## Other

drummers-lowrise

Message boards : Seventeen or Bust : Seventeen or Bust and the Sierpinski Problem

Author Message
Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20691 - Posted: 31 Jan 2010 | 17:36:45 UTC

Seventeen or Bust

Seventeen or Bust was a distributed computing project attempting to solve the Sierpinski problem. The name of the project is due to the fact that, when founded, there were seventeen values of k < 78,557 for which no primes were known.

The project was conceived in March of 2002 by two college undergraduates. After some planning and a lot of programming, the first public client was released on April 1. Seventeen or Bust ceased operations in 2016. The project was administered by:

• Louis Helm, a computer engineer in Austin, Texas.
• David Norris, a software engineer in Urbana, Illinois.
• Michael Garrison, a Computer Science undergraduate at Eastern Michigan University in Ypsilanti, Michigan.

Starting in 2010, PrimeGrid partnered with Seventeen or Bust to work towards solving the Sierpinski Problem. After the demise of the original Seventeen or Bust project in 2016, PrimeGrid is continuing by itself to continue the Seventeen or Bust project in looking to solve the Sierpinski Problem.

As of October of 2016, PrimeGrid and Seventeen or Bust have eliminated twelve of those seventeen candidates. The project might now be styled "Five or Bust," but the original name will be kept for consistency.

PrimeGrid and Seventeen or Bust's twelve prime discoveries are:

1. 46157*2^698207+1 with 210,186 decimal digits, discovered November 27, 2002. Crunched by Stephen Gibson.
2. 65567*2^1013803+1 with 305,190 decimal digits, discovered December 2, 2002. Crunched by James Burt.
3. 44131*2^995972+1 with 299,823 decimal digits, discovered December 5, 2002. Crunched by an anonymous participant.
4. 69109*2^1157446+1 with 348,431 decimal digits, discovered December 6, 2002. Crunched by Sean DiMichele.
5. 54767*2^1337287+1 with 402,569 decimal digits, discovered December 23, 2002. Crunched by Peter Coels.
6. 5359*2^5054502+1 with 1,521,561 decimal digits, discovered December 6, 2003. Crunched by Randy Sundquist.
7. 28433*2^7830457+1 with 2,357,207 decimal digits, discovered December 30, 2004. Crunched by a member of Team Prime Rib.
8. 27653*2^9167433+1 with 2,759,677 decimal digits, discovered June 8, 2005. Crunched by Derek Gordon.
9. 4847*2^3321063+1 with 999,744 decimal digits, discovered October 15, 2005 while double checking earlier tests. Crunched by Richard Hassler.
10. 19249*2^13018586+1 with 3,918,990 decimal digits, discoverd March 26, 2007. Crunched by Konstantin Agafonov.
11. 33661*2^7031232+1 with 2,116,617 decimal digits, discovered October 17, 2007 while double checking earlier tests. Crunched by Sturle Sunde.
12. 10223*2^31172165+1 with 9,383,761 decimal digits, discovered October 31, 2016. Crunched by Szabolcs PÃ©ter (SyP). This prime eliminated k=10223 from both the Sierpinski Problem and the Prime Sierpinski Problem. (official announcement)

WacÅ‚aw Franciszek SierpiÅ„ski (14 March 1882 â€” 21 October 1969), a Polish mathematician, was known for outstanding contributions to set theory, number theory, theory of functions and topology. It is in number theory where we find the Sierpinski problem.

Basically, the Sierpinski problem is "What is the smallest Sierpinski number?"

First we look at Proth numbers (named after the French mathematician FranÃ§ois Proth). A Proth number is a number of the form k*2^n+1 where k is odd, n is a positive integer, and 2^n>k.

A Sierpinski number is an odd k such that the Proth number k*2^n+1 is not prime for all n. For example, 3 is not a Sierpinski number because n=2 produces a prime number (3*2^2+1=13). In 1962, John Selfridge proved that 78,557 is a Sierpinski number...meaning he showed that for all n, 78557*2^n+1 was not prime.

Most number theorists believe that 78,557 is the smallest Sierpinski number, but it hasn't yet been proven. In order to prove it, it has to be shown that every single k less than 78,557 is not a Sierpinski number, and to do that, some n must be found that makes k*2^n+1 prime.

The smallest proven 'prime' Sierpinski number is 271,129. In order to prove it, it has to be shown that every single 'prime' k less than 271,129 is not a Sierpinski number, and to do that, some n must be found that makes k*2^n+1 prime.

Seventeen or Bust is working on the Sierpinski problem and the Prime Sierpinski Project is working on the 'prime' Sierpinski problem. The following k's remain for each project:

Sierpinski problem 'prime' Sierpinski problem 21181 22699* 22699 67607* 24737 79309 55459 79817 67607 152267 156511 168451 222113 225931 237019
• being tested by Seventeen or Bust

• Fortunately, the two projects (and later PrimeGrid's Extended SIerpinski Project) combined their sieving efforts into a single file. Therefore, PrimeGrid's PSP sieve supports all three projects.

____________
My lucky number is 75898524288+1

rebirther

Joined: 10 Aug 05
Posts: 783
ID: 85
Credit: 175,774,608
RAC: 0

Message 20692 - Posted: 31 Jan 2010 | 18:05:17 UTC

Can we expecting a 64bit application?
____________

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20693 - Posted: 31 Jan 2010 | 18:12:54 UTC - in response to Message 20692.

Can we expecting a 64bit application?

There's no 64 bit LLR application.
____________

Warped

Joined: 25 Aug 08
Posts: 288
ID: 27792
Credit: 28,151,335
RAC: 56

Message 20694 - Posted: 31 Jan 2010 | 18:51:17 UTC

A very interesting development.

One comment I have: beware, these are LONG, with n>17m! Cullen and Woodall tasks are sprints in comparison.

Five questions:
1. We use LLR, whereas SOB have adopted the Prime95 (originally developed for GIMPS). Are the relative speeds similar?
2. Will there be some trickle method of reporting (similar to CPDN)? If not, I expect a lot of computation errors and frustration.
3. Will traditional BOINC credits apply? If so, how will credit be calculated? I expect the first badges will be a while in coming!
4. Have we been allocated a range of n, which SOB will not crunch?
5. Who will get credit if a prime is found?
____________
Warped

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20695 - Posted: 31 Jan 2010 | 19:25:13 UTC - in response to Message 20694.

A very interesting development.

One comment I have: beware, these are LONG, with n>17m! Cullen and Woodall tasks are sprints in comparison.

Five questions:
1. We use LLR, whereas SOB have adopted the Prime95 (originally developed for GIMPS). Are the relative speeds similar?
2. Will there be some trickle method of reporting (similar to CPDN)? If not, I expect a lot of computation errors and frustration.
3. Will traditional BOINC credits apply? If so, how will credit be calculated? I expect the first badges will be a while in coming!
4. Have we been allocated a range of n, which SOB will not crunch?
5. Who will get credit if a prime is found?

Well, having just past the first checkpoint, if the percentage-done number is to be believed, this task is going to take 10 days to crunch. (That's on a C2Q Q6600 @ 2.4GHz)

I'm not personally worried about trickles. CPDN is a weird project, and its credit system works very differently than anyone else's. CPDN's tasks are also MUCH longer than even this SoB project -- the longest CPDN can run up to 100 days even on a fast computer. Furthermore, CPDN is very sensitive to errors because, IMHO, their code is buggy. They like to blame it on widespread hardware errors, but I don't buy it. Their code is problematic. So I'm not really worried about running a WU here for 10 days or so. Then again, I rarely get any kind of computational error, not counting those that can be explained away by something I did. Except for CPDN, but that's their problem, literally.

I've been crunching SETI stuff since 1999, and considering how comparatively slow the computers were then, WUs running for days and days was not only the norm, all of the tasks ran for that long. Problems with errors tend to have their roots in either buggy code (which most projects fix), or some kind of hardware problem (usually excessive over-clocking, heat, or the power supply). If you're not seeing errors now, I don't expect you'll see errors with these new tasks, since it appears we're running the same executable image as with the PPS LLR tasks.

I wouldn't worry about trickles until the long WUs actually prove to be problematic.

____________
My lucky number is 75898524288+1

Warped

Joined: 25 Aug 08
Posts: 288
ID: 27792
Credit: 28,151,335
RAC: 56

Message 20697 - Posted: 31 Jan 2010 | 19:57:04 UTC - in response to Message 20695.

Thanks Michael

10 days is quite quick - it looks like that's an improvement on the existing SOB software. Yes, agreed, if it's that short then stability, and for that matter, trickles and BOINC credit calculations, are not major issues either.

Also, I see that my questions 4 and 5 have been answered on the updated SOB home page.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20698 - Posted: 31 Jan 2010 | 19:57:07 UTC - in response to Message 20695.

Talking about durations, this task came with a 72 hour estimate (3 days) and it looks like it will actually take 10 days crunching full time. It has a 21 day deadline. I reset the project prior to downloading this WU, so the DCF is 1.000.

It's running on a Core2 CPU. While that's no longer that fastest CPU, it's far from the slowest. Any Intel chip prior to Core2 is a lot slower, as are all but the most recent AMD processors.

21 days may be too short a deadline. A Pentium 4 or slower computer and/or any computer not crunching one of these 24/7 will likely miss the deadline.
____________
My lucky number is 75898524288+1

Vato
Volunteer tester

Joined: 2 Feb 08
Posts: 785
ID: 18447
Credit: 262,879,636
RAC: 801,885

Message 20699 - Posted: 31 Jan 2010 | 20:48:44 UTC - in response to Message 20694.

3. Will traditional BOINC credits apply? If so, how will credit be calculated? I expect the first badges will be a while in coming!

a) roughly what credit one of these WUs will give?
b) what the maximum limit for credit per WU that is configured in the boinc server? (don't want this to be too low!)
c) that this uses the usual badge targets for an LLR project?

My guess is that we should be plausibly >5k credits per WU, and therefore it should be feasible to get bronze in just a couple of weeks or so :-)
____________

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20701 - Posted: 31 Jan 2010 | 21:21:05 UTC - in response to Message 20699.

a) roughly what credit one of these WUs will give?
b) what the maximum limit for credit per WU that is configured in the boinc server? (don't want this to be too low!)

Unknown at this time. We'll wait for actual completion times.

c) that this uses the usual badge targets for an LLR project?

Yes, they will remain the same. There's no reason to change targets.
____________

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20702 - Posted: 31 Jan 2010 | 21:25:23 UTC - in response to Message 20694.

1. We use LLR, whereas SOB have adopted the Prime95 (originally developed for GIMPS). Are the relative speeds similar?

I'm not aware of any comparisons.

2. Will there be some trickle method of reporting (similar to CPDN)? If not, I expect a lot of computation errors and frustration.

No, not at this time.

3. Will traditional BOINC credits apply? If so, how will credit be calculated? I expect the first badges will be a while in coming!

Yes, traditional BOINC credits apply and they'll be calculated the same as other LLR projects.
____________

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20703 - Posted: 31 Jan 2010 | 21:56:35 UTC - in response to Message 20699.

My guess is that we should be plausibly >5k credits per WU, and therefore it should be feasible to get bronze in just a couple of weeks or so :-)

5K looks like a good guestimate given what I'm seeing, if credit is granted using the same formulas as the PPS LLR. So two of these tasks (at most 3) should get the bronze badge. 5 days on a fast i7, using only a fraction of the cores.

