Posts by Webmaster Yoda

1) Message boards : Number crunching : "Fair" enhanced credits.... A simple solution (Message 306931)
Posted 16 May 2006 by Profile Webmaster Yoda
Post:
Nah, let's do it better: 2 WUs = 1 credit. :)


Since everybody is here for the science, no need for credits at all :D Take them away and see how many volunteers stay...

Won't make any difference to me - I'm only running SETI on one PC now (using the old optimised apps). When it runs out of work, there's other projects I can contribute to and there will be other volunteers who can continue to help this project.

Call me whatever you like - doesn't matter. I'm not here for the science, I happen to like those meaningless credits and I don't like work units that take hours to complete.

But that's just my view, at this point in time. Others will have different views and mine may change too.
2) Message boards : Number crunching : Which Dual Core? (Intel or AMD?) (Message 251449)
Posted 21 Feb 2006 by Profile Webmaster Yoda
Post:
How much faster is the 920 than the 531 in terms out output per day? If I run HT on the 531, will it generate more output?


I can't show you figures, but SETI works best with lots of on-chip memory, especially when you're running Crunch3r's optimised apps. The 920 has 2MB cache on each core, while the 531 only has 1MB that needs to be shared between two processes if running with HT on.

My own P4 550 only does about 3% more work with HT on than it does with HT off - probably due to cache (sharing 1MB between two threads).

If it were up to me, I'd get the 920 (or better) Make sure your motherboard supports the Pentium D - a lot of Pentium 4 motherboards don't (I think you need at least an i945 chipset or equivalent).
3) Message boards : Number crunching : Which Dual Core? (Intel or AMD?) (Message 251308)
Posted 21 Feb 2006 by Profile Webmaster Yoda
Post:
[But I run a Pentium D 830 and it does 2 WU's in 28 minutes.


The Pentium D 930 would probably be even quicker (with Crunch3r's apps), since they have 2x2MB L2 cache (more than the 2x512kb or 2x1MB on the Athlon X2 and Pentium D 8xx)
4) Message boards : Number crunching : DELL 4700 P4 3.4 GHz. problem -- (Message 245901)
Posted 11 Feb 2006 by Profile Webmaster Yoda
Post:
After having to reload the OS after a head crash, the system shows 1 CPU, where it did show 2 prior to the drive failure.


It sounds like HT is disabled. I'm not familiar with Dell computers (I prefer to build my own) but it should be a BIOS setting, usually accessed by pressing the Delete key while booting the PC. There should be an entry on one of the menus in the BIOS to enable Hyper Threading.
5) Message boards : Number crunching : Optimized Boinc Question (Message 242303)
Posted 4 Feb 2006 by Profile Webmaster Yoda
Post:
This machine:
http://setiathome.berkeley.edu/show_host_detail.php?hostid=2148393
I noticed that it is only downlaoding 2 WU at a time.


Looks like last time it ran benchmarks it came in really low. With such low benchmarks, BOINC must think the computer is only able to do two WU a day.

You could re-run benchmarks (manually) and see if it improves. Otherwise there may be something wrong with the computer, or it's running something else that's taking up a lot of CPU cycles.
6) Message boards : Number crunching : Do we have a Boinc virus? (Message 242074)
Posted 4 Feb 2006 by Profile Webmaster Yoda
Post:
Do you still want to stick to a single machine with a RAC of 760?


Easily done with a fast machine running optimised apps. My Athlon 64 3700+ has an RAC over 770 and my 3.4GHz Pentium 4 has an RAC over 1000. Dual core and dual processor machines can get RAC's over 2,000.

See http://setiweb.ssl.berkeley.edu/top_hosts.php

EDIT: There's something like 500 hosts with an RAC over 1,000
7) Message boards : Number crunching : What is the best optimized seti client and boinc client for P4 3.0GHZ HT SSE3 ?! (Message 240970)
Posted 2 Feb 2006 by Profile Webmaster Yoda
Post:
I use cruchers p4 sse3 optimised app with no issues on my 3.2ghz HT machine 512mb ram.
Haven't had an error with it yet, I now do a seti unit in about 31minutes. Seems very stable to me.

