View Full Version : Fastest seti@murcers June 6-13

13th June 2000, 06:26
<TABLE BORDER=1><TR><TH>name</TH><TH>wus</TH><TH>hours</TH><TH>average</TH></TR><TR><TD>Rags</TD><TD>33</TD><TD>74.48</TD><TD>2h 15m</TD></TR><TR><TD>Manex</TD><TD>101</TD><TD>300.90</TD><TD>2h 59m</TD></TR><TR><TD>cjolley</TD><TD>7</TD><TD>35.89</TD><TD>5h 8m</TD></TR><TR><TD>jms</TD><TD>51</TD><TD>262.98</TD><TD>5h 9m</TD></TR><TR><TD>sleepingdragon</TD><TD>6</TD><TD>32.45</TD><TD>5h 24m</TD></TR><TR><TD>Eye MD</TD><TD>27</TD><TD>151.57</TD><TD>5h 37m</TD></TR><TR><TD>Mark F.</TD><TD>107</TD><TD>613.62</TD><TD>5h 44m</TD></TR><TR><TD>KvHagedorn</TD><TD>67</TD><TD>387.29</TD><TD>5h 47m</TD></TR><TR><TD>giskard</TD><TD>61</TD><TD>353.45</TD><TD>5h 48m</TD></TR><TR><TD>Jorden</TD><TD>11</TD><TD>65.16</TD><TD>5h 55m</TD></TR><TR><TD>Ackerot</TD><TD>10</TD><TD>60.32</TD><TD>6h 2m</TD></TR><TR><TD>Guyver</TD><TD>42</TD><TD>262.98</TD><TD>6h 16m</TD></TR><TR><TD>cancer</TD><TD>2</TD><TD>12.60</TD><TD>6h 18m</TD></TR><TR><TD>CHHAS</TD><TD>23</TD><TD>145.55</TD><TD>6h 20m</TD></TR><TR><TD>Jakob Kruse</TD><TD>64</TD><TD>415.27</TD><TD>6h 29m</TD></TR><TR><TD>VSA</TD><TD>66</TD><TD>438.30</TD><TD>6h 38m</TD></TR><TR><TD>Kosh</TD><TD>4</TD><TD>27.25</TD><TD>6h 49m</TD></TR><TR><TD>paulcs</TD><TD>90</TD><TD>613.62</TD><TD>6h 49m</TD></TR><TR><TD>Kruzin</TD><TD>2</TD><TD>13.65</TD><TD>6h 50m</TD></TR><TR><TD>Buuri</TD><TD>24</TD><TD>164.90</TD><TD>6h 52m</TD></TR></TABLE>


13th June 2000, 08:02
Sorry about the 7 WUs. http://forums.murc.ws/ubb/frown.gif
swbell started enforcing it's 1 ip address rule and totaly hosed my home network.
Linksys router to the rescue!
It should be here tomorrow.

PS How do Rags &/or Manex get those amazing times consistently?
Is there a way to split up a WU among processors?

[This message has been edited by cjolley (edited 13 June 2000).]

13th June 2000, 09:00
I don't think it is possible to split WUs...
Rags and Manex have some very fast machines. Simple as that.

BTW. Has anyone checked out the new SETI stats(not Martins)? pretty cool.


13th June 2000, 09:13
It can't be that simple.
My machine is a P3-650@923 with 256 megs of ram.
All else being equal, that would put thier cpus at something like 2000mz http://forums.murc.ws/ubb/biggrin.gif
So, I wonder what the "else" is.
eg, Linux-vs-Win98, more memory(!?), cached WUs, Alpha-vs-P3 etc.

13th June 2000, 09:26
Well, I did a search and found.... http://forums.murc.ws/ubb/Forum12/HTML/000034.html

great big ol' huge can o' worms http://forums.murc.ws/ubb/frown.gif

Sorry about that Rags....