The biggest problem might be pending credit waiting for wingmen -- a lot of whom might potentially miss the deadline. I'm not sure I've crunched recently for any project with tasks this big that require validating against wingmen. I'm guessing a fair amount of patience will be required on our part waiting for credit to be granted. Especially if it's decided that the deadlines need to be increased, which I'm guessing will happen at some point.
____________
My lucky number is 75898524288+1

SiKTheGreatOne
Volunteer tester

Joined: 20 Mar 08
Posts: 110
ID: 20411
Credit: 101,493,682
RAC: 0

Message 20708 - Posted: 1 Feb 2010 | 1:00:41 UTC - in response to Message 20698.

21 days may be too short a deadline. A Pentium 4 or slower computer and/or any computer not crunching one of these 24/7 will likely miss the deadline.

Just downloaded one WU for my Pentium 4. It hasn't started yet, but the estimate to completion is 1267 hours which is almost 53 days. I'm going to run it a bit and then form my own estimate.
____________

StefanoD
Volunteer tester

Joined: 11 May 07
Posts: 32
ID: 8194
Credit: 10,790,073
RAC: 0

Message 20711 - Posted: 1 Feb 2010 | 8:18:35 UTC

Core2 Duo @2.4, MacOS 10.6

0,492% in an hour (~8.5 days to 100%)

Lumiukko
Volunteer tester

Joined: 7 Jul 08
Posts: 164
ID: 25183
Credit: 712,338,240
RAC: 111,046

Message 20714 - Posted: 1 Feb 2010 | 9:59:51 UTC

6.05% in 11:00 hours (~7.6 days total)

--
Lumiukko

imsaguy

Joined: 21 Oct 07
Posts: 85
ID: 13765
Credit: 11,844,028
RAC: 0

Message 20717 - Posted: 1 Feb 2010 | 13:30:40 UTC

I'd just like to say welcome to the 'fight'. My credit on PG has been lacking because I've been crunching on SoB. :-x

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20718 - Posted: 1 Feb 2010 | 14:00:59 UTC

Yesterday I downloaded one SoB WU. It's prefix is psp_sub_, it came with a deadline of 21 days, an estimate of 72 hours and actual run time appears to be about 200 hours.

Today, the other work on the quad core is clearing out, so I enabled new tasks and downloaded 3 more WUs. These are different.

The prefix is llr_sob_, the estimate is 290 hours, but the deadline is still the same 21 days. Is this just a change in the estimation, or can I expect this to actually run 4 times as long as yesterday's psp_sob_? If it really runs four times as long, I'll miss the deadline by at least 10 days. That's on a Core2Quad Q6600 @2.4 GHz.

I don't have any progress bar information yet so I don't have an estimate on the real runtime.
____________
My lucky number is 75898524288+1

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20719 - Posted: 1 Feb 2010 | 14:31:20 UTC - in response to Message 20718.

I don't have any progress bar information yet so I don't have an estimate on the real runtime.

Now I do.

Actual run time seems to be the same. It should take about 200 CPU hours (which is probably closer to 10 days in real time on my system). Same as the earlier WU.
____________
My lucky number is 75898524288+1

StefanoD
Volunteer tester

Joined: 11 May 07
Posts: 32
ID: 8194
Credit: 10,790,073
RAC: 0

Message 20720 - Posted: 1 Feb 2010 | 14:41:26 UTC

i7 Extreme 975 @3.33 - Win 7
Boinc wit 4 cores and 2 GPU in use:
0,650% after 62 min (~ 6.6 days total)

With 6 cores the total time is ~ 8.3 days

StefanoD

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20721 - Posted: 1 Feb 2010 | 14:44:29 UTC - in response to Message 20718.

Yesterday I downloaded one SoB WU. It's prefix is psp_sob_, it came with a deadline of 21 days, an estimate of 72 hours and actual run time appears to be about 200 hours.

WU's are the same, just the name was corrected. The first WU's went out with a minor copy/paste error. :)
____________

SiKTheGreatOne
Volunteer tester

Joined: 20 Mar 08
Posts: 110
ID: 20411
Credit: 101,493,682
RAC: 0

Message 20723 - Posted: 1 Feb 2010 | 15:29:38 UTC

My Pentium 4 has been running the WU for 12 hours and 1.969% is completed. At that rate the WU will finish in 25 days. Would I get credit after returning the WU four days after the deadline? I would be betting the average person can't return it within four days after my WU will expire.
____________

Rytis
Volunteer moderator

Joined: 22 Jun 05
Posts: 2649
ID: 1
Credit: 26,363,112
RAC: 34,878

Message 20724 - Posted: 1 Feb 2010 | 15:39:20 UTC

Credit will be granted for WUs past deadline, unless there are two matching results returned before. In that case you have a limited time window to return late work (at the moment, 5 days).
____________

Danial L Puckett

Joined: 5 Aug 09
Posts: 66
ID: 44528
Credit: 7,755,477
RAC: 0

Message 20725 - Posted: 1 Feb 2010 | 15:49:11 UTC

i7 Extreme 975 @3.33 - Overclocked to 4.06 on linux
with 4 cores (Hyperthread off) after 20hrs 20.875%
Should be Just Under 4 days total
Running around the clock on this one!
Maybe I will be the first to finish one of these guys!
____________

End Transmission!

opyrt

Joined: 13 Nov 09
Posts: 17
ID: 50074
Credit: 9,147,038
RAC: 0

Message 20744 - Posted: 1 Feb 2010 | 23:01:29 UTC - in response to Message 20694.

1. We use LLR, whereas SOB have adopted the Prime95 (originally developed for GIMPS). Are the relative speeds similar?

From what I understand, the speeds should be pretty much the same, I've asked around in different forums, and the general answer is "they're about the same". If I understand correctly, they're based on the same libraries aswell.

Only major difference I've heard is: Proth tests (what LLR does) gives a conclusive primality proof, whereas a PRP tests (what Prime95 does) does not.

Hope this helps! :)

Mary

Joined: 23 Oct 08
Posts: 25
ID: 30801
Credit: 5,610,096
RAC: 0

Message 20768 - Posted: 2 Feb 2010 | 14:56:13 UTC

Okay time to throw my results so far into the mix:

C2Q Q6600 @ 2.4GHz Vista HP 64
Current results: 6.21% at 11:17 hours in
Estimated time to completion at current speed ~ 182 hours (aka ~7.6 days)
Running concurrently with other boinc projects, but no other Primegrid LLR workunits
____________
~It only takes one bottle cap moving at 23,000 mph to ruin your whole day~

Kevin D Puckett
Volunteer tester

Joined: 4 Aug 09
Posts: 61
ID: 44488
Credit: 5,675,896
RAC: 0

Message 20805 - Posted: 3 Feb 2010 | 22:46:17 UTC

OK, so I'm running a Core2 Duo P8600 at 2.40GHz on Ubuntu Linux, both cores running Seventeen or Bust.
Both jobs are 46 hours in with 22.832% complete. Prediction: 8.39 days total.

My sister (who has a comparable computer, same OS) is estimating 8.0 days, but is running 1 SoB, 1Woodal. My guess is I'm running a hare slower because both of my cores are on Seventeen or Bust?
____________
May the Force be with you always.

Danial L Puckett

Joined: 5 Aug 09
Posts: 66
ID: 44528
Credit: 7,755,477
RAC: 0

Message 20823 - Posted: 4 Feb 2010 | 19:51:59 UTC - in response to Message 20725.

i7 Extreme 975 @3.33 - Overclocked to 4.06 on linux
with 4 cores (Hyperthread off) after 20hrs 20.875%
Should be Just Under 4 days total
Running around the clock on this one!
Maybe I will be the first to finish one of these guys!

Completed my first 4 SoBs in 10 to 15 minutes under 4 days!
____________

End Transmission!

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20824 - Posted: 4 Feb 2010 | 20:07:26 UTC - in response to Message 20823.

Completed my first 4 SoBs in 10 to 15 minutes under 4 days!

I am SO jealous. I'm at 34.5% on the WU that's furthest along.

Not only is your machine a whole lot faster than mine, but I'm seeing about a 30% slowdown because I'm running 4 at once (but then, so were you), and there's about a 15% difference between CPU time and elapsed time due to other stuff running on the computer. When I was running just one, I was looking at about 10 days -- now it's looking more like 12.

____________
My lucky number is 75898524288+1

Vato
Volunteer tester

Joined: 2 Feb 08
Posts: 785
ID: 18447
Credit: 262,879,636
RAC: 801,885

Message 20825 - Posted: 4 Feb 2010 | 20:37:12 UTC - in response to Message 20824.

That is fast - I've only got 2 @ 57.5% done in 97 hours.

A key point here is that there should shortly be some real SoB data for Bok (and possibly others) to work with for the subproject stats - subject to the wingman lottery :-)
____________

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20826 - Posted: 4 Feb 2010 | 20:39:47 UTC - in response to Message 20823.

i7 Extreme 975 @3.33 - Overclocked to 4.06 on linux
with 4 cores (Hyperthread off) after 20hrs 20.875%
Should be Just Under 4 days total
Running around the clock on this one!
Maybe I will be the first to finish one of these guys!

Completed my first 4 SoBs in 10 to 15 minutes under 4 days!

Congratulations! Your 4 WU's are the first to come in. Now we must wait to see if they are valid. ;)

It's definitely worth investigating why the i7 Extreme 975 did not experience the "slowdown" on multiple cores.
____________

opyrt

Joined: 13 Nov 09
Posts: 17
ID: 50074
Credit: 9,147,038
RAC: 0

Message 20827 - Posted: 4 Feb 2010 | 20:43:28 UTC - in response to Message 20826.

It's definitely worth investigating why the i7 Extreme 975 did not experience the "slowdown" on multiple cores.

A lot of memory stuff is redesigned on the i7, so my guess is that the bus is able to handle the demanding llr applications better than previous architectures. :)

Danial L Puckett

Joined: 5 Aug 09
Posts: 66
ID: 44528
Credit: 7,755,477
RAC: 0

Message 20828 - Posted: 4 Feb 2010 | 20:55:24 UTC - in response to Message 20827.

It's definitely worth investigating why the i7 Extreme 975 did not experience the "slowdown" on multiple cores.

A lot of memory stuff is redesigned on the i7, so my guess is that the bus is able to handle the demanding llr applications better than previous architectures. :)

I will see if running one SoB and some sieves increases the throughput for the SoB. May wait till after the challenge because my whole team is running SoBs and our team credit is getting low.
____________

End Transmission!

enderak

Joined: 13 Dec 08
Posts: 45
ID: 32842
Credit: 8,789,369
RAC: 0

Message 20829 - Posted: 4 Feb 2010 | 22:27:28 UTC

So it sounds like the speeds are comparable to Prime95/mprime, at least for the i7's. My non-overclocked i7 Extreme 975 is able to complete about 4 tests every 4 days (maybe a hair over 4 days) using Prime95.

Congrats on the first returns!

Tanya Puckett

Joined: 11 Aug 09
Posts: 19
ID: 44850
Credit: 5,549,656
RAC: 0

Message 20831 - Posted: 5 Feb 2010 | 2:19:19 UTC

I have a core 2 Duo T7700 running at 2.40GHz.
I am running two SoBs which are now at 27.7%, and I calculate a total time of 8.0 days.

Running one core, in the same amount of time,
I could earn 5197 credits if I ran Woodalls, which are the least efficient of any jobs,
and 6803 credits if I ran 321s, which are the most efficient of the LLRs.

If an SOB only earned 5000 credits, it would replace Woodals as the least efficient job for earning credit.
Since SOB's are long jobs, and it may be a long time before before credit is awarded, I suggest that 5000 credits may be a little low.

____________

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20835 - Posted: 5 Feb 2010 | 15:02:24 UTC

fyi...for those comparing times, only Linux has the updated LLR client. Mac and Windows are still running 3.7.1c. We hope to have those updated soon.
____________

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20837 - Posted: 5 Feb 2010 | 15:38:19 UTC - in response to Message 20835.

fyi...for those comparing times, only Linux has the updated LLR client. Mac and Windows are still running 3.7.1c. We hope to have those updated soon.

