Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Number crunching :
2014 PrimeGrid Challenge Series ideas
| Author |
Message |
|---|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
                            
|
|
I've got three ideas for the 2014 Challenge series. (Pyrus is in charge of the Challenges and these are just my suggestions.)
1) January: Now that we can run Genefer on ATI GPUs, which are a lot faster than Nvidia, I'd like to run a World Record GFN challenge. For those not paying attention to what's been going on in the GFN topic, we now have an OpenCL version of Genefer that can run on both Nvidia and ATI GPUs. ATI GPUs have MUCH better floating point performance than Nvidia, so the $1000 GTX Titan and the $300 HD 7970 run Genefer at roughly the same the speed. With this new app and the superior performance from the Red side of the GPU landscape, now is the time to really push the World Record GFN project.
I propose that this challenge be on both the short and World Record tasks, so people without a GPU (or with only a slower GPU) can also participate. However, it will be primarily a World Record challenge, so although you can run the short tasks, they'll count for less, maybe 50% or 75% of the rate of World Record tasks. There would be an advantage to running the World Record tasks, but you could still participate even if completing a WR task in 13 days is beyond your hardware's abilities. That would put the minimum GPU required at approximately a GTX 460.
The cleanup will be long, so this would run in January.
2) Likewise, with the advent of the AVX version of LLR, we can think about running an SOB challenge. You would need a reasonably fast CPU to finish within 13 days, so the minimum CPU would be approximately a desktop Core2 or Phenom II CPU.
As with the GFN challenge, this will be a long cleanup, so it should be run early in the year. We probably shouldn't run two 13 day challenges back to back, so perhaps this could be the third challenge, which would be in March.
3) We should run another SR5 challenge. SR5 is a good place to find primes and knock off some k's from the conjectures. Alternatively, if we should happen to move ESP to BOINC (something that gets discussed occaisionally), we would run that one instead.
What does everyone think?
____________
My lucky number is 75898524288+1 | |
|
|
|
|
1) I was already considering running a GFN-short challenge again due to the addition of an OpenCL client, as a WR seemed a bit heavy to me for the NVidia users. I'm perfectly happy making it either, as long as it is in January.
2) To my taste running two 13 day challenges a year is quite a big ask of participants. That's why I wouldn't readily plan both a GFN and SoB challenge. I can see the benefit to the project though and if TPTB (that is, The People That Be) would like a SoB challenge in March I'd be more than happy to put it in there too.
3) SR5 makes perfect sense. All for it.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
 | |
|
|
|
|
Lastyear it was pointed out we have not had a Cullen challenge for a while. Perhaps it is time to fit one in somewhere.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 | |
|
|
|
Lastyear it was pointed out we have not had a Cullen challenge for a while. Perhaps it is time to fit one in somewhere.
Jup, good plan :)
The projects I have on the short list so far are (in no particular order):
- GFN (January)
- PPS Sieve (December)
- PPS(E)
- SR5
- Cullen
- TRP Sieve (June/July/August somewhen)
- TRP LLR
- Cullen
- 321
That's 9 challenges. 8 to 10 challenges is what I think is a reasonable amount, because else you'll pretty much be running challenges back to back. Whether it's 8, 9 or 10 would depend on how long the challenges are.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
That's 9 challenges. 8 to 10 challenges is what I think is a reasonable amount, because else you'll pretty much be running challenges back to back. Whether it's 8, 9 or 10 would depend on how long the challenges are.
From my perspective, I'd be perfectly happy running 12 every year. It's fun, the workload on me is negligible, it increases awareness about PrimeGrid, gives participants a chance to learn something, and it increases the amount of crunching that gets done here. The only downside is if people get bored with that many challenges and stop participating. The server gets a workout during challenges with short tasks, but keeping it running under stress is what I consider fun.
____________
My lucky number is 75898524288+1 | |
|
|
|
That's 9 challenges. 8 to 10 challenges is what I think is a reasonable amount, because else you'll pretty much be running challenges back to back. Whether it's 8, 9 or 10 would depend on how long the challenges are.
From my perspective, I'd be perfectly happy running 12 every year. It's fun, the workload on me is negligible, it increases awareness about PrimeGrid, gives participants a chance to learn something, and it increases the amount of crunching that gets done here. The only downside is if people get bored with that many challenges and stop participating. The server gets a workout during challenges with short tasks, but keeping it running under stress is what I consider fun.
