Reply With QuoteSomething happend...that was a huge jump in n...to 11786612????? That does not seem right.
Gonna try secondpass on my xeons and sieving on my XP for a while until everything settles down. Hate doing duplicate work.
e
Last edited by engracio; 06-09-2006 at 09:27 PM.
Something happend...that was a huge jump in n...to 11786612????? That does not seem right.
Give it a day or so until things sort themselves out. If it is just a runaway client (or something) these tests will be re-released. If it is really a massive influx of computing power, then there is no way for P-1 to keep up with it.Originally Posted by engracio
Sieving anyone?
Joe O
Thanks Joe, its about time for sieving again anyway. Was just waiting for the new siever, guess its working fine now.
e
One user outperformed the ARS team yesterday, all by his/her lonesome. Today he/she slowed down a little, but would still have come in third place as a team.
Joe O
did you see they have 14140 pending tests ?!?!?
With account age of 227 days and 19 tests completed overall...
With the lastest issues in the prp I'm placed in a ackward position, of not knowing the best solution to the issues.
I'm going to make the suggestion that people skip a range of P-1 upto 12M.
Reason being I'm uncertain if Louie or Dave will transfer the dropped tests back into the firstpass que or not. Also even if they can/do we have some legit assigments in the range of 11.3 to 12M....
If we simply left a gap would solve a bunch of issues.
- It would ensure P-1 factors score properly and give P-1 a headstart/lead on prp again.
In either case reserve as you wish.
Makes sense to me, because the next.txt file is blank. But it is at 11975918 taken from Mike's stats page. http://www.aooq73.dsl.pipex.com/2006/scores_t.htm
Yeah...
If you go by this page, http://www.seventeenorbust.com/stats/rangeStatsEx.mhtml
it's pretty obvious everything was handed out to n=12M for firstpass.
Yeah...
If you go by this page, http://www.seventeenorbust.com/stats/rangeStatsEx.mhtml
it's pretty obvious everything was handed out to n=12M for first pass.
13470000 13485000 engracio ? [reserved]
Looks like he went above that, all the way to n=13.5m. Now the big question after everything settles down and the runaway client dropped everything is would the admin put back the dropped test to first pass. If they put it back to first pass would we be able to factor it and still get the benefits of factoring? I.e if we find a factor would the server consider it a passed test?
Personally I do not care where we start over again as long as we do not do double work or for nothing. As what everybody has been saying, the longer the test become the more benefit p1 factoring is. With my ancient cpu's they are getting very long.
e
Last edited by engracio; 06-14-2006 at 08:52 AM.
E,
Those tests around 13.4M are far a few between they are actually from a very old special que 13367 or something like that, I think a total of ~10 were completed.
I can assure you that there were no tests between >12M assigned recently.
Regardless do the range you specified if you wish, but if you havn't got that far I'd suggest taking something a little lower. By the time we get to 13.4M I'd like to see another 2 k's eliminated.
Okay buddy make sense to me, was just not sure. I was just prepping the wu so that when the boxes running the sieving and secondpass prp completes p1 can continue.
Let me reserve this range then:
12000000 12020000 engracio ? [reserved
I can always go back down to the correct prp range when this set is complete. Thanks.
e
======================================
Sounds good e.
VJS
Last edited by vjs; 06-14-2006 at 01:17 PM.
12050000 12051000 jasong [complete] 1 factor
(the very first p-1 in the group, and my very first p-1, period, found a factor. I thought that was a good sign, but of the 22 k/n pairs in that range, it was the only factor. go figure)
1 out of 22 is a very lucky shot... You should expect 1-2% success rate, unless the limits you're using are on the extremes.
yup one out of 22 is pretty lucky. I believe the first time I tried P-1 factoing I found one on the second or third number. Then not again for quite some time.
I gave Prime95 an 800MB maximum which, unless I totally misunderstood something I read about that, should be more than enough.