What is the difference between the old and new clients?
____________
My lucky number is 75898524288+1

Vato
Volunteer tester

Joined: 2 Feb 08
Posts: 785
ID: 18447
Credit: 262,879,636
RAC: 801,885

Message 20838 - Posted: 5 Feb 2010 | 16:44:53 UTC - in response to Message 20835.

My linux box is still showing LLR is 3.7.1c under the 6.0.5 wrapper.
Is this really much different to the Windows app?

____________

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20842 - Posted: 6 Feb 2010 | 0:48:06 UTC - in response to Message 20837.

Michael Goetz wrote:
fyi...for those comparing times, only Linux has the updated LLR client. Mac and Windows are still running 3.7.1c. We hope to have those updated soon.

What is the difference between the old and new clients?

New one is faster. :) And it varies based on hardware. Documentation is not available right now. Once it is, I'll post it. Right now, we're just testing the new release in Linux.

Vato wrote:
My linux box is still showing LLR is 3.7.1c under the 6.0.5 wrapper.
Is this really much different to the Windows app?

This is clerical. LLR under the 6.0.5 wrapper is the new one. Currently, only SoB and PPS have it for testing.
____________

Vato
Volunteer tester

Joined: 2 Feb 08
Posts: 785
ID: 18447
Credit: 262,879,636
RAC: 801,885

Message 20845 - Posted: 6 Feb 2010 | 13:06:14 UTC - in response to Message 20842.

Ok - would have been nice to have known, since this is probably the cause of a problem i've recently started seeing. Orderly shutdown of boinc client under Linux -> next boot -> PPS LLR WUs start from the beginning, not from the save point. I'm certain that this wasn't the case previously (though 5.11 on windows does the same). I hope that isn't true of SoB as well, as that could be a tremendous amount of work lost.
____________

Vato
Volunteer tester

Joined: 2 Feb 08
Posts: 785
ID: 18447
Credit: 262,879,636
RAC: 801,885

Message 20859 - Posted: 7 Feb 2010 | 20:51:05 UTC - in response to Message 20845.

my 2 are done and waiting for wingman in just under 7 days on a E5300@2.87GHz

http://www.primegrid.com/results.php?hostid=113117&offset=0&show_names=0&state=0&appid=13
____________

Lumiukko
Volunteer tester

Joined: 7 Jul 08
Posts: 164
ID: 25183
Credit: 712,338,240
RAC: 111,046

Message 20868 - Posted: 8 Feb 2010 | 13:31:56 UTC

My 2 SoB's are also done (in ~7,5 days) on a

http://www.primegrid.com/results.php?hostid=85145&offset=0&show_names=0&state=0&appid=13

--
Lumiukko

John
Honorary cruncher

Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0

Message 20884 - Posted: 8 Feb 2010 | 23:08:38 UTC - in response to Message 20842.

Michael Goetz wrote:
fyi...for those comparing times, only Linux has the updated LLR client. Mac and Windows are still running 3.7.1c. We hope to have those updated soon.

What is the difference between the old and new clients?

New one is faster. :) And it varies based on hardware. Documentation is not available right now. Once it is, I'll post it. Right now, we're just testing the new release in Linux.

LLR 3.8.0 has been officially released. http://jpenne.free.fr/index2.html

We hope to have all 3 builds (Linux, MacIntel, & Windows) updated and tested soon. :)
____________

thommy3

Joined: 7 Jan 08
Posts: 42
ID: 17265
Credit: 268,651
RAC: 0

Message 20896 - Posted: 9 Feb 2010 | 15:14:38 UTC

http://www.mersenneforum.org/showpost.php?p=205041

LLR3.8 does say that 2072644824759 * 2^33333+5 is not prime. Which it is.

Jack

Joined: 26 Nov 08
Posts: 76
ID: 32168
Credit: 11,976,778
RAC: 0

Message 20915 - Posted: 11 Feb 2010 | 0:16:09 UTC

First two done in 146 hours and some change. One will be done within the hour and the last is at 140 hours and still says 39 hours to go..77% done.
Phenom II 955 stock clocks.
These are pretty long , any been validated yet?
Jack

Bok
Volunteer developer

Joined: 13 Jul 05
Posts: 133
ID: 60
Credit: 22,627,896
RAC: 2,195

Message 20921 - Posted: 11 Feb 2010 | 14:42:14 UTC - in response to Message 20915.

I don't believe any are validated yet. I'm looking in the subproject xml feed and see no evidence of anyone having any credits as of yet.

I finished my first wu overnight and have 3 more which should finish today too..

Bok
____________

Bok
Volunteer developer

Joined: 13 Jul 05
Posts: 133
ID: 60
Credit: 22,627,896
RAC: 2,195

Message 20922 - Posted: 11 Feb 2010 | 15:23:12 UTC - in response to Message 20921.

2nd one finished and this one has a wingman... hopefully will be validated soon!

http://www.primegrid.com/workunit.php?wuid=103653501
____________

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 20981 - Posted: 13 Feb 2010 | 23:08:09 UTC - in response to Message 20699.

a) roughly what credit one of these WUs will give?
b) what the maximum limit for credit per WU that is configured in the boinc server? (don't want this to be too low!)

My guess is that we should be plausibly >5k credits per WU, and therefore it should be feasible to get bronze in just a couple of weeks or so :-)

If an SOB only earned 5000 credits, it would replace Woodals as the least efficient job for earning credit.
Since SOB's are long jobs, and it may be a long time before before credit is awarded, I suggest that 5000 credits may be a little low.

I've been fairly lucky with my wingmen and have received credit for three WUs so far, at approximately 5600, 6600, and 8600 credits. That averages out to a bit more than 6900 per WU.
____________
My lucky number is 75898524288+1

Danial L Puckett

Joined: 5 Aug 09
Posts: 66
ID: 44528
Credit: 7,755,477
RAC: 0

Message 20983 - Posted: 13 Feb 2010 | 23:23:38 UTC

I have had two of my four SoBs clear with credits of 6500 and 8100.
____________

End Transmission!

Rytis
Volunteer moderator

Joined: 22 Jun 05
Posts: 2649
ID: 1
Credit: 26,363,112
RAC: 34,878

Message 20986 - Posted: 14 Feb 2010 | 9:28:00 UTC

Please note that we are adjusting granted credit levels, as a consequence you can't really compare the workunits granted until now (the credit amount has been raised from the first WUs validated).
____________

TroubledBunny
Volunteer tester

Joined: 4 Jun 07
Posts: 607
ID: 8931
Credit: 470,669,737
RAC: 223

Message 21349 - Posted: 24 Feb 2010 | 14:28:49 UTC - in response to Message 20986.

I came across the AQUA project and found this long running WU, but was shocked at the credit granted.
Comparing it with the runtime of my only completed SoB WU I feel robbed - what have I missed? There has to be a reason for such a massive difference. Perhaps the AQUA WU uses several cores at the same time to process it? Perhaps it's a special - like the manual sieve efforts at PrimeGrid? Whatever the answer, I'm feeling a little green with envy.

[AF>HFR>RR] Jim PROFIT

Joined: 23 Feb 08
Posts: 5
ID: 19363
Credit: 126,283,810
RAC: 2,132

Message 21352 - Posted: 24 Feb 2010 | 15:47:03 UTC - in response to Message 21349.

This explaining the granted credits, but the application is also highly optimized.

____________

SiKTheGreatOne
Volunteer tester

Joined: 20 Mar 08
Posts: 110
ID: 20411
Credit: 101,493,682
RAC: 0

Message 21364 - Posted: 25 Feb 2010 | 5:16:28 UTC

http://www.primegrid.com/workunit.php?wuid=103402874

Finally finished the SoB WU on my P4. I have HT activated on it and ran Woodall WUs alongside the SoB WU. Took 2,065,246.69 CPU seconds or just under 24 days.
____________

denim

Joined: 22 Sep 08
Posts: 59
ID: 29415
Credit: 524,789,992
RAC: 5,598

Message 21391 - Posted: 26 Feb 2010 | 22:43:07 UTC

On my Vista box that is an AMD Athlon(tm) 64 X2 Dual Core Processor 5400+,

I currently have two SoB LLR's running and they are currently reading:

1. 53:xx:xx hours 17%+ complete
2. 43:xx:xx hours 14%+ complete

So I hope they get done in decent time.

____________

enderak

Joined: 13 Dec 08
Posts: 45
ID: 32842
Credit: 8,789,369
RAC: 0

Message 21394 - Posted: 27 Feb 2010 | 1:08:52 UTC - in response to Message 21391.

Well, that's just under 13 days for two tests, or ~5.5 days per test. That's a pretty decent time for these large tests - just hang in there and be patient. Your two tests alone should net you a bronze badge once they are verified. ;)

enderak

Joined: 13 Dec 08
Posts: 45
ID: 32842
Credit: 8,789,369
RAC: 0

Message 21396 - Posted: 27 Feb 2010 | 6:19:21 UTC - in response to Message 21394.

Oops, just realized I should have said 6.5 days, not 5.5 - still a decent time though.

JLConawayII

Joined: 12 May 10
Posts: 18
ID: 60358
Credit: 7,028,051
RAC: 0

Message 23919 - Posted: 20 May 2010 | 17:27:18 UTC

I'm going to need more computers...

[SG]Puzzle-Peter
Volunteer tester

Joined: 14 Jun 08
Posts: 374
ID: 24128
Credit: 92,652,187
RAC: 0

Message 29134 - Posted: 11 Dec 2010 | 10:36:50 UTC

Just out of curiosity: what's the sieving ratio for SoB? I mean, at the start, for any k a 1M range for n contains 1M candidates. Which percentage was left after sieving? Or in other words, how many tests cover a 1M range for n on a given k?
____________
There are only 10 kinds of people - those who understand binary and those who don't

ltd

Joined: 30 Sep 07
Posts: 10
ID: 12893
Credit: 18,034,132
RAC: 0

Message 29151 - Posted: 11 Dec 2010 | 18:50:54 UTC

Here some data from the PSP database. It is the range 10M-11M.

k=79309 992 tests
k=79817 2271 tests
k=152267 1411 tests
k=156511 917 tests
k=168451 1592 tests
k=222113 4196 tests
k=225931 1978 tests
k=237019 2438 tests

[SG]Puzzle-Peter
Volunteer tester

Joined: 14 Jun 08
Posts: 374
ID: 24128
Credit: 92,652,187
RAC: 0

Message 30657 - Posted: 9 Jan 2011 | 7:10:34 UTC - in response to Message 29151.

Here some data from the PSP database. It is the range 10M-11M.

k=79309 992 tests
k=79817 2271 tests
k=152267 1411 tests
k=156511 917 tests
k=168451 1592 tests
k=222113 4196 tests
k=225931 1978 tests
k=237019 2438 tests

Good idea to look there, thanks!

BTW all tasks I get seem to be either n<19M or n>20M. What's up with 19M<n<20M?
____________
There are only 10 kinds of people - those who understand binary and those who don't

pschoefer
Volunteer developer
Volunteer tester

Joined: 20 Sep 05
Posts: 662
ID: 845
Credit: 2,220,370,221
RAC: 923,255

Message 30660 - Posted: 9 Jan 2011 | 7:54:37 UTC - in response to Message 30657.

BTW all tasks I get seem to be either n<19M or n>20M. What's up with 19M<n<20M?