I was mainly thinking of participant boredom, not necessarily of server load. If there's a challenge every month, I'd be much more likely to skip one or more.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2125
ID: 1178
Credit: 8,404,284,720
RAC: 540,645
                                     
|
|
Just to throw in my 2 cents...
I am not a fan of a WR only or WR bonus GFN challenge. I like the idea of a combined GFN challenge, but I worry about what participation we will get by focusing on the GFN WR tasks only. We certainly would pick up the AMD/ATI GPU users, but I know that my own participation would be very limited. I am torn by wanting us to make progress toward the world record and not wanting to potentially alienate anyone from the challenge series (and the boost it brings across multiple projects) overall.
I am very much in favor of a Cullen challenge, since I do not recall the last time we ran one (it has been years). Cullen would also get a nice boost from the AVX apps.
SR5 is a no-brainer as I see it. Good chance of finds occurring and work is of modest length.
I also think that 9 challenges is about right, and I would not go over 10 since we have other competing interests (e.g., Tour de Primes in Feb., PRPnet challenges, etc.) throughout the year.
EDIT: I just thought of something that might make SoB and the GFN challenge doable. What about running two or three challenges concurrently? I am envisioning a GFN short challenge and GFN WR challenge running together (maybe an 8-day GFN short and a 13-day GFN WR so that the WR participants get the bonus of a longer challenge scoring). Alternatively, a GFN WR and SoB 13-day challenge running together. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
I am not a fan of a WR only or WR bonus GFN challenge. I like the idea of a combined GFN challenge, but I worry about what participation we will get by focusing on the GFN WR tasks only. We certainly would pick up the AMD/ATI GPU users, but I know that my own participation would be very limited. I am torn by wanting us to make progress toward the world record and not wanting to potentially alienate anyone from the challenge series (and the boost it brings across multiple projects) overall.
I spent a fair amount of time thinking about this. That doesn't mean I'm right, of course. My thoughts went something like this:
Now is finally the time to do a GFN-WR challenge!
But that will leave out all sorts of people who can't run the WR tasks, especially in only 13 days.
Let's do both n=20 and n=22 together!
But then the competitive people will only run the short tasks because there's a competitive advantage to running shorter tasks. (There's still the long-task credit bonus for WR, but no bonus in the challenge.)
So let's put in a challenge bonus for the WR tasks, so people who can run the WR tasks in 13 days will run them rather than the short tasks.
But won't people who can't run the WR tasks feel alienated?
That last question required some serious thought. I decided that, perceptions aside, with a very few exceptions people who can't run the WR tasks in 13 days want to participate just for the sake of pariticpating in the challenge rather than actually trying to compete in the challenge and try to get into the top 200 individual scoring positions. The reason is simple: If you've got a GPU that slow, or only a CPU (in a GPU challenge, that's like bringing the proverbial knife to a gun fight), you're not going to get into a scoring position anyway. I *can* (barely) finish a WR task in 13 days, and I've never, even once, made it into scoring position.
The way I figure it, the only people who might be upset about this for a valid reason are the small number of people with large numbers of CPU cores at their disposal. So many cores that you're probably going to finish at the top of the charts ahead of the GPU crowd. And I think you will still finish high on the list even if there's a built in competitive disadvantage.
People will still get upset about it, but many won't think it through enough to realize it will actually have little effect on them. So it becomes a marketing/PR issue about perceived problems rather than actual problems.
What I want to do is create an incentive so the fast GPUs are running WR tasks rather than short tasks. Without such an incentive, the short tasks are the better competitive choice, and this is intended to be a WR challenge. Without a WR incentive, we might as well just make it another Short GFN challenge like we did in 2013.
____________
My lucky number is 75898524288+1 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
EDIT: I just thought of something that might make SoB and the GFN challenge doable. What about running two or three challenges concurrently? I am envisioning a GFN short challenge and GFN WR challenge running together (maybe an 8-day GFN short and a 13-day GFN WR so that the WR participants get the bonus of a longer challenge scoring). Alternatively, a GFN WR and SoB 13-day challenge running together.
Oddly enough, as I was typing my first reply, the same thought occurred to me.
This needs some thought as well as input from participants.
(I would keep a GFN short challenge at 13 days, however -- 8 days might push most desktop Core2 and Phenom II CPUs out of the challenge.)
EDIT: I might like the idea of running GFN-WR and SoB simultaneously the best. A CPU and a GPU challenge running concurrently makes a lot of sense.