Question: If you run p-1 more than once, it does the problem slightly differently each time because of randomization, right? So, would that make it possible that another p-1 could find a factor that the previous one missed? Even with exactly the same parameters?
to the best of my knowledge, this is not how it works. it's true for ECM though and this is why we run many curves (trials) in ECM at each boundary.
Consider a prime factor p of your number k*2^n-1. p-1 is not prime, and has a factorisation p=p_1*p_2*p_3*...*q*r, where q and r are the two largest factors.
If r<B1, the factor p will be found in stage 1, if r<B2 AND q<B1, it will be found in stage 2. No random parameters.
In other words: if all prime factors of p-1 are below B1, it's a stage-1-hit, if the largest (and only the largest) is between B1 and B2, it's a stage-2-hit.
Yours H.
___________________________________________________________________
Sievers of all projects unite! You have nothing to lose but some PRP-residues.
11400000 11500000 engracio ? [reserved]
e
=========================
E, you might be working behind prp...
-VJS
Not really, had to dump a few G here and there but I mostly factored 95% of this range. Too bad did not get much factors like the last time. Maybe this next range.
11400000 11500000 engracio 17 [complete]
E
17 factors is not bad. You didn't have to dump all those G because PRP is back down again. Want to go back and get them?
So far you have a lock on the 500000 points and above:Code:p (T) k n Score Factor found Score changed Score was Score could be Reqd bias PRP saved 19387.196T 21181 11625332 528170.686 Mon 17-Jul-2006 Fri 28-Jul-2006 35.000 8187960.721 484.68 (2) engracio 75720.086Z 21181 11606444 526455.813 Mon 17-Jul-2006 Fri 28-Jul-2006 35.000 9999999999999.998 9999.99 (2) engracio 41158.945P 10223 11593949 525322.903 Mon 17-Jul-2006 Fri 28-Jul-2006 35.000 17289284572.827 9999.99 (2) engracio 2277.338T 55459 11561530 522389.195 Mon 17-Jul-2006 Fri 28-Jul-2006 22.773 951279.378 56.93 (2) engracio 84710.325P 10223 11553701 521681.953 Wed 26-Jul-2006 Fri 28-Jul-2006 35.000 35336914058.002 9999.99 (2) engracio 72082.420T 55459 11547478 521120.132 Sat 08-Jul-2006 Fri 28-Jul-2006 35.000 30036800.128 1802.06 (2) engracio 895.728P 19249 11546366 521019.771 Sun 09-Jul-2006 Fri 28-Jul-2006 35.000 373178529.588 9999.99 (2) engracio 45671.456T 33661 11542752 520693.665 Fri 07-Jul-2006 Fri 28-Jul-2006 35.000 19015756.551 1141.79 (2) engracio 137.145Y 10223 11538761 520333.659 Thu 06-Jul-2006 Fri 28-Jul-2006 35.000 9999999999999.998 9999.99 (2) engracio 2683.231T 24737 11537191 520192.073 Fri 07-Jul-2006 Fri 28-Jul-2006 26.832 1116113.234 67.08 (2) engracio 131.898Z 10223 11536997 520174.578 Fri 07-Jul-2006 Fri 28-Jul-2006 35.000 9999999999999.998 9999.99 (2) engracio 3146.918T 21181 11532404 519760.487 Thu 06-Jul-2006 Fri 28-Jul-2006 31.469 1307901.811 78.67 (2) engracio 265.611P 19249 11531282 519659.356 Sat 08-Jul-2006 Fri 28-Jul-2006 35.000 110369928.217 6640.27 (2) engracio 1391.878T 21181 11531228 519654.489 Sat 08-Jul-2006 Fri 28-Jul-2006 13.919 