mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Data

Reply
Thread Tools
Old 2013-08-13, 02:10   #826
Aramis Wyler
 
Aramis Wyler's Avatar
 
"Bill Staffen"
Jan 2013
Pittsburgh, PA, USA

23×53 Posts
Default

Quote:
Originally Posted by garo View Post
I think that is more due to GPU72 holding everything from 61m upwards to make sure everything gets TFed to 74 bits before being handed out for LL testing. The thought process being it is better to hand out a 65M with TF to 74 bits than a 61M with TF done to 73 bits. If we find TF is lagging behind I think we could just throw back any 61M or 62M exponents into the LL pool and be fine. No need to mess with the What Makes Sense setting.

Yikes, I thought we were only holding 64M and higher to 74 bits, and < 64M was 73 bits? I've been using my card to factore the <64M ones to 73 bits, to sweep them out as it were, because I thought everyone was doing 64M+ to 74.
Aramis Wyler is offline   Reply With Quote
Old 2013-08-13, 02:36   #827
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

2×3×1,693 Posts
Default

Quote:
Originally Posted by Aramis Wyler View Post
Yikes, I thought we were only holding 64M and higher to 74 bits, and < 64M was 73 bits? I've been using my card to factore the <64M ones to 73 bits, to sweep them out as it were, because I thought everyone was doing 64M+ to 74.
So THAT's where all those 73-74s are coming from! Argh! (Just kidding. I tend to get a lot of Lowest Factored Level assignments, too, for my faster card. But I take everything to 74, no matter where it starts.)
kladner is offline   Reply With Quote
Old 2013-08-13, 04:00   #828
Primeinator
 
Primeinator's Avatar
 
"Kyle"
Feb 2005
Somewhere near M52..

3×5×61 Posts
Default

When is it desirable to go above 74 bits?
Primeinator is offline   Reply With Quote
Old 2013-08-13, 04:09   #829
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11×157 Posts
Default

Chris has been talking about starting up to 75 soon. I think it's a ways away and maybe borderline profitable. However, since from 74 to 75 takes the same amount of time as UP TO 74, and the trial factoring effort as we know isn't exactly doubling the LL pace, I don't think we can afford to move past 74 yet.

On the other hand, more knowledgeable people than I have probably thought this through. This is just what I've gathered over the last little while.
TheMawn is offline   Reply With Quote
Old 2013-08-13, 04:13   #830
Aramis Wyler
 
Aramis Wyler's Avatar
 
"Bill Staffen"
Jan 2013
Pittsburgh, PA, USA

1101010002 Posts
Default

Quote:
Originally Posted by kladner View Post
So THAT's where all those 73-74s are coming from! Argh! (Just kidding. I tend to get a lot of Lowest Factored Level assignments, too, for my faster card. But I take everything to 74, no matter where it starts.)

lol, I'll release the rest of the 73s and take them all to 74. That must have changed while I was away and I missed it.
Aramis Wyler is offline   Reply With Quote
Old 2013-08-13, 04:25   #831
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

2×3×1,693 Posts
Default

Quote:
Originally Posted by Primeinator View Post
When is it desirable to go above 74 bits?
This is a question of much discussion and dispute. AFIK, for the time being it is only recommended in the hundreds of millions exponent ranges. Also, at any level there is a break-even point, beyond which the TF does not save any time and effort versus an LL test.

This page
gives current suggested guidelines, though these have been adjusted from time to time. Part of the calculus for GPU72 is how far we are staying ahead of LL assignments. When we pull ahead, there's time to take factoring to higher levels. If LL is catching up too quickly it may be better to drop back a level and complete more to lesser levels, instead of fewer to higher levels. (Please be aware that this is a crude explanation by one who has next to no grasp of the maths involved.) Chalsall monitors the assignment rates and can give a much better picture than I.
kladner is offline   Reply With Quote
Old 2013-08-13, 04:56   #832
Primeinator
 
Primeinator's Avatar
 
"Kyle"
Feb 2005
Somewhere near M52..

39316 Posts
Default

Quote:
Originally Posted by kladner View Post
This is a question of much discussion and dispute.

This page
gives current suggested guidelines, though these have been adjusted from time to time. Part of the calculus for GPU72 is how far we are staying ahead of LL assignments. When we pull ahead, there's time to take factoring to higher levels. If LL is catching up too quickly it may be better to drop back a level and complete more to lesser levels, instead of fewer to higher levels. (Please be aware that this is a crude explanation by one who has next to no grasp of the maths involved.) Chalsall monitors the assignment rates and can give a much better picture than I.
Those videos are quite impressive. Kudos to the effort that must have taken!
Primeinator is offline   Reply With Quote
Old 2013-08-13, 06:12   #833
blahpy
 
blahpy's Avatar
 
Jun 2013

107 Posts
Default

Am I correct in saying that LL tests will get reassigned after 1 year if not manually extended? If so, we should be getting a lot of the low double-checks back this month, the lowest next week

http://mersenne.org/report_exponent/...&B1=Get+status
blahpy is offline   Reply With Quote
Old 2013-08-13, 12:55   #834
Chuck
 
Chuck's Avatar
 
May 2011
Orange Park, FL

15658 Posts
Default

Quote:
Originally Posted by kladner View Post
Chalsall monitors the assignment rates and can give a much better picture than I.
I just use "Let GPU72 decide" in MISFIT and I don't have to worry about changing requirements.
Chuck is offline   Reply With Quote
Old 2013-08-13, 18:03   #835
garo
 
garo's Avatar
 
Aug 2002
Termonfeckin, IE

22·691 Posts
Default

This page is what you are looking for:
http://www.mersenne.ca/cudalucas.php...ranularity=0.1

With a GTX570, you can start at 74 bits at 51M. But we simply do not have that kind of firepower.
garo is offline   Reply With Quote
Old 2013-08-13, 19:11   #836
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

2·3·1,693 Posts
Default

Quote:
Originally Posted by garo View Post
This page is what you are looking for:
http://www.mersenne.ca/cudalucas.php...ranularity=0.1

With a GTX570, you can start at 74 bits at 51M. But we simply do not have that kind of firepower.
That's exactly the graph I was thinking of. Thanks very much, Garo!

Last fiddled with by kladner on 2013-08-13 at 19:11
kladner is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Newer X64 build needed Googulator Msieve 73 2020-08-30 07:47
Performance of cuda-ecm on newer hardware? fivemack GMP-ECM 14 2015-02-12 20:10
Cause this don't belong in the milestone thread bcp19 Data 30 2012-09-08 15:09
Newer msieves are slow on Core i7 mklasson Msieve 9 2009-02-18 12:58
Use of large memory pages possible with newer linux kernels Dresdenboy Software 3 2003-12-08 14:47

All times are UTC. The time now is 23:31.


Fri Aug 6 23:31:25 UTC 2021 up 14 days, 18 hrs, 1 user, load averages: 3.44, 3.75, 3.90

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.