EDIT2: The more I think about it, the more I like it. What does everyone else think?
____________
My lucky number is 75898524288+1 | |
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1884
ID: 352
Credit: 3,037,812,339
RAC: 83,518
                         
|
|
GPU GFN task are low CPU usage so CPU can do SoB alongside.
I still wonder if there is anythnig that can be done to speed-up long cleaning process.
Sending recycled tasks to trustworthy hosts is a way to go but BOINC isn't designed to do that :-(
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 | |
|
|
|
|
Are the GFN GPU tasks floating precision or double precision?
____________
 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
Are the GFN GPU tasks floating precision or double precision?
Double Precision.
Any LLR, PFGW, Genefer, Prime95, or similar program that needs to multiply huge numbers requires double precision floating point hardware.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
I favor a GFN challenge in January, 13 days long, where both short and WR tasks score counts toward a single combined total, for standings purposes. I'd probably run WR units on GPU and short units on CPU with AVX-Genefer. I'd rather not see a SoB challenge run in parallel, as I'd have to split my effort. A challenge on SoB sounds fine in principle to me, but run it separately, for 13 days, or save for 2015.
I'd vote for no more than 8 challenges; that plus all of the PRPnet challenges seems like plenty. I know the number of PRPnet challenges doesn't affect most people, but even with 8 on BOINC, that's still one roughly every 6 or 7 weeks, and as discussed above, at one or two of them will last nearly two weeks.
+1 for a Cullen challenge. If the early part of the year is going to be dominated by the 13-day challenges, make Cullen an 8-day run maybe in late September? I don't think most people are too keen on long challenges during the northern hemisphere summer. Also +1 for SR5 at some point.
We had a Father's Day challenge this year, but there's never been a Mother's Day challenge. Also, keep the tradition of naming/scheduling a challenge or two based on equinoxes/eclipses/meteors (just a personal preference).
My $.02.
--Gary
p.s. @Pyrus: in this post, you listed Cullen twice. Was that intentional? | |
|
|
|
I favor a GFN challenge in January, 13 days long, where both short and WR tasks score counts toward a single combined total, for standings purposes. I'd probably run WR units on GPU and short units on CPU with AVX-Genefer. I'd rather not see a SoB challenge run in parallel, as I'd have to split my effort. A challenge on SoB sounds fine in principle to me, but run it separately, for 13 days, or save for 2015.
I'd vote for no more than 8 challenges; that plus all of the PRPnet challenges seems like plenty. I know the number of PRPnet challenges doesn't affect most people, but even with 8 on BOINC, that's still one roughly every 6 or 7 weeks, and as discussed above, at one or two of them will last nearly two weeks.
+1 for a Cullen challenge. If the early part of the year is going to be dominated by the 13-day challenges, make Cullen an 8-day run maybe in late September? I don't think most people are too keen on long challenges during the northern hemisphere summer. Also +1 for SR5 at some point.
We had a Father's Day challenge this year, but there's never been a Mother's Day challenge. Also, keep the tradition of naming/scheduling a challenge or two based on equinoxes/eclipses/meteors (just a personal preference).
My $.02.
--Gary
p.s. @Pyrus: in this post, you listed Cullen twice. Was that intentional?
Yes, because I'm very much in favor of it! O-)
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
|
I vote for 15 days for the GFN, so that I can compete in it!
____________
  | |
|
|
|
We had a Father's Day challenge this year, but there's never been a Mother's Day challenge. Also, keep the tradition of naming/scheduling a challenge or two based on equinoxes/eclipses/meteors (just a personal preference).
+1 on this comment!
And in favor of running GFN + SOB simultaneously.
____________
Member of the Dutch Power Cows
My Stats | |
|
|
|
|
I would be opposed to running GFN and SoB at the same time.
____________
| |
|
|
|
|
GFN and SoB at the same time, but one month time. So all devices (i hope) can complete one task.
____________
DeleteNull | |
|
|
|
GFN and SoB at the same time, but one month time. So all devices (i hope) can complete one task.
15 days I might consider, a month definitely not. Clean up on these long challenges always takes longer than the challenge itself and by making a month long challenge we'll be cleaning up for months.
