Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
General discussion :
Some questions from a newbie
| Author |
Message |
|---|
|
|
|
Hey :) Just started crunching primes recently, and have collected some questions since then that I'm hoping some of you veterans can answer-
1. First of all, why are so many sub-projects "cpu only"? Are the computations involved not suitable for GPUs, or is it just that there isn't a program available to do those task with a GPU?
2. Is it possible to set "number of threads" independently for each sub-project? Some seem to recommend multiple threads, and some just one, but the only configuration I can find applies to all at once, which makes it awkward if you want to contribute to many of them.
3. Is it reasonable to estimate the "chance of finding a prime" from the stats here (https://www.primegrid.com/stats_genefer.php) by simply taking (completed workunits)/(primes found). (do the numbers include proof tasks?)
Thanks! | |
|
Honza Volunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,420,056,055
RAC: 2,653,228
                                      
|
|
Hi :-)
All good question and basically general answer would be "yes" to all.
Ad 2 - for me, it makes (more) sense if one comupter is doing one sub-project and different one doing something else.
Why?
Some CPU are doing better on short task, some on longer.
See recent How to optimize the number of threads form example.
Why not?
Running multiple kind of task on single computer would be probably less efficient and frankly, single compuetr has minicrule change of finding primes on several subprojects.
With single/small number of computers, I would run one subproject to find a primes, then contunie to more demanding one.
Ad 3 - basically, larger the candidate, smaller the chance of finding a prime.
See also Is there a 'probability of finding a prime' listed anywhere? for examples and formula.
____________
My stats | |
|
 mikey
Send message
Joined: 17 Mar 09
Posts: 1907
ID: 37043
Credit: 831,688,438
RAC: 818,552
                     
|
Hey :) Just started crunching primes recently, and have collected some questions since then that I'm hoping some of you veterans can answer-
2. Is it possible to set "number of threads" independently for each sub-project? Some seem to recommend multiple threads, and some just one, but the only configuration I can find applies to all at once, which makes it awkward if you want to contribute to many of them.
Thanks!
Yes just use the 'venues' ie default or blank, home, school, work, and then all the Planets if you have enough pc's and want to have each pc running a different sub-project. BUT as was said before focusing on one sub-project at a time is much easier and more likely for you to find a prime number. Especially since some of the cpu only sub-projects can take 150K seconds, or more, to run a single task.
It also depends on how much difference there is between your pc's, if they are close in how long they take to crunch a task then putting them in the same venue makes alot of sense, but if you have ie a dual core and an 16 core pc it might more sense to split them to get the most efficiency out of each of them. It's alot of trial and error but some of the long time crunchers here can help you with figuring out how to optimize your crunching. | |
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 843
ID: 164101
Credit: 306,554,426
RAC: 5,458
|
3. Is it reasonable to estimate the "chance of finding a prime" from the stats here (https://www.primegrid.com/stats_genefer.php) by simply taking (completed workunits)/(primes found). (do the numbers include proof tasks?)
Yes but a mathematical result can also be applied: the expected number of WUs is e-γ log N / log pmax,
where γ is Euler's constant, N the tested number and pmax the sieve limit that can be found here for GFN.
For the GFN in progress, we have
15 8,100
16 15,400
17 29,600
18 52,000
19 97,000
20 181,000
21 335,000
22 609,000
DYFL 673,000 Note that the likelihood of success is 50% for log 2 ~ 0.693147 times the expected number of WUs.
The number of GFN-17 primes is about #WUs / 29,600 but less than 20,500 tasks are needed for half of the participants.
That's because few unlucky searchers need more than 50,000 or even 100,000 tasks. | |
|
mikey
Send message
Joined: 17 Mar 09
Posts: 1907
ID: 37043
Credit: 831,688,438
RAC: 818,552
|
For the GFN in progress, we have
DYFL 673,000
That's ALOT of DYFL tasks in progress!!! | |
|
compositeVolunteer tester
Send message
Joined: 16 Feb 10
Posts: 1172
ID: 55391
Credit: 1,221,586,795
RAC: 1,513,610
              
|
For the GFN in progress, we have
DYFL 673,000
That's ALOT of DYFL tasks in progress!!!
I think you misunderstood.
The "GFN in progress" means the kinds of subprojects available to work on.
The table shows the number of workunits expected to be processed before finding a prime in that subproject. | |
|
|
|
|
I agree with most of what is said. That in mind, there are a couple of subprojects that do work OK together with the same number of threads, because the sizes of the work are similar. One examples of this is CUL and WOO.
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1
My movie https://vimeo.com/manage/videos/502242 | |
|
Post to thread
Message boards :
General discussion :
Some questions from a newbie |
