PSP and SOB Sieving

[[XC] gomeler]

Ok I just got finished tracking down all the relevant info and it looks like we have the ability to earn double the points on these projects. When sieving ranges for PSP, the fact.txt files that contain the factors can ALSO be submitted to SOB Sieve to help them out also. You must be running the proper DAT file which is included in the guide that I sent out and is the same file included in all the guides floating around the net. Now here is the process you need to follow to get most out of your time.

1) Submit your results to SOB Sieve, make sure you create an account and log in first!
http://www.seventeenorbust.com/sieve/

2) Submit your results to PSP Sieve
http://www.psp-project.de/sieveimport.php

3) Reserve another range and repeat!
http://www.mersenneforum.org/showthread.php?t=2666

I hope this is correct, I'll know within a day or so :rolleyes:

[[XC] DragonOrta]

Gomeler, what happens with the SOB stuff with the combined .dat? Did you ever get points for it?

[glennpat]

I am running with the combined dat and getting points for both. I request the ranges from: http://www.sierpinskisieve.com/. When reserving a range you put "combined dat" in the comment field.

There are some instructions at: http://www.free-dc.org/forum/showthread.php?s=&threadid=3501

[glennpat]

Nice going XS_parts on PSP Sieve. That was a big 1st day. With you and gomeler gains the last couple of gains we should be able to stay ahead of Free-DC.

[[XC] gomeler]

Wow HUGE first day XS_Parts! I don't know if either of you two are using Sr2Sieve but I found it to be about 25% faster clock/clock on my Core 2 compared to JJSieveSSE2. If you aren't using it you might want to check it out :up:

[XS_parts]

I'm using the 64-bit sr2sieve on linux.

[glennpat]

The linux version of sr2sieve was slower on my AMD Athlon 64 X2 versus JJSieveSSE2 under Wine. I went from 549 Kp/s down to 528 Kp/s. I had tried it a few days ago when I read about the newer version of sr2sieve being faster on some Intel Dual cores.

Edit -9/29/2007 The above numbers for sr2sieve are wrong. Sr2sieve is faster by a lot. Set post 10.

[XS_parts]

Interesting.... I get 672 on amd-64-3200 on combined using sr2sieve 1.5.17. I'm running on Mepis-64. On an X2-3800 on aGigabyte mb, running Mepis-64, I'm getting 726 per core, no overclock, also using sr2sieve 1.5.17, and that is on a single channel system memory-wise too.

[[XC] gomeler]

Wow those 64-bit clients must really help you out. I get 850kp/s per core at 3200MHz on a Core 2, sr2sieve in XP Pro x86.

[glennpat]

Interesting.... I get 672 on amd-64-3200 on combined using sr2sieve 1.5.17. I'm running on Mepis-64. On an X2-3800 on aGigabyte mb, running Mepis-64, I'm getting 726 per core, no overclock, also using sr2sieve 1.5.17, and that is on a single channel system memory-wise too.

I went and tried it on another box and went from 499 with JJSieveSSE2 to 712 with sr2sieve. I went back and reran sr2sieve on the box I did before and now I get 786. On that box I was getting 549 with JJSieveSS2. I am wondering if I may have still had EON running before when it looked like sr2sieve was slower.

Thats over a 40 percent improvement! Thanks XS_parts :up:

[XS_parts]

Thats over a 40 percent improvement! Thanks XS_parts :up:

For your score's sake, I'm glad that you took the time to go back and re-try the 64-bit client.

If you had been running eon, then just for ha-ha's it would be interesting what you are getting out of jjsieve without anything else running other than necessary o/s system events, and then just the jjsieve. If you were running eon, then it should stand to reason that your jjsieve score should increase too. Otherwise, maybe it was just a bad hair day maybe?? :ROTF:

But then, like the bumper stickers say.... $h** happens. :)

[UCmajewski]

I am going to give this a try on 32 Bit vista

[glennpat]

For your score's sake, I'm glad that you took the time to go back and re-try the 64-bit client.

If you had been running eon, then just for ha-ha's it would be interesting what you are getting out of jjsieve without anything else running other than necessary o/s system events, and then just the jjsieve. If you were running eon, then it should stand to reason that your jjsieve score should increase too. Otherwise, maybe it was just a bad hair day maybe?? :ROTF:

But then, like the bumper stickers say.... $h** happens. :)

I found out the mystery and it was not eon. I was really running an old sr2sieve version 1.4.34 and not the new version. When I started the program up I did a "sr2sieve -s". I should have done a "./sr2sieve -s". Without the ./ it picked up an old version that I had in the /BIN directory that I had tried a few months ago when I was running a different project

I also tried it (32 bit version) on a Celeron 3.33 Ghz running linux and went from a rate of 246 to 493. My total for my 5 cores I went from 2336 to 3489. Nice inprovement for no cost. Since I run eon along with the sieve I won't really get that much improvement, but eon is in a wait mode a lot waiting for work.