578365.586 34.80 (2) engracio 211.787P 67607 11531187 519650.793 Sat 08-Jul-2006 Fri 28-Jul-2006 35.000 88003001.266 5294.68 (2) engracio 1269.580T 55459 11529694 519516.239 Tue 04-Jul-2006 Fri 28-Jul-2006 12.696 527406.758 31.74 (2) engracio 126.246P 67607 11528667 519423.692 Thu 06-Jul-2006 Fri 28-Jul-2006 35.000 52435361.164 3156.14 (2) engracio 80445.754P 10223 11526077 519190.333 Thu 06-Jul-2006 Fri 28-Jul-2006 35.000 33397671039.595 9999.99 (2) engracio 2321.674T 10223 11525789 519164.388 Tue 04-Jul-2006 Fri 28-Jul-2006 23.217 963812.435 58.04 (2) engracio 24789.747T 55459 11517574 518424.583 Sat 01-Jul-2006 Fri 28-Jul-2006 35.000 10276474.214 619.74 (2) engracio 33213.948E 55459 11513278 518037.915 Mon 03-Jul-2006 Sun 09-Jul-2006 35.000 9999999999999.998 9999.99 (2) engracio 3721.833T 21181 11511932 517916.796 Tue 04-Jul-2006 Sun 09-Jul-2006 35.000 1541357.340 93.05 (2) engracio 4929.227E 10223 11506649 517441.546 Fri 30-Jun-2006 Sun 09-Jul-2006 35.000 2039513603842.301 9999.99 (2) engracio 57274.696T 24737 11506087 517391.002 Fri 30-Jun-2006 Sun 09-Jul-2006 35.000 23695622.414 1431.87 (2) engracio 5757.564T 24737 11502511 517069.450 Fri 30-Jun-2006 Sun 09-Jul-2006 35.000 2380532.438 143.94 (2) engracio 99999.999Y 10223 11501405 516970.019 Fri 30-Jun-2006 Sun 09-Jul-2006 35.000 9999999999999.998 9999.99 (2) engracio 230.196P 24737 11493487 516258.461 Fri 04-Aug-2006 95028016.772 5754.90 (2) engracio 44479.372T 24737 11487751 515743.296 Thu 03-Aug-2006 18343351.824 1111.98 (2) engracio 38834.518T 24737 11474743 514575.968 Wed 02-Aug-2006 15979157.491 970.86 (2) engracio 1747.465T 22699 11469934 514144.746 Wed 02-Aug-2006 718423.334 43.69 (2) engracio 11592.538T 67607 11469851 514137.305 Wed 02-Aug-2006 4765891.026 289.81 (2) engracio 4418.233T 10223 11457101 512994.900 Sat 29-Jul-2006 1812375.068 110.46 (2) engracio 1075.644P 33661 11451408 512485.216 Wed 26-Jul-2006 440794860.563 9999.99 (2) engracio 49946.562T 10223 11450345 512390.075 Wed 26-Jul-2006 20464105.477 1248.66 (2) engracio 1942.552T 55459 11448094 512188.635 Fri 28-Jul-2006 795589.458 48.56 (2) engracio 13872.847T 22699 11445958 511997.524 Fri 28-Jul-2006 5679628.468 346.82 (2) engracio 2935.332T 10223 11441177 511569.888 Wed 26-Jul-2006 1200739.131 73.38 (2) engracio 163.862P 19249 11432642 510806.921 Wed 26-Jul-2006 66930288.095 4096.56 (2) engracio 14941.227T 21181 11432228 510769.927 Wed 26-Jul-2006 6102363.182 373.53 (2) engracio 16420.073T 55459 11430346 505465.178 Tue 18-Jul-2006 2 engracio 6890.336P 33661 11420280 504575.306 Tue 18-Jul-2006 2 engracio 7804.304T 10223 11416709 504259.805 Wed 26-Jul-2006 2 engracio
Joe O
Last edited by Joe O; 08-04-2006 at 07:45 PM.
Joe,
What I meant about dumping several G's or is it M's is that when I see the prp being issued at or just behind what I am factoring, I would dump a few of them until I am just above it. Granted some might have been re issued again but at least I know someone is prp'ing it and I don't want to duplicate work.
Yes, the prp going back and forth gave me enough time to like I said factor about 95% of this range.