Looks like this range is done by the native SoB project right now: SoB Test Range Statistics.
____________

T-Armstrong

Joined: 3 Aug 10
Posts: 270
ID: 64825
Credit: 1,013,734,308
RAC: 1,098,877

Message 40361 - Posted: 17 Sep 2011 | 1:57:38 UTC - in response to Message 20691.

I can give all Tasks back, the time is to slow, I need 70 Days, and I cant crunch it in 45 Days, thats crazy !!!

If I am "out of time" then no credits,

crazy, crazy,crazy !!!

cheers
____________

mfbabb2
Volunteer tester

Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 11,549,356
RAC: 370

Message 40362 - Posted: 17 Sep 2011 | 2:39:35 UTC - in response to Message 40361.

I can give all Tasks back, the time is to slow, I need 70 Days, and I cant crunch it in 45 Days, thats crazy !!!

If I am "out of time" then no credits,

crazy, crazy,crazy !!!

cheers

What are you using for a CPU? Even my slow machines will do an SoB in about 3 weeks.
____________
Murphy (AtP)

rroonnaalldd
Volunteer developer
Volunteer tester

Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0

Message 40371 - Posted: 17 Sep 2011 | 12:00:02 UTC - in response to Message 40362.

What are you using for a CPU? Even my slow machines will do an SoB in about 3 weeks.

Either he uses an ATOM-cpu or he is doing calculations only on weekends...
Both would be possible with hidden hosts.
____________
Best wishes. Knowledge is power. by jjwhalen

T-Armstrong

Joined: 3 Aug 10
Posts: 270
ID: 64825
Credit: 1,013,734,308
RAC: 1,098,877

Message 40378 - Posted: 17 Sep 2011 | 13:49:36 UTC - in response to Message 40371.

I have no atom CPU, I have i7 @ 1.8 Ghz. I've cleaned with a vacuum cleaner, poured new coolant, now he's running 12% on the day. That's crazy. Also. 0.5% on the day yesterday, today, cleaned, 12% on the day. Now ready in 7 Days, previously in 11 weeks.

I did not know that new coolant makes the machine 10 times faster. One coolant = 50 USD

I have 2 gamer lap-tops with 8 core, and one of the lap-top, I bring him tomorrow to my hardwhere dealer. He do inside the new coolant. The result is: 10 times faster. And not hot.

I think, 1:30 h. for one pps-sieve task is not normal, perhaps 0:15 or 0:30 but not 1:30, I must clean it.

I crunch 24h. every Day. 7 days @ week, every month, every Year. Not one minute without crunch. IÂ´m a true Astra from Sicituradastra. A true Astra, crunch every minute of the life.

I open my Profile. please wait for the new/ next server update.

My first SoB, I see: CPU time is: 1:875:49 h.
Now: 549 h.

6 tasks I crunch now finish and to the end.
____________

mfbabb2
Volunteer tester

Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 11,549,356
RAC: 370

Message 40379 - Posted: 17 Sep 2011 | 14:39:41 UTC - in response to Message 40378.

If a modern (Intel, at least) CPU gets too hot, it will cycle down the clock until it cools off. Your 1.8 GHz may have have been running MUCH slower than that.
____________
Murphy (AtP)

T-Armstrong

Joined: 3 Aug 10
Posts: 270
ID: 64825
Credit: 1,013,734,308
RAC: 1,098,877

Message 40381 - Posted: 17 Sep 2011 | 14:46:17 UTC - in response to Message 40379.

mfbabb2
Thanks for helping,

I wait of the first SoB, and then we can see it.

I know, 1,8 Ghz is nothing, but I`m looking for a better Desktop with NVIDIA Quadt SLI 590 grafic, and 4,2 Ghz, but I wait 3 Month, then I buy 4 times this big mashines, with watercolling system... and so on.
____________

rroonnaalldd
Volunteer developer
Volunteer tester

Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0

Message 40384 - Posted: 17 Sep 2011 | 17:44:34 UTC - in response to Message 40379.

If a modern (Intel, at least) CPU gets too hot, it will cycle down the clock until it cools off.

It depends of why it was to hot or was the thermal throttling activated.
When the thermal throttling was activated, the cpu will stay below their nominal frequency until you reboot the host (seen on my Core2 Duo and Quad). In some cases you have to unplug the powercord (found in a german journal over energy measurements on modern Sandy Bridge CPUs and their Z68-chipset).
____________
Best wishes. Knowledge is power. by jjwhalen

T-Armstrong

Joined: 3 Aug 10
Posts: 270
ID: 64825
Credit: 1,013,734,308
RAC: 1,098,877

Message 40387 - Posted: 17 Sep 2011 | 20:28:18 UTC - in response to Message 40384.

rroonnaalldd,

Thanks

Armstrong*
____________

T-Armstrong

Joined: 3 Aug 10
Posts: 270
ID: 64825
Credit: 1,013,734,308
RAC: 1,098,877

Message 40398 - Posted: 18 Sep 2011 | 11:19:37 UTC - in response to Message 40387.

Mow, 2 Days = 22 % in 48 hours

2 Tasks, I crunch with Turbo speed technology, thats the same, @2,8 Ghz. This Intel technology, tkes the power of 8 core in two processors, and is 2 times faster. SoB are ready in perhaps 4 Days ????
____________

[DPC]Charley

Joined: 15 Apr 11
Posts: 803
ID: 95137
Credit: 94,993,523
RAC: 0

Message 42090 - Posted: 23 Oct 2011 | 14:00:39 UTC

Silly question maybe, but I finished my first SoB WU a couple of days ago and got it confirmed today. It does not say "\$number is not prime" like you normally get on other projects when a number is not prime. If this happens there it means it is prime. Could it possibly mean just that here too or is there no notification of non-primality?

Counter still says 0, so haven't gotten my hopes up just yet ;)
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)

Darryl

Joined: 21 Mar 09
Posts: 100
ID: 37210
Credit: 366,233,783
RAC: 0

Message 42093 - Posted: 23 Oct 2011 | 15:07:17 UTC - in response to Message 42090.

Silly question maybe, but I finished my first SoB WU a couple of days ago and got it confirmed today. It does not say "\$number is not prime" like you normally get on other projects when a number is not prime. If this happens there it means it is prime. Could it possibly mean just that here too or is there no notification of non-primality?

Counter still says 0, so haven't gotten my hopes up just yet ;)

The SOB workunits do not seem to show the number once tested (from looking at a couple of examples on the stats pages). The Riesel LLR project is the same in this regard - there is no notification of the number being tested, once it is validated. It would be nice if this feature could be added of course...
____________

Warped

Joined: 25 Aug 08
Posts: 288
ID: 27792
Credit: 28,151,335
RAC: 56

Message 55018 - Posted: 28 May 2012 | 17:14:49 UTC

As mentioned by John in the opening post, this project started in April 2002 and therefore recently passed the 10-year mark.

I have also recently completed 10 years of on-off (mainly off) contribution to the project:
My Profile
____________
Warped

KEP

Joined: 10 Aug 05
Posts: 288
ID: 110
Credit: 8,529,584
RAC: 50

Message 85679 - Posted: 18 May 2015 | 16:29:51 UTC

Could some of the moderators please remove k=90527 from the Prime Sierpinski Problem list of k's remaining. The k=90527 had a prime at n=9162167 (90527*2^9162167+1) and therefor this k is not searched anymore :)

Take care.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 85681 - Posted: 18 May 2015 | 17:47:07 UTC - in response to Message 85679.

Could some of the moderators please remove k=90527 from the Prime Sierpinski Problem list of k's remaining. The k=90527 had a prime at n=9162167 (90527*2^9162167+1) and therefor this k is not searched anymore :)

Take care.

You mean in the first post in this thread, right? I'll take care of it. (Run of the mill moderators can't fix that since they don't have the ability to modify other people's posts.)

Does 90527 appear anywhere else?
____________
My lucky number is 75898524288+1

KEP

Joined: 10 Aug 05
Posts: 288
ID: 110
Credit: 8,529,584
RAC: 50

Message 85702 - Posted: 19 May 2015 | 17:24:05 UTC - in response to Message 85681.

Could some of the moderators please remove k=90527 from the Prime Sierpinski Problem list of k's remaining. The k=90527 had a prime at n=9162167 (90527*2^9162167+1) and therefor this k is not searched anymore :)

Take care.

You mean in the first post in this thread, right? I'll take care of it. (Run of the mill moderators can't fix that since they don't have the ability to modify other people's posts.)

Does 90527 appear anywhere else?

No it doesn't appear anywhere else, as far as I know :)

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 86911 - Posted: 3 Aug 2015 | 14:40:16 UTC

Good Day,

It is interesting for me what is the lower bound of current search.

According to Statistics page, Primegrid is currently crunching 31 000 000.

"Seventeen or Bust" project is crunching 29 000 000

Also I know that "Seventeen or Bust" crunched 12 000 000 this winter.

Is it possible to post more detail about the status of this project?

Thank you.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 86920 - Posted: 3 Aug 2015 | 17:59:47 UTC - in response to Message 86911.

Good Day,

It is interesting for me what is the lower bound of current search.

According to Statistics page, Primegrid is currently crunching 31 000 000.

"Seventeen or Bust" project is crunching 29 000 000

Also I know that "Seventeen or Bust" crunched 12 000 000 this winter.

Is it possible to post more detail about the status of this project?

Thank you.

It's complicated.

Until recently, SoB and PG were dividing up the work by 'n', which has some disadvantages. Earlier this year, we decided to divide the work by 'k' instead. Moving forward, PrimeGrid will be doing all the crunching on two of the six remaining 'k's, and SoB will work on the other 4.

But before we do that, we're first cleaning up the work we had previously agreed to do.

Here's the details of PrimeGrid's work on SoB:

17.0M to 17.2M: all 6 k's completed
18.0M to 18.5M: all 6 k's completed
20.0M to 22.0M: all 6 k's completed
27.0M to 28.0M: all 6 k's completed or in progress

~29.1M to 31.0M: k=10223 and k=67607 in progress

31.0M to 32.0M: all 6k's in progress

32.0M to 50.0M: k=10223 and k=67607 will be done next

The reason we switched from dividing the work by 'n' to dividing the work by 'k' is that when you divide the work by 'n', if either group finds a prime, there's a 50/50 chance that the other group would have been crunching that 'k' at a higher 'n' level, and all of their work above the prime would have been wasted effort. By dividing the work by 'k' that won't happen.

Also, it should be noted that everything PrimeGrid has done was double checked at the beginning, while SoB only does single tests initially and does the double checks later. All of PrimeGrid's work can therefore be considered "complete", but that's not the case with all of SoB's work. There are errors, and the double checking does sometimes turn up primes that were missed the first time around. It's rare, but not unheard of.
____________
My lucky number is 75898524288+1

Death[Kiev]
Volunteer tester

Joined: 12 Jun 08
Posts: 126
ID: 24043
Credit: 3,622,648
RAC: 0

Message 88406 - Posted: 23 Sep 2015 | 18:54:51 UTC - in response to Message 86920.

Is there any plans to do some double-check for SoB for lower n's?
First few Ms can be done quickly.

AFAIK one prime in SoB was from double check.
____________

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 88407 - Posted: 23 Sep 2015 | 19:24:24 UTC - in response to Message 88406.

Is there any plans to do some double-check for SoB for lower n's?
First few Ms can be done quickly.

AFAIK one prime in SoB was from double check.

Yes, but it's a ways off, and definitive plans have not been established. Also, we'll only be double checking 2 of the 6 k's.

As soon as we finish the current n=31M range, our agreement with the SoB project will change. Instead of dividing up the work by n-range, each group will work only on specific K values. PrimeGrid will work on two k's while SoB will work on the other 4.

Once we reach that point, we're going to devise a plan to double check the early work on our two k's.

