mersenneforum.org  

Go Back   mersenneforum.org > Extra Stuff > Miscellaneous Math

Reply
 
Thread Tools
Old 2021-03-27, 01:15   #34
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

32·601 Posts
Default

Quote:
Originally Posted by tuckerkao View Post
If most of other users have their computers turned off during the early morning of U.S. Eastern Time and Ben Delo's machines are on, that determines my chance.

While Ben is automatized, a lot of other users including me are not.
That's a big if. The fleet curtisc has going at his university are probably on 24/7 at least 9 months of the year (maybe 12) except for maintenance or extended power outages. (Where I worked was 24/7/365 except for maintenance or extended power outages longer than UPS runtime). Many schools and businesses leave most gear running at all times. Off hours when users are not using them, is a good time for virus scans, backups, installs and updates. Same applies to my little fleet of cpus and gpus. Some power down for the hot months. All my prime95 instances are PrimeNet connected, except occasionally a test version is installed aside a production version and the test version may not be PrimeNet connected. Colab mprime via PrimeNet too.
The odds are much less affected by number of users doing this, than throughput of those users that are. Higher throughput users will have some degree of automation in place. The top N in the top 500 producers list provide half the total project throughput, where N is a small number. A bit of spreadsheet work with https://www.mersenne.org/report_top_500_custom/?type=0 as input shows N is 3 to 7, bounded by assuming users ranked by overall throughput #501 - #4612 contributed negligible amounts, or assuming all #501 to #4612 users in the past year contributed as much as #500.
All contributions are welcome, big small or medium. Every user in the project began with 0 GHzD in their past year.
Attached Thumbnails
Click image for larger version

Name:	GHzD per year for top N GIMPS top producers overall.png
Views:	59
Size:	39.3 KB
ID:	24573   Click image for larger version

Name:	running sum GHzD per year for top N GIMPS top producers overall.png
Views:	55
Size:	27.3 KB
ID:	24574  
kriesel is offline   Reply With Quote
Old 2021-03-27, 01:46   #35
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

2×1,579 Posts
Default

Quote:
Originally Posted by tuckerkao View Post
If most of other users have their computers turned off during the early morning of U.S. Eastern Time and Ben Delo's machines are on, that determines my higher chance.

While Ben is automatized, a lot of other users including me are not.
Prime95/mprime does not receive new work at midnight in any specific time zone, but whenever it is low on work or out of work no matter what time it is. Why would it wait for midnight for some specific time zone?

I think most users on this forum has one or more 24/7 "automatized" computers, while I guess some users like you do not run them 24/7 for power, temperature or noise reasons.

Last fiddled with by ATH on 2021-03-27 at 01:47
ATH is online now   Reply With Quote
Old 2021-03-28, 22:34   #36
tuckerkao
 
Jan 2020

15816 Posts
Default

Quote:
Originally Posted by Viliam Furik View Post
As far as his profile on the forum says, he lives in Hong Kong, so you are probably wrong about the timezone in one way. You are also (this one is for sure) wrong about the timezone in another way. It doesn't matter where he physically situates himself. He uses AWS with which he doesn't have to interact at all unless he really wants. You would need to know, which server center his machines are set up in. He gets assigned exponents throughout the entire day. Do you know what the word 'automatized' means?

No, you obviously don't.
My main point is that why should I use UncWilly's "Predict M52" thread when I can easily score up 4 guesses using the P-1 factoring today alone?

I expect these 4 exponents will yield some composite results within the next 3 days, guess who's running all of them?
https://www.mersenne.org/report_expo...3330391&full=1
https://www.mersenne.org/report_expo...3337323&full=1
https://www.mersenne.org/report_expo...3373321&full=1
https://www.mersenne.org/report_expo...3397341&full=1

Quote:
Originally Posted by Uncwilly View Post
Most users leave their machines on 24/7. I get assignments all times of day and night.

You are
How come you haven't finished a PRP test on at least 1 exponent with my P-1 factoring signed on it?

Quote:
Originally Posted by slandrum View Post
I really don't understand why. Do you have a crush on him or something?

It really doesn't matter what machine does the work, as long as it's done correctly.

It's not even Ben's software running, you realize, he just has access to a tremendous amount of computing power that he's donating to the cause. You don't even want to guess how much that computer time is probably costing him.
I just want him to test all of my guesses from time to time, so the trial and P-1 factoring tasks are more entertaining, that's all.

To be clear, my real crush is not Ben Delo, here's the link to my ultimate hero - https://en.wikipedia.org/wiki/Xiumin

Last fiddled with by tuckerkao on 2021-03-28 at 22:52
tuckerkao is offline   Reply With Quote
Old 2021-03-29, 00:51   #37
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2×4,903 Posts
Default

Quote:
Originally Posted by tuckerkao View Post
How come you haven't finished a PRP test on at least 1 exponent with my P-1 factoring signed on it?
I am using work machines. I don't want the hassle of finding a prime on a machine without all of the t's dotted and i's crossed. So I have most workers doing DC's, 1 is doing P-1 all of the time, one worked was doing PRP-CF, but I am using it now for some P-1 on some Cat 0 exponents that are stuck on P-1 by others. Sometimes I use that work for targeted DC's. I chose to work on parts of the project I like. For years I worked on TF in the 332M range to keep loons from wasting time doing LL's on exponents that have a factor that is practical to find.
Uncwilly is offline   Reply With Quote
Old 2021-03-29, 03:01   #38
tuckerkao
 
Jan 2020

15816 Posts
Default

It was surely nice that ViliamF found a factor between 2^78 and 2^79 for M103,332,391
https://www.mersenne.org/report_expo...exp_hi=&full=1

This is only practical for him. If I run all those P-1 and trial factoring from 2^76 to 2^79, it'll take around 5 to 6 full days which is more than half of the time duration compare to a PRP test on my PC.

