Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants


Leader Boards



1) Message boards : Problems and Help : PrimeGrid and SETI@Home CUDA scheduling issues (Message 12344)
Posted 4593 days ago by akappner
I am currently running PrimeGrid on a PC with 2 processor cores and 2 GPUs. What I would like to do is letting PrimeGrid run on each of the CPU cores while SETI@home (which has a GPU-only client) runs on the graphics cards. However, this does not work: One PrimeGrid thread always shows as "deferred" in the BOINC manager while the 2 SETI threads and one PG thread continue to run normally. Then, after some time, it stops both SETI threads and resumes the paused PG thread. To me, this behaviour seems highly illogical, since PG and SETI do not compete for resources. Is there any reason for this or might this be a bug in BOINC?

Thanks in advance
- Alex
2) Message boards : Number crunching : Prime which was already known found - why? (Message 12040)
Posted 4608 days ago by akappner
Thanks John, I must have missed that message.
I'll be looking forward to searching for real new primes with PrimeGrid in 2009!

Best regards
- Alexander
3) Message boards : Number crunching : Prime which was already known found - why? (Message 12035)
Posted 4608 days ago by akappner
I recently discovered that my computer had computed that 245*2^441021+1 was prime. ( As this prime seemed large enough to be in the Prime Pages, I looked it up there ( To my surprise, I found that this prime had already been reported a year ago. Now, my question is, why does PrimeGrid test numbers which are already known to be prime?

4) Message boards : General discussion : Prime search (Message 11503)
Posted 4641 days ago by akappner
1st is correct.
Regarding the 2nd problem: That method has been known for a long time and is called sieve of Eratosthenes when starting at 1.
I don't know the answer to the 3rd question, although I believe this is what the Primegen project was about.
I believe that the point raised in 4 is also correct, however, of what use is this to you when trying to find new large primes?

I am no mathematician, therefor I cannot say with certainity that this is correct. However, these are logical conclusions: If a composite number x had a factor y larger than sqrt(x), its remainder (x/y) would again be smaller than its square root, which would therefor be found first when doing a trial division. It follows that if we test all possible remainders by doing trial division up to sqrt(x), we must find y if it exists.
5) Message boards : General discussion : US Navy Team has selected Prime Grid as 'Project of the Month". (Message 11218)
Posted 4662 days ago by akappner
Sieving usually gets done 1-2 orders of magnitude faster than Lucas-Lehmer tests (LLR). I copied the following information from another thread (can't remember whom I took it from to give the person credits) - your actual results may vary, but the time-scale should be about the same:

TPS : 12 minutes
PSP LLR : 57 hours
32-1 LLR: 20 hours
Cullen : 45 hours
Woodall : 52 hours
Sieve GCW : 45 minutes
Sieve PSP : 22 minutes
6) Message boards : Number crunching : Setting up multiple different machines for PG (Message 10706)
Posted 4697 days ago by akappner
Problem solved - thanks a lot!

- Alexander
7) Message boards : Number crunching : Setting up multiple different machines for PG (Message 10704)
Posted 4697 days ago by akappner
I am working on setting up BOINC and PG on all computers I am administrating. While doing this, the following questions came up:
- The PCs I am administrating are of different processing power and uptime. While I would like the fast, high-uptime PCs to crunch work units which take long (i.e. Lucas-Lehmer tests), I want the ones with less uptime to do sieving (because they would not be able to complete the large work units in time). However, I found no way to tell PG which work units to assign specific PCs (I can only set preferences for all PCs on my account). What should I do?
- I noticed that if the faster multi-core PCs get to do sieving, they complete a work unit every 20 seconds or so. However, they seem to be uploading the work unit immediately, without collecting a few and uploading them in one bunch, which means my network is needlessly wasting the PG servers' bandwidth... is that expected behaviour or should I manually override that in the BOINC settings?

Thanks for your time

[Return to PrimeGrid main page]
Copyright © 2005 - 2021 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 2.43, 3.04, 2.99
Generated 27 Jul 2021 | 12:24:09 UTC