____________
My lucky number is 75898524288+1

Rafael
Volunteer tester

Joined: 22 Oct 14
Posts: 885
ID: 370496
Credit: 334,085,845
RAC: 156,268

Message 88410 - Posted: 23 Sep 2015 | 19:36:21 UTC - in response to Message 88407.

Yes, but it's a ways off, and definitive plans have not been established. Also, we'll only be double checking 2 of the 6 k's.

As soon as we finish the current n=31M range, our agreement with the SoB project will change. Instead of dividing up the work by n-range, each group will work only on specific K values. PrimeGrid will work on two k's while SoB will work on the other 4.

Once we reach that point, we're going to devise a plan to double check the early work on our two k's.

Out of curiosity, any particular reason for a 2/4 k split, instead of an even 3/3?

Also, if a double check has potential to find primes missed, and thus waste effort on the upper n, shouldn't that be a top priority?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 88413 - Posted: 23 Sep 2015 | 19:54:40 UTC - in response to Message 88410.

Yes, but it's a ways off, and definitive plans have not been established. Also, we'll only be double checking 2 of the 6 k's.

As soon as we finish the current n=31M range, our agreement with the SoB project will change. Instead of dividing up the work by n-range, each group will work only on specific K values. PrimeGrid will work on two k's while SoB will work on the other 4.

Once we reach that point, we're going to devise a plan to double check the early work on our two k's.

Out of curiosity, any particular reason for a 2/4 k split, instead of an even 3/3?

Also, if a double check has potential to find primes missed, and thus waste effort on the upper n, shouldn't that be a top priority?

The choice of K's was SoB's decision, so you'll need to ask them. Yes, the double check is important, but it's also difficult to do because we don't run the same software as SoB. That's why we don't have a plan for it yet.
____________
My lucky number is 75898524288+1

Rafael
Volunteer tester

Joined: 22 Oct 14
Posts: 885
ID: 370496
Credit: 334,085,845
RAC: 156,268

Message 88414 - Posted: 23 Sep 2015 | 20:16:07 UTC - in response to Message 88413.

we don't run the same software as SoB

And ours is faster / slower in general, or does it depend on particular k's and n's? Or pretty much the same speed?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 88415 - Posted: 23 Sep 2015 | 20:29:57 UTC - in response to Message 88414.

we don't run the same software as SoB

And ours is faster / slower in general, or does it depend on particular k's and n's? Or pretty much the same speed?

Just different. We use LLR. They've used at least two other programs over the years. There's many programs than can be used to test the primality of Proth numbers. I don't recall the details of what they run; I last looked at this in January. I don't plan at looking at it again until at least 2016, possibly 2017. The current n=31M range will take a while.

EDIT: We have almost 2 years worth of SoB work remaining at n=31M, so the post-31M plans are very low priority at this time.
____________
My lucky number is 75898524288+1

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 89330 - Posted: 25 Oct 2015 | 14:27:11 UTC

It is interesting is it known a difficulty of the algorithm of finding those promes?

Gary Craig
Volunteer tester

Joined: 30 Dec 09
Posts: 3213
ID: 52890
Credit: 1,005,618,748
RAC: 0

Message 89392 - Posted: 28 Oct 2015 | 4:42:49 UTC - in response to Message 89330.

It is interesting is it known a difficulty of the algorithm of finding those promes?

Vitaly,
The "LLR" program we use tests numbers in the Sierpinski Problem much the same way, although it adjusts for the size of the numbers by picking an appropriate FFT size, and also understands things like hardware (e.g. AVX vs. non-AVX). But the basic algorithm is the same. The main issue with finding primes in the Seventeen or Bust project is just that the numbers being tested are so huge, primes are very rare. I hope that addresses your query.
--Gary

Rafael
Volunteer tester

Joined: 22 Oct 14
Posts: 885
ID: 370496
Credit: 334,085,845
RAC: 156,268

Message 89393 - Posted: 28 Oct 2015 | 5:07:37 UTC - in response to Message 89392.

It is interesting is it known a difficulty of the algorithm of finding those promes?

Vitaly,
The "LLR" program we use tests numbers in the Sierpinski Problem much the same way, although it adjusts for the size of the numbers by picking an appropriate FFT size, and also understands things like hardware (e.g. AVX vs. non-AVX). But the basic algorithm is the same. The main issue with finding primes in the Seventeen or Bust project is just that the numbers being tested are so huge, primes are very rare. I hope that addresses your query.
--Gary

And there's less people running SoB tasks (again, due to the huuuuge numbers / run times).

Though my intuition tells me we'll get a prime in 2016.

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 89417 - Posted: 29 Oct 2015 | 16:41:10 UTC - in response to Message 89393.

Yes ))

Also it would be cool if Google Super Computer helped to resolve this problem.

Recently, he helped in solving one mathematical problem "Rubik's cube mystery":

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 89499 - Posted: 2 Nov 2015 | 14:46:53 UTC - in response to Message 89393.

It is interesting is it known a difficulty of the algorithm of finding those promes?

Vitaly,
The "LLR" program we use tests numbers in the Sierpinski Problem much the same way, although it adjusts for the size of the numbers by picking an appropriate FFT size, and also understands things like hardware (e.g. AVX vs. non-AVX). But the basic algorithm is the same. The main issue with finding primes in the Seventeen or Bust project is just that the numbers being tested are so huge, primes are very rare. I hope that addresses your query.
--Gary

And there's less people running SoB tasks (again, due to the huuuuge numbers / run times).

Though my intuition tells me we'll get a prime in 2016.

It is interesting,

Is it possible to split the calculation of particular number among several users.
In this case, for example, particular user will calculate 2 million of iterations rather than 30 millions?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 89500 - Posted: 2 Nov 2015 | 15:04:05 UTC - in response to Message 89499.

Is it possible to split the calculation of particular number among several users.
In this case, for example, particular user will calculate 2 million of iterations rather than 30 millions?

This topic has come up before, and I've put some thought into what would be needed to make it work.

Is it technically possible? Yes.

But it's not practical. It would involve transferring very large files back and forth between your host computers and our servers. That would be a problem for many users, and it would most definitely be a problem for PrimeGrid both in terms of disk storage and network bandwidth. It's not going to happen.
____________
My lucky number is 75898524288+1

Christopher Graesser

Joined: 19 Jun 09
Posts: 6
ID: 42154
Credit: 16,538,538
RAC: 0

Message 101180 - Posted: 22 Nov 2016 | 17:20:24 UTC

Hi Michael,

Greets
Chris

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 101181 - Posted: 22 Nov 2016 | 18:09:06 UTC - in response to Message 101180.

Hi Michael,

Greets
Chris

We know. The link is actually fine; it's the website that's down. The site administrator of that website is working on getting the site back online. (Unlike the saga with SoB's server earlier this year, I fully expect this one to be resolved successfully. I don't, however, have any idea how long it will take.)

Until then, if you need information from prothsearch.net you can access their webpages on the wayback machine.
____________
My lucky number is 75898524288+1

Christopher Graesser

Joined: 19 Jun 09
Posts: 6
ID: 42154
Credit: 16,538,538
RAC: 0

Message 101185 - Posted: 22 Nov 2016 | 19:10:49 UTC

OK, thanks a lot. I did read the wiki article regarding the sierpinski problem. :-)

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 103258 - Posted: 7 Jan 2017 | 23:00:00 UTC - in response to Message 101181.

Hi Michael,

Greets
Chris

We know. The link is actually fine; it's the website that's down. The site administrator of that website is working on getting the site back online. (Unlike the saga with SoB's server earlier this year, I fully expect this one to be resolved successfully. I don't, however, have any idea how long it will take.)

Until then, if you need information from prothsearch.net you can access their webpages on the wayback machine.

That page, and the root page http://www.prothsearch.net/, now show a message, ProThSearch.net is coming soon. Pro Th? What would Monsieur FranÃ§ois Proth say? Mersenneforum.org has the thread Why is Prothsearch Gone? but it does not give much information. /JeppeSN

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 103813 - Posted: 18 Jan 2017 | 21:07:05 UTC - in response to Message 103258.

That page, and the root page http://www.prothsearch.net/, now show a message, ProThSearch.net is coming soon.

Today I saw a new site at that address. It did not (yet?) have all the subpages the old website had. Does anyone know if the new site is made by Wilfrid Keller or some other reliable person? /JeppeSN

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 103816 - Posted: 18 Jan 2017 | 21:44:57 UTC - in response to Message 103813.

That page, and the root page http://www.prothsearch.net/, now show a message, ProThSearch.net is coming soon.

Today I saw a new site at that address. It did not (yet?) have all the subpages the old website had. Does anyone know if the new site is made by Wilfrid Keller or some other reliable person? /JeppeSN

I don't know.

____________
My lucky number is 75898524288+1

unconnected
Volunteer tester

Joined: 4 Apr 09
Posts: 95
ID: 37970
Credit: 313,835,241
RAC: 4,380

Message 103819 - Posted: 18 Jan 2017 | 22:28:09 UTC

I think domain prothsearch.net was expired and re-registered by another person.
Update: prothsearch.com is the right place now.

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 103820 - Posted: 18 Jan 2017 | 22:38:13 UTC - in response to Message 103819.

Update: prothsearch.com is the right place now.

You are right; great find!

EDIT: Also, the page sierp.html is there and contains a short notice on this project's elimination of k=10223.

/JeppeSN

axn
Volunteer developer

Joined: 29 Dec 07
Posts: 285
ID: 16874
Credit: 28,027,106
RAC: 0

Message 103827 - Posted: 19 Jan 2017 | 8:11:49 UTC

Must be some kind of auto-generated code

eCommerce Made Easy With Proth Search

WTF?!

Iain Bethune
Honorary cruncher

Joined: 28 Jan 09
Posts: 1588
ID: 34775
Credit: 194,056,043
RAC: 0

Message 103828 - Posted: 19 Jan 2017 | 8:45:14 UTC - in response to Message 103827.

Must be some kind of auto-generated code

eCommerce Made Easy With Proth Search

WTF?!

The domain has been poached by a cyber-squatter. They are actually quite clever, in that they have uploaded a lot of content from the old site that looks quite relevant, but inserted links to their "new eCommerce search tool". I'll see if I can get in touch with Wilfrid.

Cheers

- Iain
____________
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 103885 - Posted: 20 Jan 2017 | 6:06:00 UTC

Long sad story, but Iain was correct: a cyber squatter got the domain name, so it should be ignored.

http://www.prothsearch.com/ is the new URL (".com" instead of ".net")

It will take a little while for Wilfrid to create all the pages, but the site IS coming back.
____________
My lucky number is 75898524288+1

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 103886 - Posted: 20 Jan 2017 | 6:21:22 UTC

FYI -- except for the posts in this thread where we're talking about the change from prothsearch.net to prothsearch.com, I'm going to globally change the URL from .net to .com in every post in the forums, regardless of who the author of the post is. I don't want anyone following links to the fake website. I have no idea what their intentions are, and there's certainly a possibility that they're malicious.

____________
My lucky number is 75898524288+1

composite
Volunteer tester

Joined: 16 Feb 10
Posts: 769
ID: 55391
Credit: 696,046,522
RAC: 113,357

Message 103900 - Posted: 20 Jan 2017 | 9:17:56 UTC - in response to Message 103828.

Must be some kind of auto-generated code

eCommerce Made Easy With Proth Search

WTF?!

The domain has been poached by a cyber-squatter. They are actually quite clever, in that they have uploaded a lot of content from the old site that looks quite relevant, but inserted links to their "new eCommerce search tool".

