Trial Factoring Progress !!

[rsbriggs]

Well, we've moved up to 195 th place in Trial Factoring. :thumbs:

And the good news is We're on track to overtake Team Prime Rib !! :elephant:

The bad news is, the current overtake date is 08-Jul-2163 :geezer:

[rsbriggs]

Up +7 places to # 188 since yesterday.

The overtake date has moved out to 01-Jan-2661. :scared:

Oh well, only about 656 calendar years to go, but what the heck does a mere 239,447 days mean, anyway?

It'll go by quickly.... :sleepy:

[MonsterMan]

briggs, your 30m exponent was the first one issued. Congrats on being the first.

Up to 180.

I would like to see us in the Top 100 by next week. Once we hit 100, we will be able to watch our progress on TPR's graph. I like graphs.

Oops, another factor for me. Only 4000 more to go.

[rsbriggs]

You hit two factors overnight. Nice job....

We're up to 179th and an overtake of 01-Dec-2030. This just keeps becoming more and more fun as we go along.

Well, I don't feel special for getting the first 30M exponent. ;) Does that put us in the range of TF'ing for 10M digit numbers now or not? :cool: I don't know where we cross over to that....

4000 factors to go before overtaking ARS, but did you notice that they aren't the leaders in exponents, just CPU years?? :D Netforce, in 10th place, has nearly twice as many as ARS does.

Hmmm. Need more gas. Go faster. Have any particular tips or advice about runing this on multi-CPU / HT machines?

(The 85th to 100th place graph looks really cool, BTW. Can't wait to be on there....)
http://www.teamprimerib.com/rr1/bin/top_graphs.php?t=p&g=7

[MonsterMan]

For multi CPU machines, run one instance for each CPU in a separate directory.

I know that hyper-threading doesn't help with LL testing, but I haven't done any research with TFing. I checked the mersenne.org forum site and the concensus is it could help. But each machine is different and some overclocks need to be turned down. Prescotts work better than Northwoods because of the 1 mb cache. Benchmarking it is the only true way to find this out.

NetForce kind of cheated with that. In the beginning of Gimps, they were the only team/person doing TFing. And they started from the beginning of Gimps, so they found all those small factors which took like 10 seconds to find.

Also the 10m digit numbers start in the mid 33m range. We haven't got there yet, but some people have jumped ahead and received assignment from George Woltman, the head guy at Prime95.

[birdman2584]

well, i finished a TF last night sometime, and have moved onto the next one. When do i get on that list with you guys at the bottom of the screen at the GIMPS site here (http://mersenne.org/cgi-bin/primenet_report.pl?UserID=FreeDC&UserPW=go-4-it) oh well...top 100 here we come...even better...TPR here we come:mouserun:

[MonsterMan]

The list on the bottom is when you actually find a factor so the number is considered composite and out of the testing. For each factor found, we save almost two weeks of LL testing. Factors are good, they eliminate that exponent from being a Mersenne prime.

Keep crunching and sooner or later you will find a factor. We need gazillions of factors.

[rsbriggs]

Don't worry about not finding a factor - many, if not most, of the trial factoring runs probably won't produce one. I'm actually quite surprised that I found one after just a couple of days worth of crunching. Usually I have spend a month running an LL test to show up down there.

Factor or not, any time spent trail factoring is added into the teams CPU-years count, which is just as important a metric as factors (to some of us, anyway) since it counts towards the teams overall standing.

Monster Man seems determined to crank out 5,000 factors or so and beat out TPR - preferably this weekend. :) (Or are you wanting 11,000 or so to beat out NetForce?) I'm pretty content to just see our overall team standing go up a little every once-in-a-while. When I started the GIMPS team, Free-DC was still considered as being part of TPR. The team has come up thousands of places from its initial start - don't remember exactly how many - 20,000? 30,000?. Actually, it's quite awesome to be thinking about hitting the top 100 teams in GIMPS.

GIMPS is a long term kind of project, been around for years, will continue to be around in years. To paraphrase a well known Ent: "No need to be hasty. young cruncher..." You'll show up in the at the bottom, just keep crunching !! :thumbs:

[birdman2584]

i have another question about the stats on TPR's site. When it shows total producers output here (http://www.teamprimerib.com/rr1/bin/top.php?type=p&r=3&so=r) it shows that we have done no work within the last 7 30 or 90 days. Does TF not count toward producing, or is it completely separate from TF. Thanks, this project is still a bit confusing to me.

[MonsterMan]

That page only shows Lucuas-Lehmer or Double-Checking years. Lucas-Lehmer is the same as DoubleChecking but doublechecking is just that, it doublechecks the results of the original LL test. Only after 2 matching results are submitted is an exponent considered not a Mersenne prime.