I use the SSE3 versions on my Athlon 64 3700 and 3.4GHz Pentium 4 with HT. No problems at all.

Question for llama009 - do you mean 31 minutes for each WU or 62 minutes to complete two, one on each logical CPU? 31 minutes each sounds extremely quick for that CPU, since my 3.4GHz P4 takes approx. 55-60 minutes per WU (two at a time)
8) Message boards : Number crunching : Reference WU (Message 240637)
Posted 1 Feb 2006 by Profile Webmaster Yoda
Post:
Is there any way we can view the results for the reference WU?
I think it would be useful to see how various CPUs or apps compare.


Something like http://www.marisan.nl/seti/reference.htm ?

9) Message boards : Number crunching : Do we have a Boinc virus? (Message 240564)
Posted 1 Feb 2006 by Profile Webmaster Yoda
Post:
All speculation... So here's a bit more.

He claims he doesn't know how to write a virus and we have no proof that he can or has done so. Still, Google his name and observe the interests/skills someone with the same name appears to have. Proves nothing, but it's interesting.

While from the project's point of view I guess these hosts are helping, if SETI is running on them without the owner's knowledge and consent (as appears to be the case with the initial report), it would be a violation of the rules, by the person who installed it or caused it to be installed.

I'm sure BOINC/SETI would not want to be a knowing party to, or beneficiary of illegal/criminal activity, if that's what this turns out to be.

At any rate, Matt says he's looking into it. Let's leave it to him and see what happens.

Meanwhile, keep crunching, people!

[edit]
P.S. If someone else did this, Carsten would only know about it if he looked at his stats, which not everybody does.
[/edit]
10) Message boards : Number crunching : new CPU list (Message 240247)
Posted 31 Jan 2006 by Profile Webmaster Yoda
Post:
Imho best buy would be Dual Core Opteron... 1 MB cache per CORE...


Pentium D 900 series has 2MB per core and may be cheaper. But it depends on which models you compare (a tad difficult as MHz doesn't mean much in comparing AMD and Intel processors)
11) Message boards : Number crunching : A64 optimizing: SSE2 vs. SSE3? (Message 239178)
Posted 29 Jan 2006 by Profile Webmaster Yoda
Post:
P4 running at 2.6ghz v a XP running at 1.8ghz. Nice comparison. Check your facts next time before posting.


Then compare MY (stock) 2.4GHz Pentium 4 and (overclocked to 2.34GHz) Athlon XP 3000+. Very close in terms of raw MHz. Both have PC2700 RAM (stock on the P4, overclocked to 180MHz on the XP)

With optimised SETI apps matching the CPU's capabilities, the P4 (with SSE2) easily outperforms the Athlon XP (with SSE). On Einstein (pre Albert) and Predictor, it was the other way around, with the XP being a lot quicker than the P4.

Given the above, I think the difference is SSE2.

But yes, for Athlon 64's the story is different, as they have SSE2 (and SSE3 depending on model).
12) Message boards : Number crunching : Supercomputers! (Message 238814)
Posted 28 Jan 2006 by Profile Webmaster Yoda
Post:
I believe it is "Maximum daily WU quota per CPU 100/day" because I have a dual processor system with HT on each chip, Seti thinks it is 4 chips, and this is what it says when I view the computer.


According to the BOINC Wiki:

"The Daily Result Quota is multiplied by how many CPU's a Host has, with an upper limit of 4 CPU's. For computers attached to the SETI@Home project this means that that computer can download a maximum of 100-400 result/day, depending on how many CPUs are present and enabled by the Preference Setting "Use at most 'x' CPUs"."

13) Message boards : Number crunching : **** Closed SETI/BOINC Milestonesâ„¢ V Closed **** (Message 238775)
Posted 28 Jan 2006 by Profile Webmaster Yoda
Post:
16K SETI in a week, 50K SETI in 30 days, 1k SETI RAC on one computer :-)

Record power bill in 60 days :-(

Time to retire some older computers.
14) Message boards : Number crunching : Supercomputers! (Message 238135)
Posted 27 Jan 2006 by Profile Webmaster Yoda
Post:
Well, assuming there isn't a maximum :P How many units could the top computer crunch in 24 hours, is my question.