*** off-topic ***

Quite right. Domain squatting is now really about exploiting high search engine rankings from expired domains, with residual earnings by selling domains in the after-market. These "entrepreneurs" are always on the lookout for freshly expired domains with good "link juice" that can drive significant traffic to their "money site" by building a "private blog network" using expired domains. http://nichesiteproject.com/private-blog-networks That's an instructive and perhaps lucrative read.

Squatters can subscribe to domain search services, some of which offer archived copies of web sites for sale so that squatted domains continue to look legitimate to search engines. If you run a web site and regularly check IP addresses from your web server logs, you will eventually discover when your domain name has been catalogued and your web content copied by these services. That's how I found out about this cyber-squatting industry.

The value of expired domains shot up after Google deflated the ranking of "public blog networks". Domain registrars jumped on this profitable bandwagon and facilitated it by charging exorbitant fees to recover domain names during the grace period past their expiry date. Let this be a warning, if your web site has good ranking, you will pay dearly for not renewing the name before the deadline. If you have no desire to continue the site, you can at least sell it in the aftermarket before it expires, maybe for more than you paid to register the name. Most people don't realize this. However I expect that will change in a few years when Google's AI will be able to distinguish squatted domains from good ones as it indexes the web.

Currently, cyber-squatters devalue search engine results by turning part of the web into garbage for specific enough search terms. It harms legitimate buinesses who compete for search engine ranking, particularly in industries where businesses have significant turnover and can't maintain their domain names. I've seen it personally, where a few of my competitors have gone out of business and their domains have been taken over like this. Heck, the domain of the former occupant of our current physical location is still alive years later, but the web site content has changed to something in Japanese. This is insane because that business and domain name were named for the street we are located on, it still has good ranking, and I could use that domain to redirect traffic to my web site. People still know that former business by name. And although not squatting, the business strategy of occupying a particular "space" also works with physical locations, and even with phone numbers - I once called a supplier in a high-value niche industry that I hadn't yet realized went out of business, and one of his competitors had grabbed that phone number. Smart move.

Our business is a local service not driven by ecommerce, so the firewall categorically blocks IP address not from this continent. I'm not apologizing to overseas folks that can't see the web site because the chance that I'll get any business from them is nil. We are not an international destination, despite exhortations via email from "magazines" offering us "awards" (for a price). The real benefit of keeping it local is that it saves the web server's CPU and bandwidth for customers that really use it, and we function perfectly with a \$4/month VPS serving content from the ramdisk. The only pain in the ass are people that use proxy servers in the cloud to circumvent firewall rules. The access log suggests that they are usually up to no good. It also amazes me how many accesses from retail ISPs are declaring the user agent to be "Googlebot" and friends - you can be certain that Googlebot and Bingbot are not using IP services from Time-Warner or Comcast. The real tricky ones are those running in Google's cloud and Microsoft's Azure cloud.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 103907 - Posted: 20 Jan 2017 | 11:48:42 UTC - in response to Message 103900.

I will let composite's post about cybersquatting stay because it's relevant and informative, but anyone who wishes to to continue or comment on the general topic of cybersquatting please create a new thread in the General discussions forum.

As it is, I may end up moving this whole discussion about prothsearch elsewhere, but I do want to point out one fact here: Wilfrid Keller didn't accidently let the domain expire. The original owner of the site (and the domain name) is deceased, and Wilfrid was unable to get control of the domain name. The registrar was not very helpful, and thanks to composite's information, it's clear why.
____________
My lucky number is 75898524288+1

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 106043 - Posted: 15 Mar 2017 | 12:37:10 UTC - in response to Message 103907.

It looks like clean up for 10223 is finished:

http://www.primegrid.com/stats_sob_llr.php

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 106067 - Posted: 15 Mar 2017 | 19:49:44 UTC - in response to Message 106043.

It looks like clean up for 10223 is finished:

http://www.primegrid.com/stats_sob_llr.php

Quite old news. See The "Top Ten" Record Primes and Official Announcement and so on. /JeppeSN

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 106068 - Posted: 15 Mar 2017 | 20:07:23 UTC - in response to Message 106067.

It looks like clean up for 10223 is finished:

http://www.primegrid.com/stats_sob_llr.php

Quite old news. See The "Top Ten" Record Primes and Official Announcement and so on. /JeppeSN

Nope. That's brand new news. The very last k=10223 SOB work unit completed today. (It started back in September.)
____________
My lucky number is 75898524288+1

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 106075 - Posted: 15 Mar 2017 | 21:29:42 UTC - in response to Message 106068.

It looks like clean up for 10223 is finished:

http://www.primegrid.com/stats_sob_llr.php

Quite old news. See The "Top Ten" Record Primes and Official Announcement and so on. /JeppeSN

Nope. That's brand new news. The very last k=10223 SOB work unit completed today. (It started back in September.)

You are right, of course! Thanks. I do not know why I misread Vitaly's post and thought he was referring to the prime find. I even discussed the 10223 clean up with you in the other thread ... /JeppeSN

[AF>Occitania] f11ksx

Joined: 27 Jan 07
Posts: 1
ID: 5309
Credit: 557,306,111
RAC: 668,484

Message 106143 - Posted: 18 Mar 2017 | 8:40:39 UTC

for following k, n will be
k = 21 181 n = 0 mod 4
k = 22 699 n = 2 mod 4
k = 24 737 n = 3 mod 4
k = 55 459 n = 2 mod 4
k = 67 607 n = 3 mod 4

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 106156 - Posted: 18 Mar 2017 | 22:49:25 UTC - in response to Message 106143.

for following k, n will be
k = 21 181 n = 0 mod 4
k = 22 699 n = 2 mod 4
k = 24 737 n = 3 mod 4
k = 55 459 n = 2 mod 4
k = 67 607 n = 3 mod 4

Yes. This can be seen by sieving to the prime 5. /JeppeSN

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 106258 - Posted: 22 Mar 2017 | 11:41:56 UTC - in response to Message 106156.

It is interesting "Min in progress" is reduced to 7 000 000.
Does it mean that you start checking tasks of SoB project that was closed last year?

As I remember just before closed it was tested about 1 million tasks.

Thanks.

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,494,092
RAC: 3,785

Message 106269 - Posted: 22 Mar 2017 | 13:24:35 UTC - in response to Message 106258.

It is interesting "Min in progress" is reduced to 7 000 000.
Does it mean that you start checking tasks of SoB project that was closed last year?

Any candidate where SB didn't seem to have a pair of matching residues from two different users is being retested. There were candidates much smaller than n=7M, but we eliminated them internally. If you check the link above, you can see how many candidates need to be tested per n range.

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 106278 - Posted: 22 Mar 2017 | 16:14:14 UTC - in response to Message 106269.

It is interesting "Min in progress" is reduced to 7 000 000.
Does it mean that you start checking tasks of SoB project that was closed last year?

Any candidate where SB didn't seem to have a pair of matching residues from two different users is being retested. There were candidates much smaller than n=7M, but we eliminated them internally. If you check the link above, you can see how many candidates need to be tested per n range.

OK,

but I do not see "how many candidates need to be tested per n range" in that link.

Thanks.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 106279 - Posted: 22 Mar 2017 | 17:09:02 UTC - in response to Message 106278.

It is interesting "Min in progress" is reduced to 7 000 000.
Does it mean that you start checking tasks of SoB project that was closed last year?

Any candidate where SB didn't seem to have a pair of matching residues from two different users is being retested. There were candidates much smaller than n=7M, but we eliminated them internally. If you check the link above, you can see how many candidates need to be tested per n range.

OK,

but I do not see "how many candidates need to be tested per n range" in that link.

Thanks.

____________
My lucky number is 75898524288+1

[SG]Felix

Joined: 2 Aug 16
Posts: 32
ID: 453628
Credit: 42,198,810
RAC: 5,664

Message 109080 - Posted: 25 Jul 2017 | 13:52:42 UTC

when i want to make an app_config.xml, whats the app name?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 109081 - Posted: 25 Jul 2017 | 14:35:05 UTC - in response to Message 109080.

when i want to make an app_config.xml, whats the app name?

You can find the list here: http://www.primegrid.com/forum_thread.php?id=7299&nowrap=true#105133

For Seventeen or Bust it's llrSOB.

____________
My lucky number is 75898524288+1

[SG]Felix

Joined: 2 Aug 16
Posts: 32
ID: 453628
Credit: 42,198,810
RAC: 5,664

Message 109082 - Posted: 25 Jul 2017 | 15:21:35 UTC

thank you very much

bcavnaugh

Joined: 8 Nov 13
Posts: 215
ID: 273570
Credit: 795,246,478
RAC: 0

Message 123985 - Posted: 27 Dec 2018 | 0:50:33 UTC

Is this App still in Double Check Mode?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 123990 - Posted: 27 Dec 2018 | 1:11:52 UTC - in response to Message 123985.

Is this App still in Double Check Mode?

Yes.

If we're really, really, lucky, the double check will finish near the end of 2019. I think it's more likely that it will finish in early to mid 2020.
____________
My lucky number is 75898524288+1

Jay

Joined: 27 Feb 10
Posts: 103
ID: 56067
Credit: 53,729,466
RAC: 3,186

Message 123999 - Posted: 27 Dec 2018 | 14:21:52 UTC - in response to Message 123990.

If we're really, really, lucky, the double check will finish near the end of 2019. I think it's more likely that it will finish in early to mid 2020.

If we're really, really, really lucky, the next 5 results returned will be primes and the conjecture will be put to bed.

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 124001 - Posted: 27 Dec 2018 | 15:53:43 UTC - in response to Message 123999.

If we're really, really, lucky, the double check will finish near the end of 2019. I think it's more likely that it will finish in early to mid 2020.

If we're really, really, really lucky, the next 5 results returned will be primes and the conjecture will be put to bed.

Provided that the next five results are for five different k, of course. /JeppeSN

xii5ku

Joined: 17 Dec 16
Posts: 66
ID: 476505
Credit: 576,779,205
RAC: 3,688

Message 124129 - Posted: 2 Jan 2019 | 17:29:16 UTC

Question (or complaint):

At the PrimeGrid preferences page, I see SoB-LLR marked as "Focus project". And even as the sole one.

Why do you want your contributors to focus on a project which most current CPUs do not support effectively? I am referring to the extraordinary cache requirements, which only a few server CPUs are able to satisfy. All other CPUs run into a memory bandwidth wall with this project.

Why doesn't PrimeGrid go with the times and point their contributors to those projects which are more efficiently supported by current popular computer hardware?

Jay

Joined: 27 Feb 10
Posts: 103
ID: 56067
Credit: 53,729,466
RAC: 3,186

Message 124131 - Posted: 2 Jan 2019 | 18:25:45 UTC - in response to Message 124129.

Question (or complaint):

Why do you want your contributors to focus on a project which most current CPUs do not support effectively?
...
Why doesn't PrimeGrid go with the times and point their contributors to those projects which are more efficiently supported by current popular computer hardware?

I've got to disagree with the effectiveness of older cpu's. I have an i7-950, as the only cpu I have used on this project. That processor was introduced in the 2nd quarter of 2009 according to Intel. I have earned, with that single cpu, over 20 million credit on SOB. It currently takes approximately 4.5 days to return a task. I'm usually the first to return a workunit.

Dave

Joined: 13 Feb 12
Posts: 2829
ID: 130544
Credit: 954,747,840
RAC: 254,811

Message 124132 - Posted: 2 Jan 2019 | 18:30:49 UTC

Doesn't matte what the CPU is (within reason e.g don't use a PII) - if it can reliably compete a SoB & within the extended deadline it all helps.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 124133 - Posted: 2 Jan 2019 | 18:48:46 UTC - in response to Message 124129.