My next range will give me plenty of head room I would not even worry about it.
e
E,
I'm currious what you are currently using for B1 and B2 bounds... Seem to be doing pretty well, IMHO.
I'll be glad when I finish my last range under 12m, hopefully it would give me lots of breathing room. Have to constantly check the assignment page for the current n. Last week have to either skip bunch of n just to stay ahead of prp. I try to complete as many n as I can. Alas most of them are not factors but you do not know that unless it gets factored. With the 12m range, I should be able to just stick 100 or so n's per cpu and not worry about it getting passed by prp. Harvest them and send out the factors.
Which reminds me, I feel the assignment page is the most current for the current n being given out. I now prefer looking at it instead of the next.txt page. I guess when we get ahead again from current n being handed out, the next.txt page will be relevant again.
vjs, I use B1=70000, B2=822500 bounds. With my duallies I reserve 560mb per cpu with 1.17% chance. I have 1.5GB memory per box.
e
BTW, what sieve depth and factor worth should be use for the 12m range. Right now I am using 50.1 and 1.8 with the 1160-1175 range. How much to change if any?
You seem to be doing great on your own, better than expected.
I wouldn't change a thing.
Might change your luck.
Louie can you post what is the B1=B2 setting you are currently using on your P1 factoring? Thanks.
e
I'm not using B1/B2. I'm using prime95 24.14 with a sieve depth of 50 and a factor value of 2.1 and specify 256MB of mem for B2. These settings end up causing approximately B1=95000, B2=1092500.
Somewhat overkill. I'm trying to find more large factors like
19750058751527901255535231 | 21181*2^12447884+1
19750058751527901255535230 = 2 x 3 x 5 x 7 ^ 2 x 11 x 17 ^ 2 x 2963 x
32191 x 90863 x 487649
What are you using?
Cheers,
Louie
Louie,
I believe most people choose B1 between 60k-85k and a B1:B2 ratio between 10-16.
Larger than this range your time spent per factor drops quite a bit, but you do find larger factors of course.
The B1:B2 ratio is based upon suggestions from Dr. Silverman where a roughly equal time should be spent during stage1 and stage2.
That is the reason I asked you the question why you are getting lots of large factors. I am also using Prime 24.14 with 50.1 and 1.8 factor value, b1=75000 b2=825000. Since I have 1.5gb of memory per duallie I am using 560mb per instance of prime. It seems that the amount of memory does not really affect the completion per unit any faster. I am just looking for different configuration I can use factoring. Thanks.
vjs,
Comparing his factors found and mine, seems like he is finding a little bit more. Nothing close to being scientific but a Pluto wag (out there guess)
e
Interesting lets compare louies and yours...
12250000 12500000 louie 107
11780000 11930000 louie 44
Basically Louie P-1'ed a range of 400000 n and produced 151 factors.
0.3775 factors per 1000 n-range
11600000 11770000 engracio 30
12000000 12020000 engracio 8
11400000 11600000 engracio 51
E P-1'ed a total range of 390000 n and produced 89 factors.
0.228 factor per 1000 n-range.
Louie used a B1=95K with B2=B1x11.5
E used a B1=75K with B2=B1x11
Basially Louie spent close to 30% more processing power per k/n pair, yeilding a 65% increase in factor probability.
-------------------
Very interesting, E you might want to try higher bounds say B1=110k and if you are not limited in memory and have dual channel try upping your B2:B1 ratio. I wouldn't go much beyond 14.
B2=1500000
Are you still running those MPX's?
The most important point to remember is a little P-1 is better than no P-1 so as long as we are not passing ranges...
To try and translate VJS's recommendation into the parameters that you are using, I would suggest that you try changing your 1.8 factor value to the 2.1 that Louie is using. Depending on how that goes, I would try 2.2 as well. I would recommend that you leave everything else as is, unless you want to lower your factor depth to 50.0 from 50.1 as well or instead of changing the factor value. It would be instructive to see what these changes do to your B1 and B2 values, as well as your run time.