I'm still curious though,

13th June 2000, 09:41
Chuck, actually in my case, it is just that simple http://forums.murc.ws/ubb/smile.gif. I crunch WUs on a 4-way HP L2000 compute server using the 2.4 command line client compiled for HP-UX. I'm not sure exactly how fast the processors are in MHz, but they have rather large L2 caches, 1MB at least per processor I think. I've heard that a large L2 helps with getting the time down. I can finish a WU in just under 3 hours if no other processes are running on the server.

As for Rags, I think he has his rig pretty much tweaked to the max, so his results aren't typical by any means. I can't touch his stats using my home rig either...



13th June 2000, 10:14
That's how I run my WU's in 6 to 7 hours on my normal P3-450 with 256Mb RAM. Just run Windows, and minimize the 1.06 GUI, let it run in the system tray. And let it be about the only thing to run on your system.

I've had a couple of 5.5 hours and one of just under 5 hours. I'm glad to be back in the top 10 of fastest http://forums.murc.ws/ubb/wink.gif


13th June 2000, 10:42
Hmmmmmm..... interesting
The tax collection system I designed is geting it's own server!
A Compac 8000
4 2meg cache Xeon 550s and 2 gigs of ram http://forums.murc.ws/ubb/biggrin.gif
I'm just sure we need to burn-in that sucker before we trust it with a govenment database!
You guys got any sugestions for a cpu intensive app we could use? http://forums.murc.ws/ubb/wink.gif

13th June 2000, 16:21
WOW! http://www.specbench.org/osg/cpu95/results/res99q4/cpu95-19990927-03869.html

13th June 2000, 22:18
Bwahahaha, cjolley!
I think it would be totally irresponsible of you if you *didn't* burn in that new server with a very computationally intensive app like seti. For at least a week.


13th June 2000, 23:29

You mean the new "personal stats" page? With the 10 last results? Or have you found more new goodies?


14th June 2000, 00:16
just the new personal stats, kinda nice to view your last 10 WUs. But there are beta-stats for gaussians and spikes too. But you probably already knew that. http://forums.murc.ws/ubb/smile.gif


14th June 2000, 00:59
No, i didn't noticed it yet. But it really is cool. Now I can see that one of our 3 Members send a WU with 23 Hrs http://forums.murc.ws/ubb/frown.gif.


14th June 2000, 02:57

And of course you can view more than your last 10 WUs, just change the value for "nresults" in the url...

14th June 2000, 03:39
Hihi, will look my last 100 WU's now!

14th June 2000, 03:44
bah, You can only see your last 30 WU's http://forums.murc.ws/ubb/frown.gif

Mark F
15th June 2000, 15:01
Is it the L2's fault?

Is that why one system with a P3-733 (256k L2, VIA133A) runs about as fast as the one with P3-558 (512k L2, BX). Or is it mostly the better memory throughput of the BX system?

Any ideas! Besides swapping prossesors in the systems. I need the greater CPU in the VIDEO system, than in my GF and daugter's system.

Mark F.

OH NO, my retractable cup holder swallowed a DVD...
and burped out a movie

15th June 2000, 20:27
I get the impression that memory system throughput is very important, cache & all.
I am running a 142mz front-side-bus and have the fastest time of a normal (:P) PC.
I would be interesting to see the proccessor & FSB of the top systems.

ABit BF6, P3-650@923, 256mb@142cas3, 10gb IBM@7200, SB Live Value@3.0, Pioneer 104s DVD, Mitsumi CDRW@2x2x8, Acatel 1000 ADSL@1.5mb/sec, Linksys EtherFast NIC, LG 995e, USB mouse,Matrox G400 MAX!!!! :D

16th June 2000, 00:15
you should get some CAS2 capable RAM. I read somewhere that you need 20MHz higher FSB to make up for the performance hit CAS3 gives you.
I think Rags mentioned in another thread that he is running at a VERY high FSB(170?), which would make sense.
A really nice benchmark program can be found here: http://www.fi.muni.cz/~xsmid4/


[This message has been edited by jms (edited 16 June 2000).]

16th June 2000, 09:59
I have two very similar setups: a VIA Apollo Pro 133A-based system with a PIII @ 800 MHz and a BX-based system with a PIII @ 733 MHz. In both cases, the RAM is clocked at 133 MHz and the timings are set to 2-2-2.