Question (or complaint):

At the PrimeGrid preferences page, I see SoB-LLR marked as "Focus project".

"Focus project" is the project (or projects) that we would most like people to work on. SoB is currently the focus project because we want to get the double check completed as quickly as possible.

Hardware changes over time, and different people have very different hardware, so how efficient certain hardware might be doesn't really play a role in choosing the "Focus project".

If you wish to run something that's more efficient on your computer, by all means do so. "Focus project" is merely a suggestion, not a requirement. If it was mandatory, PrimeGrid wouldn't give a choice.
____________
My lucky number is 75898524288+1

xii5ku

Joined: 17 Dec 16
Posts: 66
ID: 476505
Credit: 576,779,205
RAC: 3,688

Message 124155 - Posted: 2 Jan 2019 | 23:22:58 UTC - in response to Message 124133.

As a post scriptum, most of my own CPUs are actually very well suited to these tasks. My question came from a macroeconomic angle, so to speak, not a personal one.

xii5ku

Joined: 17 Dec 16
Posts: 66
ID: 476505
Credit: 576,779,205
RAC: 3,688

Message 124321 - Posted: 4 Jan 2019 | 21:05:54 UTC - in response to Message 124129.

xii5ku wrote:
Why do you want your contributors to focus on a project which most current CPUs do not support effectively? I am referring to the extraordinary cache requirements, which only a few server CPUs are able to satisfy. All other CPUs run into a memory bandwidth wall with this project.

Michael,
I have to take this back partially. I made more tests with SoB-LLR since I posted. (My last systematic SoB-LLR testing was 2 years ago with v7, I only now took the time to test v8 more thoroughly.) While it is true that Xeons greatly benefit from having ample cache, performance on desktop CPUs with fast RAM is actually quite close when normalized to core count and clock speed. (My current testing is limited to BDW-EP and KBL-S though.)

WayneKFord

Joined: 28 Jan 18
Posts: 16
ID: 972674
Credit: 26,305,078
RAC: 29,948

Message 126137 - Posted: 3 Feb 2019 | 18:52:49 UTC

Excuse me for being new to this science. However, how can we "prove" the thesis numerically? It seems that for k*2^n+1 and a given k, "all" n must be tested.

Crun-chi
Volunteer tester

Joined: 25 Nov 09
Posts: 3008
ID: 50683
Credit: 59,175,698
RAC: 93,033

Message 126138 - Posted: 3 Feb 2019 | 18:55:44 UTC - in response to Message 126137.

Excuse me for being new to this science. However, how can we "prove" the thesis numerically? It seems that for k*2^n+1 and a given k, "all" n must be tested.

Not all , it just need to find prime for remain k
When prime is found , then is removed from conjuncture.

____________
92*10^1439761-1 REPDIGIT PRIME :) :) :)
314187728^131072+1 GENERALIZED FERMAT
31*332^367560+1 CRUS PRIME
Proud member of team Aggie The Pew. Go Aggie!

composite
Volunteer tester

Joined: 16 Feb 10
Posts: 769
ID: 55391
Credit: 696,046,522
RAC: 113,357

Message 126141 - Posted: 3 Feb 2019 | 19:13:44 UTC - in response to Message 126138.

Excuse me for being new to this science. However, how can we "prove" the thesis numerically? It seems that for k*2^n+1 and a given k, "all" n must be tested.

Not all , it just need to find prime for remain k
When prime is found , then is removed from conjuncture.

The project would necessarily test n to infinity for a given k if the conjecture is false. We just don't know that it is false, so we prefer to think that it is true and try to prove it. By all means, go ahead and mathematically prove the conjecture to be false. It would save us all a lot of computer time.

EDIT: we prefer to think the conjecture is true because that's the only result we can test and prove numerically.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 126143 - Posted: 3 Feb 2019 | 19:17:47 UTC - in response to Message 126137.

Excuse me for being new to this science. However, how can we "prove" the thesis numerically? It seems that for k*2^n+1 and a given k, "all" n must be tested.

Yes and no, literally.

We can prove the conjecture true by finding just a single prime for each of the remaining 5 k's. We don't need to test "every" n to prove the conjecture true. On the other hand...

Since there's an infinite number of n's, it's not possible to prove the conjecture false using this method, as that would involve testing an infinite number of n's.
____________
My lucky number is 75898524288+1

Eudy Silva

Joined: 26 Aug 17
Posts: 1325
ID: 918937
Credit: 221,903,924
RAC: 455,351

Message 126149 - Posted: 3 Feb 2019 | 20:58:05 UTC - in response to Message 126137.

WayneKFord wrote:
Excuse me for being new to this science. However, how can we "prove" the thesis numerically? It seems that for k*2^n+1 and a given k, "all" n must be tested.

You see, a Sierpinski number is an odd k such that k*2n+1 is composite for all n.

If you give me one k and ask me if it is a Sierpinski number, I could:

1. prove k is not a Sierpinski number, by showing you one value of n that makes k*2n+1 a prime number
2. prove k is a Sierpinski number using some fancy mathematical proof
3. prove k is a Sierpinski number, making you wait for ever, while I test all values of n to get composite numbers for k*2n+1

I can give you a Sierpinski number: 78557

In 1962, John Selfridge proved that 78,557 is a Sierpinski number...meaning he showed that for all n, 78557*2^n+1 was not prime.

John Selfridge was a smart guy. He must have used option 2 above :)

So, we know 78557 is a Sierpinski number: 78557*2n+1 is composite for all n.
Isn't there any other odd number, smaller than 78557, that is also a Sierpinski number ?
This is the conjecture (Sierpinski Problem): 78557 is the smallest Sierpinski number.
Edit: In other words, according to the conjecture, if you pick any k<78557 then k*2n+1 will be prime for at least one value of n.

Most number theorists believe that 78,557 is the smallest Sierpinski number, but it hasn't yet been proven. In order to prove it, it has to be shown that every single k less than 78,557 is not a Sierpinski number, and to do that, some n must be found that makes k*2^n+1 prime.

Almost all odd numbers k below 78557 have been tested and one n was found that made k*2n+1 a prime number.
Currently, only 5 odd numbers below 77857 have yet to be tested to show they are not Sierpinski numbers: 21181, 22699, 24737, 55459 and 67607.

We're using option 1 above.
We have to show that:
21181*2n+1 is prime for some n
22699*2n+1 is prime for some n
24737*2n+1 is prime for some n
55459*2n+1 is prime for some n
67607*2n+1 is prime for some n
____________
"Accidit in puncto, quod non contingit in anno."
Something that does not occur in a year may, perchance, happen in a moment.

Cowering

Joined: 22 Jan 17
Posts: 2
ID: 485165
Credit: 534,784,316
RAC: 59

Message 126168 - Posted: 4 Feb 2019 | 4:53:39 UTC

Wow. I was checking the first invalid work unit I have gotten in months, and see that my compute speed is really poor. Could someone explain this workunit page:

http://www.primegrid.com/workunit.php?wuid=591705832

My TR1950x OC to 3.7GHZ is taking 30 days for a test that the client says should take two days. I was assuming this was a guess by the client (since the PPSeive times are also always wrong). But I see two intel chips with basically the same integer and floating numbers actually doing the tests in two days in the above workunit.

My rig is as rock solid as I can make it. Tested it for a month before starting BOINC with all the usual tests like PRIME95, etc. I only use it for 4 hours a day. The rest of the time is all BOINC.

If someone else is running a TR1950x and is getting two day tests on the reissued numbers, please let me know and i'll start pulling more hair trying to figure this out.

For my 3.7GHZ speed, I get the same scores as others. IE Cinebench15 scores around (3000-3100), etc.

Luigi R.

Joined: 11 Feb 14
Posts: 165
ID: 297455
Credit: 38,836,937
RAC: 4,263

Message 126173 - Posted: 4 Feb 2019 | 8:33:11 UTC - in response to Message 126168.

The other two hosts have multithreading enabled.

You need something like this:

<app_config> <app> <name>llrSOB</name> <fraction_done_exact/> <report_results_immediately/> </app> <app_version> <app_name>llrSOB</app_name> <cmdline>-t 32</cmdline> <avg_ncpus>32</avg_ncpus> </app_version> </app_config>

____________
My DC mathematical side :)

dukebg
Volunteer tester

Joined: 21 Nov 17
Posts: 238
ID: 950482
Credit: 23,670,125
RAC: 7,187

Message 126205 - Posted: 4 Feb 2019 | 16:30:36 UTC - in response to Message 126168.

My TR1950x OC to 3.7GHZ is taking 30 days for a test that the client says should take two days. I was assuming this was a guess by the client (since the PPSeive times are also always wrong). But I see two intel chips with basically the same integer and floating numbers actually doing the tests in two days in the above workunit.

They are multithreading, i.e. running the same workunit on multiple cores, while you're running a separate unit on every core at the same time. You probably have a lot of progress on every of ~32 units running in parallel by the time one of them finishes. Eventually you'll finish all 32 units and if you'll divide the total time you'll see that you have about the same throughput.

See Luigi R.'s post above mine for what you should have in your app_config for setting up multithreading. Or more instructions for example, in first post here, under Multi-threading optimisation instructions.

Cowering

Joined: 22 Jan 17
Posts: 2
ID: 485165
Credit: 534,784,316
RAC: 59

Message 126207 - Posted: 4 Feb 2019 | 16:32:59 UTC - in response to Message 126173.

The other two hosts have multithreading enabled.

You need something like this:
<app_config> <app> <name>llrSOB</name> <fraction_done_exact/> <report_results_immediately/> </app> <app_version> <app_name>llrSOB</app_name> <cmdline>-t 32</cmdline> <avg_ncpus>32</avg_ncpus> </app_version> </app_config>

I'll try this. I don't exactly know what you mean by the "other two hosts". This is a single machine with a single TR1950x and a single GTX1080 (no VMs either). Will this fix BOINC displaying that LLR shows as running for 14+ days (in the tasks tab) before completing, but that also show as 30 days in the workunit URL I posted?

Luigi R.

Joined: 11 Feb 14
Posts: 165
ID: 297455
Credit: 38,836,937
RAC: 4,263

Message 126216 - Posted: 4 Feb 2019 | 17:11:05 UTC - in response to Message 126207.

I'll try this. I don't exactly know what you mean by the "other two hosts".

I meant your wingmen, the 'two intel chips'.

This is a single machine with a single TR1950x and a single GTX1080 (no VMs either). Will this fix BOINC displaying that LLR shows as running for 14+ days (in the tasks tab) before completing, but that also show as 30 days in the workunit URL I posted?

Workunit page shows both of them.
Your machine downloaded that task on December 29th and reported it 30 days later. There is nothing wrong, unless the task has been sitting for 14 days as completed before being reported. You should reduce your BOINC queue if you want your work to be processed immediately after the download.
Your runtime is 1,392,480.87 seconds, i.e. 16days 2hours 48minutes.
____________
My DC mathematical side :)

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 135087 - Posted: 23 Nov 2019 | 11:33:00 UTC - in response to Message 126216.

It is interesting, how many 31M tasks still we have?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 135089 - Posted: 23 Nov 2019 | 11:52:04 UTC - in response to Message 135087.

It is interesting, how many 31M tasks still we have?

It's better to think in terms of candidates (workunits) rather than tasks for this sort of question.

At this moment, there are 223 candidates still in progress in the 31M range. There's 12185 completed candidates, one of which is prime.

In the 32M range, there's 1405 candidates in progress and 8329 as yet unstarted candidates. 1546 are done.
____________
My lucky number is 75898524288+1

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 138267 - Posted: 24 Feb 2020 | 15:16:33 UTC - in response to Message 135089.