2^73 to 2^74 -> 1 hour+
2^74 to 2^75 -> 2 hours+
2^75 to 2^76 -> 4 hours+
2^76 to 2^77 -> 8 hours+
2^77 to 2^78 -> 16 hours+
2^78 to 2^79 -> 1 day and 8 hours+

If my GPU gets too hot, it's even slower than that.


I plan to perform the trial factoring to 2^77 and P-1 with the suggested bounds for the exponents below. Unless ViliamF wants to do the higher bits, I probably will directly run the PRP tests.

M168,3**,3*1

402, 415, 423, 432, 454, 463, 472, 487, 489, 498
705, 712, 715, 735, 753, 769, 771, 777, 784, 793, 798

The whole list is probably too much, the 2 exponents I'm more likely to go after are M168,347,321 and M168,377,311 consider I've spent over 5 weeks for M168,377,323

Last fiddled with by tuckerkao on 2021-03-29 at 03:27
tuckerkao is offline   Reply With Quote
Old 2021-04-01, 00:23   #39
tuckerkao
 
Jan 2020

34410 Posts
Default

For some reason I had a very bad luck landing on the exponents with the F-PM1 results recently. I've done a whole bunch of P-1 factoring in the M103,300,000+ range, the 1 I skipped was the candidate that had an interesting factor -
https://www.mersenne.org/report_expo...exp_hi=&full=1

I hadn't found a single factor during the entire March 2021. I recalled someone said that there was around 7% to 8% chance finding a factor through P-1 factoring.
Attached Thumbnails
Click image for larger version

Name:	Mersenne NF-PM1 List.jpg
Views:	39
Size:	297.1 KB
ID:	24601  

Last fiddled with by tuckerkao on 2021-04-01 at 00:36
tuckerkao is offline   Reply With Quote
Old 2021-04-01, 00:43   #40
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

100110010011102 Posts
Default

I didn't look at what the chance of finding a factor you had for your individual exponents. But, let's suppose that it is either 3 or 5 %. At 3% you would expect 1 factor for every 33 exponents, on average (in large groups). If it is 5% (which I doubt you are doing), it goes to 1 in 20, again on average. So, doing 40 and not finding a factor is well within that 1 in 20 average. Do lots of tests 300 or more, then worry about your averages.
Uncwilly is offline   Reply With Quote
Old 2021-04-01, 00:57   #41
tuckerkao
 
Jan 2020

1010110002 Posts
Default

The no factor results also appeared on all of my trial factoring exponents for the entire March 2021.

I'll keep trying if the no factor list grows too long, I'm wondering whether there can be the hardware issues.
Attached Thumbnails
Click image for larger version

Name:	Mersenne Trial Factoring with No Factors List.png
Views:	55
Size:	157.3 KB
ID:	24602  
tuckerkao is offline   Reply With Quote
Old 2021-04-01, 04:06   #42
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

32·601 Posts
Default

I count 55 NF lines. That is a very likely outcome for such bit levels.
For probability 1/75, 55 tries, chance of zero factors 48%. https://www.gigacalculator.com/calcu...calculator.php
kriesel is offline   Reply With Quote
Old 2021-04-05, 06:40   #43
tuckerkao
 
Jan 2020

23·43 Posts
Default

Quote:
Originally Posted by kriesel View Post
I count 55 NF lines. That is a very likely outcome for such bit levels.
For probability 1/75, 55 tries, chance of zero factors 48%. https://www.gigacalculator.com/calcu...calculator.php
I got 17 more NF or NF-PM1 lines since that list.

It seems like I cannot get anyone else on exponent this size, so I have to test those all myself, anyone feels like trial factoring between 2^78 to 2^79 is necessary or not?
https://www.mersenne.org/report_expo...exp_hi=&full=1
tuckerkao is offline   Reply With Quote
Old 2021-04-05, 14:40   #44
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2×4,903 Posts
Default

Quote:
Originally Posted by tuckerkao View Post
It seems like I cannot get anyone else on exponent this size, so I have to test those all myself, anyone feels like trial factoring between 2^78 to 2^79 is necessary or not?
Do all the factoring you want, just stop whinging about it. Also, stop begging people to do work for you. Run PRP's (with version 30) on your own numbers.
Uncwilly is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Automatic fetch of Trial Factoring work for GPU mfakt* LaurV GPU to 72 81 2020-12-02 05:17
Simple Script to get Trial Factoring Work jfamestad PrimeNet 3 2016-11-06 20:32
Why trial factoring work chopped into chunks? lidocorc PrimeNet 4 2008-11-06 18:48
How does the trial factoring work with 15K*2^n-1 jocelynl 15k Search 0 2003-07-11 14:23
How does trial-factoring work? ThomRuley Software 5 2003-05-30 20:34

All times are UTC. The time now is 11:37.


Thu Jul 29 11:37:18 UTC 2021 up 6 days, 6:06, 0 users, load averages: 2.23, 1.87, 1.67

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.