The BX-based system seems to be beating the VIA system by about 15 minutes. There are work unit related variations, of course, but the BX system does seem to be a bit faster, despite the marginally slower clock speed.

They both appear to be faster than a third system, another BX motherboard and a 600 MHz Katmai, FSB @ spec.

I would think Mark's VIA system should be a lot faster than his BX-based system, simply because the VIA systems's FSB is faster and the CPU is considerably faster.

Mark, I'm using an Asus P3V4X motherboard, which has a few unique quirks, so I can't be sure if this is applicable to your motherboard. I've noticed that memory performance takes a *huge* nosedive if the memory bus is at 100 MHz, or if CL3-rated PC133 RAM is used. Also, and this is just bizzare, the board doesn't seem to like it if you set the timings manually.

Memory peformance is greatly enhanced if you are using CL2-rated PC133 RAM, running on a 133 MHz memory bus, and the timings are set by SPD @ 2-2-2. If I deviate from this configuration, my memory benchmarks go to hell, and I begin sticking pins in my Wenchi Chen voodoo doll.

Again, I'm not sure if this has anything to do with your situation, but I do think your VIA/733 MHz Coppermine system should be noticably faster than your Intel/558 MHz Katmai system.


[This message has been edited by paulcs (edited 16 June 2000).]

Mark F
16th June 2000, 14:51
http://forums.murc.ws/ubb/smile.gif Wenchi Chen voodoo doll http://forums.murc.ws/ubb/wink.gif
havent stoped laghfing yet!!

The VIA is at 133 with memory set to 2-2-2(manually); the best I can get the Tyan board to run. The BX is at 124 with memory at 3-2-2 (PC100 and has been run at 133 3-2-2 mostly stable and 3-3-3 totaly stable).
The BX will do a WU from 5:40-6:12. The VIA runs 5:43-6:04 (based on last 10 WUs from each system). Running pretty cose to gether dispite the BX systems disadvantages; slower memory setting, slower FSB slower CPU. It's only advantage is the larger L2 (and BX chipset, maybe http://forums.murc.ws/ubb/wink.gif).

Oh well, neither is running poorly so I'm not complaining, it's just strange.

If the 815 is even comperable to a BX@133 with the newer features; it bye-bye VIA.

Mark F.

Mark F.

OH NO, my retractable cup holder swallowed a DVD...
and burped out a movie

16th June 2000, 15:44
The benches I've seen have it a little slower than BX-133 and faster than the VIA Apollo Pro 133A and even the i840 (w/RDRAM) on many tests.

Someone should find out who was on the i400BX design team and erect an Internet shrine in their honor. I hope they all got big raises. http://forums.murc.ws/ubb/smile.gif


20th June 2000, 14:08

Flash your P3V4X BIOS to revision 1005. It is noticeably faster than version 1003 running 3DMark 2000 as a benchmark.

20th June 2000, 14:11

So, SETI@home version 1.06 is faster than 2.04?

20th June 2000, 15:37
In my opinion it is, Swing.

But then again, 2.04 (??) calculates lots of other things as well, just look at the extra graphics they put in there, so my 1.06 is going faster, leaving behind data.

And I don't go for the top 1 spot in any league, just for a fast crunch !!

(never leave your crunching on when playing a 3D game !!! It will slow down the crunching process !! I had a WU that I stopped at 25%, after it said 14 hours... I quit playing my Star Trek Armada and let it run on it's own through the night: 6 hours 38 minutes !!)


21st June 2000, 11:16

As to which version is fastest seems to vary between different PC's.On my Cel 550 v2.04 GUI reduced my WU times by 2hrs!:Q.
However a fellow team member of mine noticed no difference on his PII.
Perhaps the v1.06 is faster with bigger cache cpu's & the V2.04 faster with smaller cache cpu's?.Just a guess on my behalf though http://forums.murc.ws/ubb/smile.gif
At the end of the day there is only one way to know for certain ,try it http://forums.murc.ws/ubb/smile.gif

I guess I won't recruit for my team here ;) TA