It is interesting, the double check is completed, but this task is still unfinished:

http://www.primegrid.com/workunit.php?wuid=621076596

It has n = 243724568

Is it okay?

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 138278 - Posted: 24 Feb 2020 | 21:30:35 UTC - in response to Message 138267.

It is interesting, the double check is completed, but this task is still unfinished:

http://www.primegrid.com/workunit.php?wuid=621076596

It has n = 243724568

Is it okay?

No, it is not completed. It is currently the "Min remaining n" on https://www.primegrid.com/server_status_subprojects.php, and the n is not what you say. As the oldest SOB work unit, it is visible even to people who did not process a task from it, even if it is still unfinished. In the name llrSOB_243724568, the last part has nothing to do with the n.

A new task was sent out just 10 hours ago. Addition: The computer that got the task now seems to have way too many tasks, so it could still take a long time before the work unit is complete :-(

The work unit will be OK once two tasks are "completed" and agree on the residue (result).

/JeppeSN

Ravi Fernando
Volunteer tester
Project scientist

Joined: 21 Mar 19
Posts: 144
ID: 1108183
Credit: 8,505,009
RAC: 7,921

Message 138280 - Posted: 24 Feb 2020 | 21:53:05 UTC - in response to Message 138267.

It is interesting, the double check is completed, but this task is still unfinished:

http://www.primegrid.com/workunit.php?wuid=621076596

It has n = 243724568

Is it okay?

As JeppeSN pointed out, that's not the n value. In fact this is currently the minimum remaining n as well as the oldest unfinished WU, so you can see here that it is for n = 31,811,338--a few hundred thousand past where the double-check ended. Note that the number 243,724,568 appearing in the WU's name is much larger than the n's we're testing.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 138284 - Posted: 24 Feb 2020 | 23:35:51 UTC - in response to Message 138267.

It is interesting, the double check is completed, but this task is still unfinished:

http://www.primegrid.com/workunit.php?wuid=621076596

It has n = 243724568

Is it okay?

The double check is indeed complete, and all candidates below 31.8 million are done.

That number you're quoting is an index into a database table. It's not "n".
____________
My lucky number is 75898524288+1

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 138350 - Posted: 26 Feb 2020 | 18:10:06 UTC - in response to Message 138267.

http://www.primegrid.com/workunit.php?wuid=621076596

Looks like JimB gave himself an additional task for that candidate now and confirmed (validated) Pavel Atnashev's result from August. One of the timed out tasks woke up (trickled) and is alive, so there are theoretically two more people who can gain credit if they confirm the result of Pavel and Jim. /JeppeSN

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 140309 - Posted: 18 May 2020 | 15:28:24 UTC - in response to Message 138350.

It is interesting, how many 31M tasks still we have?

stream
Volunteer moderator
Volunteer developer
Volunteer tester

Joined: 1 Mar 14
Posts: 834
ID: 301928
Credit: 488,466,944
RAC: 116,805

Message 140310 - Posted: 18 May 2020 | 17:18:06 UTC - in response to Message 140309.

It is interesting, how many 31M tasks still we have?

Only 2.

Vitaly

Joined: 24 Sep 12
Posts: 61
ID: 173343
Credit: 13,785,129
RAC: 0

Message 140658 - Posted: 9 Jun 2020 | 17:05:47 UTC - in response to Message 140310.

It is interesting, how many 31M tasks still we have?

Only 2.

Only 1 left )

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 791
ID: 1097922
Credit: 3,783,734
RAC: 40,187

Message 140756 - Posted: 13 Jun 2020 | 8:00:28 UTC - in response to Message 140658.

It is interesting, how many 31M tasks still we have?

Only 2.

Only 1 left )

Still 1. Was looking thru each subproject's range when I noticed this one ;)
____________
SHSID Electronics Group
SHSIDElectronicsGroup@outlook.com

GFN-14: 50103906^16384+1
Proth "SoB": 44243*2^440969+1

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 140765 - Posted: 14 Jun 2020 | 8:07:36 UTC - in response to Message 140756.

It is interesting, how many 31M tasks still we have?

Only 2.

Only 1 left )

Still 1. Was looking thru each subproject's range when I noticed this one ;)

Yes. That is, on /server_status_subprojects.php, you can see "Min remaining n". That links to a WU that is actually "unlocked" so that anyone can see it even though it is unfinished. Right now that is llrSOB_243729026 (wuid=622163821). Two people (Pavel Atnashev (2 Sep 2019) and Malcolm Beeson (11 Dec 2019)) have returned residues, but they do not match, so one of them must be invalid. The task is currently with msct256 where the deadline will be reached in 4.5 days unless msct256's computer sends "trickles" to extend the deadline.

Everything is working as expected.

The WU currently has 10 tasks of which 2 (as mentioned) have led to a (maybe wrong) result. The maximum is 15 resp. 5.

/JeppeSN

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,494,092
RAC: 3,785

Message 140777 - Posted: 14 Jun 2020 | 17:54:58 UTC - in response to Message 140765.

The WU currently has 10 tasks of which 2 (as mentioned) have led to a (maybe wrong) result. The maximum is 15 resp. 5.

Except that if it hits 15 it will be changed to some much higher number, I believe 99.

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 102,920

Message 140784 - Posted: 14 Jun 2020 | 19:44:26 UTC - in response to Message 140777.

The WU currently has 10 tasks of which 2 (as mentioned) have led to a (maybe wrong) result. The maximum is 15 resp. 5.

Except that if it hits 15 it will be changed to some much higher number, I believe 99.

Yes, I do not know what else we could do. I thought the limit of 15 was meant to catch situations where something is wrong (like a WU that is impossible for any user to do, because of some subtle error somewhere). When nothing is wrong, and the WU is just "unlucky" with the hosts it gets sent to, there is nothing else we can do but to keep sending it forever. /JeppeSN

Chooka

Joined: 15 May 18
Posts: 151
ID: 1014486
Credit: 242,033,674
RAC: 313,192

Message 140833 - Posted: 15 Jun 2020 | 20:56:18 UTC

I actually just messaged Pavel Atnashev. I'm curious to know how his 2600K & even his Q9400 finish SoB w/u's in LESS than half the time of my 4770 & 1790 CPU's! It's amazing...and I've no idea how it's possible.

I'm running mt (4 cores) and 1 w/u at a time...but less than HALF the time to finish a SOB!

I thought it might be the power of Linux but he's using Win7.

http://www.primegrid.com/results.php?hostid=905514&offset=0&show_names=0&state=4&appid=13

____________

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 140834 - Posted: 15 Jun 2020 | 21:07:22 UTC - in response to Message 140833.

I actually just messaged Pavel Atnashev. I'm curious to know how his 2600K & even his Q9400 finish SoB w/u's in LESS than half the time of my 4770 & 1790 CPU's! It's amazing...and I've no idea how it's possible.

Those computers are running BOINC, but they're not running the calculations. They're just gateways to the real computers.

Behind each gateway is a bunch of servers that do the actual computations. To PrimeGrid's server it may look like a ancient Core2Quad, but it's actually a stack of Xeon servers with hundreds of modern cores.

Also, he picked the Xeon CPUs very carefully so that the CPUs cache is large enough to hold the entire computation, thus avoiding the need to access the much slower main memory. This gives it a huge advantage over your Haswell.
____________
My lucky number is 75898524288+1

Chooka

Joined: 15 May 18
Posts: 151
ID: 1014486
Credit: 242,033,674
RAC: 313,192

Message 140835 - Posted: 15 Jun 2020 | 21:15:17 UTC - in response to Message 140834.

Ohhhhhhhh right. Wow... I've never heard of that being done before!
Gee...so I could look at anyone's stats but the PC i see may not ACTUALLY be what's crunching.

Wow. There you go.

Thank you Michael. I can now rest easy :D
____________

Nick

Joined: 11 Jul 11
Posts: 882
ID: 105020
Credit: 1,318,826,036
RAC: 2,505,141

Message 140837 - Posted: 16 Jun 2020 | 1:13:38 UTC - in response to Message 140835.

so I could look at anyone's stats but the PC i see may not ACTUALLY be what's crunching.

It's only Pavel's computers that are set up this way.
The times for SOB that I get on my two 99xx-X computers are usually around 36,000 sec running one SOB at a time with most cores and 4 cores for GPU, etc. Very occasionally the times are about 55,000 sec which I guess is due to the fastest cores being assigned to the GPU tasks - and happening due to poor timing of new SOB task and new GPU task. I noticed this happen on only one computer and when it happened it would take some tasks before it would return to the quicker time.

zombie67 [MM]
Volunteer tester

Joined: 30 Nov 06
Posts: 263
ID: 4065
Credit: 1,405,439,306
RAC: 1,604,103

Message 140838 - Posted: 16 Jun 2020 | 3:37:45 UTC - in response to Message 140835.

so I could look at anyone's stats but the PC i see may not ACTUALLY be what's crunching.

Absolutely yes. There are other ways to fake what we see on the stat pages too. But I am not sure if any others are doing it at this time. Perhaps there are others doing it with other projects. No way to easily tell.
____________
Reno, NV

Chooka

Joined: 15 May 18
Posts: 151
ID: 1014486
Credit: 242,033,674
RAC: 313,192

Message 140839 - Posted: 16 Jun 2020 | 3:44:29 UTC - in response to Message 140837.

so I could look at anyone's stats but the PC i see may not ACTUALLY be what's crunching.

It's only Pavel's computers that are set up this way.
The times for SOB that I get on my two 99xx-X computers are usually around 36,000 sec running one SOB at a time with most cores and 4 cores for GPU, etc. Very occasionally the times are about 55,000 sec which I guess is due to the fastest cores being assigned to the GPU tasks - and happening due to poor timing of new SOB task and new GPU task. I noticed this happen on only one computer and when it happened it would take some tasks before it would return to the quicker time.

So you must be chasing 1st over throughput then right Nick?
I would have thought running 14-18 cores / 1 task won't be best for throughput?

I think with my Threadripper, I was planned on doing 2 tasks across 16 cores and see how that goes.
____________

Nick

Joined: 11 Jul 11
Posts: 882
ID: 105020
Credit: 1,318,826,036
RAC: 2,505,141

Message 140840 - Posted: 16 Jun 2020 | 4:02:38 UTC - in response to Message 140839.

So you must be chasing 1st over throughput then right Nick?
I would have thought running 14-18 cores / 1 task won't be best for throughput?

I think with my Threadripper, I was planned on doing 2 tasks across 16 cores and see how that goes.

I go entirely by FFT size and amount of L2 + L3 cache for my 99xx-X computers - If I ran two SOB tasks it would go beyond cache into RAM. I do not use my 9900K for SOB as it only has 16 Mb L3 cache.
Right now I am running 321:
9900K - 2 x 3 core tasks (2 cores for GPU) - 8.6 tasks / day
99660X - 5 x 3 core tasks (1 core for GPU) - 20.8 tasks / day
9980XE - 4 x 4 core tasks (2 cores for GPU) - 21.7 tasks / day

zombie67 [MM]
Volunteer tester

Joined: 30 Nov 06
Posts: 263
ID: 4065
Credit: 1,405,439,306
RAC: 1,604,103

Message 140841 - Posted: 16 Jun 2020 | 4:22:11 UTC - in response to Message 140839.

I think with my Threadripper, I was planned on doing 2 tasks across 16 cores and see how that goes.

The best strategy for Zen2 Threadripper is to keep each MT task within each chiplet. For zen2, that is at most 8. Greater than 8, then the tasks have to swap cache over the bridge, which slows things down. You may get better total performance running two 4-thread tasks vs. one 8-thread task depending on the sub-project. But going over 8 will slow things down. This is with SMT (hyper-thread) turned off.