Joe O
Vjs,
Yes on the MPX but they are now sieving and dying slowly, had to use their memory and to increase and make it dual channel the memory on the Xeons and a Pentium D915 oc'd to 3500 mhz. Probably could get a few more mhz but did not want to spent money on memory and better cooling.
Did not know you guys were going to break it down and throw off my galactic wag. Portions of my range were skipped due to the random fluctuations of the prp. I know I skipped a few factors.
I really just wanted to know how Louie was able to complete those range so quick. One quick answer was he had more machines to crunch. doh!
I'll slowly increase the B2 and see if they really make that much difference. My P1 machines are all Intel cpu's and I try to have at least 1gb of memory while factoring. What is odd is I have P4 2.8 with only 512 memory with 452mb allotted, max memory Prime95 would let me allocate. It can crunch a factor every two hours but my Xeon 3.1 could only crunch every two and half with 560mb alloted per instance.
Most of the time it is not used. Ummm.
e
Humm,
I forgot about that in the calc's I guess that throws everything out the window from above.
Still using my MPX as a home machine, dual 2.4G Bartons and 2G of ram, sieve exclusively. Yup it's getting old. I should have sold it off 6 months ago. I could have got ~$400 for just the board and processors
Since this thread is a little old what is the sieving depth and factor value should be used when doing a range?
Thanks.
I wouldn't use sieve depth and factor worth.
I'd simply specify the B1 and B2 values.
Since it will be a while until we get to 14M I would suggest you use.
B1=100k to 120k with a b2=b1x12 to b2=b1x14 depending on how much memory you have.
in other words,
minimum
B1=100000 b2=1200000
maximum
B1=120000 b2=1700000
What CPU memory and amount of memory do you have?
The one I will run on is an Intel dual core 6600 running 2.4 GHz (stock) with 2 Gig of memory. SOB Sieve will also be running on it.
To use the B1 B2 do I use the AdvanceFactor command for PRIME95? I assume I edit the the worktodo.ini to use the B1 B2.
Last edited by glennpat; 06-30-2007 at 07:53 AM.
To specify the desired B1 and B2 values, place the following line(s) in your worktodo.ini file:
Pminus1=10223,2,14000237,1,120000,1700000
where
Pminus1=k,2,n,1,B1,B2
Do not use the AdvanceFactor command, that is for something else.
Last edited by Joe O; 06-30-2007 at 10:28 AM.
Joe O
Thank You! I now see that was in the instructions at the top of the reserve thread.
The 1st few line of my are worktodo are:
Pminus1=21181,2,14000180,1,120000,1700000
Pminus1=10223,2,14000237,1,120000,1700000
Pminus1=21181,2,14000252,1,120000,1700000
Pminus1=10223,2,14000261,1,120000,1700000
Pminus1=55459,2,14000278,1,120000,1700000
It is running.
That looks good to me.
Let us know what your times are per run, it would be interesting to see.
The last two stage 1 times were 6800 and 6893 seconds.
The last two stage 2 times were 5078 and 5119 seconds.
In my range, 14000000 to 14010000, I have 444 numbers left to do. With the one core I can do about 7 per day. From the user stats it has:
"Factors next to enter (main) 'active window' " which is at n=13813352".
From the link http://www.seventeenorbust.com/sieve/next.txt it has:
k n
24737 13629607
Which number do I have to stay ahead of? Looks like I may have to spread the work around on some of my other boxes.
Glenn
Yes you might have to spread the wealth to several cpu/cores to stay ahead. The number fluctuate back and forth from the highest to some lower k and n. Mainly because somebody dropped it and went back to the first pass queue. If you stay ahead of the "next to enter" you'll be fine. At the current rate 2 or 3 cpu/core should be enough to stay ahead until you figured out the system/p1'ing.
On my pd945 3.4 oc'd to 3.9 I get 5767 secs on stage 1 and the other core is 5484 on stage 2. I am using 840mb per core since I have 2gb of memory. 1.34% probability of finding a factor.
e
Posting Permissions