of the Duron 600 I posted...I figure it is going to do about 3.5 a day...I was some impressed with that I may save the money and buy a Duron instead of the T-Bird....However I am going to try to test a T-Bird before I buy Either.... :D

[bailey]

things are realy working great now, the first two I did realy blew my adverage away, but soon will be catching up, just finisheing no. 25 still adveraging 4.5 per.
the group listing still showing 18 for me.

hey, I'm moving up the list fast now.

[Edited by bailey on 09-20-2000 at 12:58 AM]

[Toaster]

um...this was fun but I wonder if its accurate.

871.93megaflops/second/CPU
37.92 cpu cycles/flop
total time 0.42 hrs

Thats um..Linux 6.2 SMP
Dual PIII-700E @ 1010 ea.
256Mb memory at cas2 (pricey memory!)
AMI megaraid set to RAID 1+0.
6 Micropolis "stinger" fast/ultra/wide 9.22GB
64MB of RAID cache.
and other fun stuff.

Fast muddah humpah, but I doubt its accuracy and the folks at SETI simply said "Um...wow...cant be right"

[Byte 2.0]

Toaster,

That is possible, for a couple of reason. One Seti spy at first reports very high Mflops, since the seti client is just starting and does not have alot of data to estimate the total time and is always re-estimating the total times.

Plus those speeds don't seem outragous with your Equipment.
I seems like the kind of person that has about the Best Ram and MB in your system, probably on a great BX Chipset.

Example, I have not tried myself, I will, but the T-Bird 800 MHZ gets about 400 Mflops in windows, so you with Dual 700 P3's (totaling 1400 MHZ) on Linux could very well be reporting 871.93megaflops. Most likely would be slower in WinDoze.

Also the Windows Verison's of Seti is not Mulit-Thread, so on a Dual system in WinNT or WIN2K your only real benifit is Running 2 Seti Clients at once, or Running WinNT on Processor 0 and One Seti Client totally on Processor 1.
(I think that is right)

I am not an expert on linux, I can barely add/remove users. However I understand that Linux runs seti faster, plus you are mostly running a text verison Rigth, that is muilt-Thread?

However those are great numbers to have at the start or thought out a Unit, did you complete that Unit and is so how long did it take...

From reading your other posts you know alot about linux and have no doubt tweaked it down to run as fast as Possible (AFAP).

I know in that a Duron 600 on a cheap MB with cheap no name ram can gets about 175 Mflops and complete a work unit in 5 to 6 hours. Once it did it in 4.5 hours.

I will report on the T-Bird 800 after I get it built and Tweaked down in Win98.

[Toaster]

I re-ran the sim, this time distributing the SETI program umong CPU`s. What SETI will NOT let me do is calculate and then combine results when run as such. On the second try, SETI completed in approx. 33min after some tweaking.
I`m going to try the other "distributed power" work handler to see if results are a bit better. The SETI program has a number of "bells and whistles" that eat CPU cycles but dont contribute to the running calculation.
SETI is quite allergic to BeoWulf and crashes often enough to avoid running it in that type of configuration.
SETI does have plans to support "parrallel tasking" systems in the near future. Its hard to say which is faster, Linux or Windoze but most apps are cached in Linux and second and subsequent launching execute far faster. Netscape for example loads in about 4-5 seconds and the 2nd or third time loads in less then 2 seconds. This is how I bested my previous numbers by "caching" the program and executing it direct from memory.

[bailey]

12 days and 40 work groups completed, moving up the ladder on user stats, but I think toaster is going pass us all to the top.

[Byte 2.0]

Last time I ask Toaster he was part of another group, and as far as I know he isn't in the PCMECH group at this time.

However he is always welcome to join.

My goal for myself is to reach 1000 work units. I want a chance to be a crucher of the Week....

I would like to see Toaster in the group, truth be known I bet he has more Work units then the top 3 of us together...

[Toaster]

For the time being, I am not part of any "group".
My "BeoWulf" project is sold to universities on a time basis and I use SETI for functional testing. The folks at SETI informed me that my equipment may not comply to their requirements and that all computations may be erronous.
They do accept completed work units from me but do not elude to their usefulness. When I add/change/modify any clusters or software, I use SETI to simply "burn in" and work the system. While I say they "buy" time, I mean that if the project they wish to compute is something "futile" I impose a charge. If they prove usefulness and allow the results and the project to be distributed freely, I do not charge.
I have recently made numorous changes and additions and thus was my SETI times. Soon, 2 add`l systems will be brought on line bringing the total to 13 systems working in parallel on a given task.
A university in Ill. is considering an alike system for their sciences research. The systems will communicate and share the task at hand. More news as it develops.
So...the bottom line is simple, I don`t wish to be part of a group for the immediate future because the state of my system(s) is constantly changing as I become more fluent in parrallel tasking system architecture.

[bailey]

100 wu in just 30 days
not too bad

[bailey]

I have two similear systems running and have a great deal of differance in time, I discovered that the faster one is running with a resolution of 1.5 to .3hz, while the slower one is running at .075hz. the faster one takes 6hr and the slower one will take over 8.5hr. I even had one so fast it completed one work unit in 23 minutes.
so I guess all work units are not equal or the same.