[XS_parts]

ah yes, the old "bin path" trick. :up: That has gotten everyone's goat once or twice in the UNIX world. heh heh

Every now and then something comes along that just doesn't seem right, and I'll spend a bunch of time going nuts over it. Usually turns out I just wasn't *really* seeing what was on the screen in front of me. Rather I *thought* saw something else cause that was what I was expecting. I used to get really hyper in a hurry on the keyboard, but over the years I have learned that it is something stupid I did, rather than the system. :)

[XS_parts]

I just finished running tests on the 64-bit linux sr2sieve versions of 1.5.17, 1.5.19, and 1.6.3. And the winner was 1.5.17 by about 30kps. Not much, but then every little bit helps. :)

[XS_parts]

heh heh heh

I
see
roadkill
looming
on
the
HORIZON!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! :yepp:

[[XC] gomeler]

Holy cow XS_parts, HAX!!! I use to have a huge chunk of pie and now it's... tiny :(

[XS_parts]

Well I could either let your piece of the pie get larger again, or let it get smaller. Your choice. :)

:confused:

By the way what does HAX mean?

[[XC] gomeler]

hax = hacks = what people usually yell in first-person shooters when someone is kicking ass and taking names, legitamately. Basically they are playing so good it looks like they are hacking.

"Hahahah I just got a triple-headshot"
"HAX!"
"n00b.."

[Joe O]

New combined dat file is online at the usual place:

http://www.psp-project.de/test/sievecomb.zip

Time to update all the guides.

Yes, you can change dat's in mid-range.
Just stop the program, put in the new dat with the same name as the old one, and restart the program.

There is one less k, since a prime was found for k=33661.

There are also a whole lot less k n pairs in the other k's as well.
All submitted factors were used to clean up the dat. About 1MB smaller file. Still too big to attach though. BZ2 and gz versions are available if you PM me your email address, or email factrange at yahoo dot com.

[[XC] gomeler]

Submitted second to last chunk on my final range for PSP:S, switching idle core over to SOB:S now.

[[XC] gomeler]

Nice little speedup with taht new SOB.DAT file, hitting 930kp/sec now at 3.22GHz :yepp: Will knock out those SOB ranges in no time.

[Joe O]

gomeler;2530399']Nice little speedup with taht new SOB.DAT file, hitting 930kp/sec now at 3.22GHz :yepp: Will knock out those SOB ranges in no time.

I do hope that you are using the most recent SR2sieve program as well.

http://www.free-dc.org/forum/showthread.php?t=12673

[Joe O]

There is 14.4T of low p reservations at http://www.mersenneforum.org/showthread.php?t=2666 that need doing. Lower p usually means more factors. The first 6.4T should have a lot of SoB factors <20M as well. Come get them while they are hot! This is all that is left unreserved below 100T. 100T-120T reservations will be directly controlled by LTD. After this there will only be the Combined Automated Reservation System http://www.sierpinskisieve.com/index.php

[em99010pepe]

sr2sieve-1.7.5 is out, watch the boost....
The client can be used with PSP Sieve, SOB Sieve and RS sieve.

http://geocities.com/g_w_reynolds/

[glennpat]

sr2sieve-1.7.5 is out, watch the boost....
The client can be used with PSP Sieve, SOB Sieve and RS sieve.

http://geocities.com/g_w_reynolds/

I tried it on a combined SOB.dat and got an 8% percent improvement. It was running a around 1360000 p/sec using the 64 bit linux version on a X2 3600+ clocked at 2.81 Ghz. I tested it on a small range where I had found a couple factors before and it worked ok.

[em99010pepe]

My sieve speeds for RS were with 1.7.4 version:

AMD 64 3000+ (2GHz) - 240kp/sec
P4 3.0GHz - 240 kp/sec
Q6600 (@3.0GHz) - 536 kp/sec/core

and now with 1.7.5 version:

AMD 64 3000+ (2GHz) - 286kp/sec (+19&#37;)
P4 3.0GHz - 323 kp/sec (+35%)
Q6600 (@3.0GHz) - 680 kp/sec/core (+27%)

Big improvements only with RS and RS5 (lot's of k's).