According to BOINCStats, SETI is running at 191.956 TeraFLOPS, with 343,588 active hosts ranging from 486's up...

According to the article, BlueGene can do 136.8 TF/s with 65,536 processors. That's a little over 2 GFlops per CPU, which is double the speed of the theoretical reference computer for BOINC.

How many work units? My guess would be 65,536 * 2 * the number the reference computer would do (whatever that number is). In theory anyway :-)

But if we look back in 20-25 years we'll probably laugh at those figures. Anyone got specs for a CRAY supercomputer of the early to mid 1980's?
15) Message boards : Number crunching : dual or single channel for ram (Message 237686)
Posted 26 Jan 2006 by Profile Webmaster Yoda
Post:
Absolutely! Example:

P4, 2.4 dual channel, no HT: RAC 459.75 (533 MHz bus)
P4, 3.06 single channel, HT: RAC 466.54 (533 MHz bus)


RAC is such an inaccurate figure it doesn't prove a thing. My own 2.4GHz P4 (stock speed, single channel) has a lower RAC but actually completes WUs quicker than yours (assuming we're talking about host 1999617)

The RAC on mine (host 1966827) is much lower mostly because it only runs a few hours a day - if I ran it 24/7, an RAC over 500 would be possible.

The only test that would give a reasonable indication would be to run the reference work unit on one and the same machine in both configurations (single and dual channel).
16) Message boards : Number crunching : dual or single channel for ram (Message 237685)
Posted 26 Jan 2006 by Profile Webmaster Yoda
Post:
OOPS - duplicate post
17) Message boards : Number crunching : **** Closed SETI/BOINC Milestonesâ„¢ V Closed **** (Message 237403)
Posted 25 Jan 2006 by Profile Webmaster Yoda
Post:
BOINC@Australia reached 5 million SETI credits today and are still accellerating. We also reached 14 million credits across BOINC projects today.

Any Aussies out there not in a team yet? It costs nothing to join our team and be part of the push to 10 million. Come on Aussie, come on!



18) Message boards : Number crunching : Which Dual Core? (Intel or AMD?) (Message 236915)
Posted 24 Jan 2006 by Profile Webmaster Yoda
Post:
That said - I truly would dislike having to part with it since I put time & money into making it...


Then keep it and build a new box with a new mobo, dual core CPU (Pentium D 900 series with 2*2MB Cache) and matching RAM in a cheap case. Put one of the old HDD's in it and you have an extra PC for crunching!

You can only spend so much, before you start digging a hole

Yeah, I'm in it :-) Every time I get a power bill I decide to switch my two slowest PC's off as they produce very little anyway, but then I get to the point that I just want that little more credit, so there we go again!
19) Message boards : Number crunching : Which Dual Core? (Intel or AMD?) (Message 236731)
Posted 24 Jan 2006 by Profile Webmaster Yoda
Post:
My mobo is Intel socketed type, & only takes Intel Pentium 4 "Extreme Edition" 478-pin cpu's as their FINAL upgradability!

I.E.-> The most I can hope for is 3.4ghz edition with 2mb L3 cache onboard.

Downside is the price, @ $950 per cpu, according to pricewatch, was here:


Yeah, you could probably buy a new Dual Core CPU + Mobo + fast DDR2 RAM for less.
20) Message boards : Number crunching : Calibrating Client v5.3.12.tx37 - fair credits in all projects (Message 236660)
Posted 24 Jan 2006 by Profile Webmaster Yoda
Post:
@trux

I'm using one of your calibrating clients but just need to be sure in my own mind that I'm doing the right thing.

You say "Normally, the SETI@Home reference WU, when calculated by a reference machines should result in credit of 32.32 cobblestones"

All I can find (in the WIKI) is:

A Cobblestone is 1/100 day of CPU Time on a Reference Computer that does:

  • 1,000 double-precision MIPS based on the Whetstone Benchmark.
  • 1,000 VAX MIPS based on the Dhrystone Benchmark.



So for 32.32 cobblestones it would take the reference computer 27924.48 seconds (0.3232 days) to calculate the reference work unit.

Where can I go to verify that this is indeed the case (given I don't have such a computer)?



Next 20


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.