So TFing and LL testing are measured separately, but in total p90 CPU year output, we have moved up.

The main reason I care about factors is that they save the project so much time. One factor saves at least two weeks of LL testing and the doublecheck.

[MonsterMan]

Don't take this the wrong way, but I will be leaving for a week starting the 6th of June. I promised some guys that I would let them use my hardware for a week about a month ago. And one thing I really like to do is keep my word. It should only be the 6th through the 13th, so rest assured I will return.

Kick some butt while I'm gone, rightyO.

[rsbriggs]

No problem - I suspect that GIMPS will still be here when you return. :D

I'm going to be kind of slow for the next short while myself - the Cern Linear Heavy Collider sent out some work, which I have setup to preempt everything on all of my boxen until it gets processed....

Did you happen to notice this little curious thing? You just started trial factoring an exponent that is less than the LL test I started last week....
What odd set of circumstances would it take to cause that to be happening? :confused:


28473941 F 62 04-Jun-05 06:57 MonsterMan01
28554767 69 01-Jun-05 21:41 rsbriggs_05

Old TF job that was assigned 60 days ago, but just now barely timed out, and wasn't assigned for LL testing in the mean time? You just happened to request a factor and grabbed it before someone requested an LL test?

Anyway, /me finds that to be somewhat of a curiosity....

[MonsterMan]

If you look at this, very long to download, (http://mersenne.org/primenet/status.txt) , you will find first time LL tests not being worked on in the 20M to 27M range. People reserve them and just leave. Sometimes, a machine will reserve 100 to 500 exponents by accident and nobody realizes it for awhile.

There's only one real problem with Gimps and that is the server and it's software are so old. It wasn't made to last as long as it has. With luck, Ver. 5 of the server will be running soon.

[rsbriggs]

Garrgh. People with slow machines shouldn't be doing LL tests - some of those machines have been working on a single LL test for two years....

[rsbriggs]

Ummm.. This is a little discouraging....

Getting exponents from server
ERROR 7: Server has run out of exponents to assign
...
Will try contacting server again in 60 minutes
...
Getting exponents from server
ERROR 7: Server has run out of exponents to assign
...
Will try contacting server again in 60 minutes
...

[MonsterMan]

You can find some answers at the mersenne.org forum found here (http://www.mersenneforum.org/)

[rsbriggs]

Overnight it seems that Birdman found a factor, and looks like the MonsterMan found another as well. :thumbs:

We've moved up to 162nd place in trail factoring, up from 179th place just a couple of days ago....

[MonsterMan]

This is getting really AWESOME !!

And, it looks l ike I won't be going anywhere after all.

A war broke out and I decided to stay neutral. So no go bye-bye.:|party|:

[MonsterMan]

OK, we'll be moving into #160 with the next update, but after that things start to get sticky. Our supposed competition is spaced farther apart, so each victim will take a day or two.

Now would be a good time to talk some friends or Sys Admins into running Gimps trial factoring. Remember, TFing takes very little bandwidth and the program moves out of the way for near everything.

Let's push a bit harder to get into the Top 100 and get on Ars' graph page.

[rsbriggs]

Well, we need 57.5 CPU years in TF to hit the top 100. We're currently at 35.0 CPU years and producing at 1 Trial Factoring CPU year every 4.5 days.

By my calculating, that means we have 22.5 CPU years * 4.5 days per CPU year = 101.25 days to the top 100, and we'll be moving up one place roughly every 1.75 days...

So, here's the countdown to top 100:

60 places / 22.5 CPU years and counting.....

[rsbriggs]

58 places. 22.0 CPU years.

Not a bad days work: Up 2 places and 1/2 a CPU year.

[MonsterMan]

Hot dang!!

With the next update we should be tied for 145th. However, we've lost our overtake date of TPR.

But that's OK. We are still charging up the ladder. One goal at a time and right now it is to get to the Top 100. TPR graph page, here we come!!!!

[rsbriggs]

Up 8 places to 150 place - 50 more to go to top 100....

TPR appears to have increased their output...

[rsbriggs]

An hour later, and we show as being up another 8 places and in 142nd place?
Well, that works just fine for me.... :D

[MonsterMan]

We're now up to 125th. Only 25 more positions to the Top 100.

Keep up the good work everyone.

Also, in the Options header, under Preferences, make the days between sending end dates to "1". This keeps everything fresh.

And if anyone decides to drop out, please send any partially completed work to me, I will finish it.

[rsbriggs]

Up to 119th place and still climbing....

[black_civic55]

what should i have selected in options so that i know I'm contributing the right way???

i think i just started a 10 mill digit...