Furthermore, I can well imagine that people who normally don't run their systems 24/7 might well do so for a challenge. In a challenge of days or two weeks in the extreme they might leave them running, but most likely not for a month.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
|
Isn't doing a GFN and SoB challenge together quite an extreme... er, "challenge"? I know we can adjust the block size for GFN on GPU, but it still creates screen lag, to whatever extent. That's not a problem for people who have redundant PCs (i.e. ones which they don't use for everyday browsing, etc.) or use cloud computing, but it does affect others. Still, I'd be up for it, the world won't end if I have a bit of screen lag on my "browsing" PC for a couple of weeks.
I'm currently sieving GFN N=20 as much as I can, even though I can see that the "priority" is N=21; my logic is that as soon as N=20 is optimally sieved, then we can move on to sieving N=21 and 22. GFN manual sieving is obviously PRPNet/PSA, but it's so easy to set up, the instructions are concise and comprehensive, and it doesn't even require a DAT file like SR5 sieving. Just put a line template in a notepad file in the same folder, adjust the P ranges accordingly, and right-click>paste into your console window.
My only other thought is a Wieferich challenge, most people have GPUs capable of participating and if a load of us got on to it, then there'd be a massive amount of work done. The only downside is that (block size depending) that also causes screen lag, and so people wanting to crunch and browse at the same time would have to put up wth that. Wieferich primes are proven to exist (even if there are only two...) whereas Wall-Sun-Sun primes are not, so by hedging our bets there's probably more chance of finding a third Wieferich than there is finding a first WSS. That's obviously conjecture though, and I'm sure people who know far more about stats and maths than me would like to comment on that... | |
|
|
|
...
I'm currently sieving GFN N=20 as much as I can, even though I can see that the "priority" is N=21; my logic is that as soon as N=20 is optimally sieved, then we can move on to sieving N=21 and 22. GFN manual sieving is obviously PRPNet/PSA, but it's so easy to set up, the instructions are concise and comprehensive, and it doesn't even require a DAT file like SR5 sieving. Just put a line template in a notepad file in the same folder, adjust the P ranges accordingly, and right-click>paste into your console window.
Optimal sieve depth won't be reached for years if I'm not mistaking.
My only other thought is a Wieferich challenge, most people have GPUs capable of participating and if a load of us got on to it, then there'd be a massive amount of work done. The only downside is that (block size depending) that also causes screen lag, and so people wanting to crunch and browse at the same time would have to put up wth that. Wieferich primes are proven to exist (even if there are only two...) whereas Wall-Sun-Sun primes are not, so by hedging our bets there's probably more chance of finding a third Wieferich than there is finding a first WSS. That's obviously conjecture though, and I'm sure people who know far more about stats and maths than me would like to comment on that...
If Wief or WSS are ported to BOINC I'd be more than happy to shuffle some things about to fit them in, but as far as I am aware that's not yet on the table.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
|
My vote goes to SOB!
____________
 | |
|
|
|
|
I would definitely be up for an SoB challenge, but would prefer not to have it run concurrently with a GFN challenge. I suppose in reality it won't make THAT much difference, but that's my preference. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
GFN and SoB at the same time, but one month time. So all devices (i hope) can complete one task.
It depends on what you mean by "all devices". The normal deadline for SoB is 40 days. There are lots of active computers with CPUs and GPUs that can not complete SoB or GFN in a month. I have two running in my house -- actual computers in use by people, not just antiques in storage.
Lots of people have low end GPUs. They can't complete GFN-WR in a month. While not a lot of people are probably running 32-bit CPUs, there are some. They can't do SoB in a month. (Off the top of my head, my "basement space heater" should take about 8 weeks to do SoB.) Netbook type CPUs can't do SoB in a month.
So while making the challenge long does have merit and will enable more people to participate, there will always be some computers that are too slow.
____________
My lucky number is 75898524288+1 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
|
gazzyk1ns:
I would definitely be up for an SoB challenge, but would prefer not to have it run concurrently with a GFN challenge.
Tarmo Ilves:
My vote goes to SOB!
Michael Milerick:
I would be opposed to running GFN and SoB at the same time.
Since this isn't a democracy and you're not voting, this decision will be made based on merit, and not the number of people who like or don't like a proposal. In other words, we want to choose the best ideas, not merely the most popular.
Therefore, it would be helpful if you explained why you like or dislike any idea, rather than merely a +1 or -1.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
A GPU challenge, assuming one is fully dedicated to it, almost always causes some degree of screen lag (again, depending on block settings). So it may be an unpleasant surprise to people new to GPU challenges, and simply undesirable to others compared with a CPU challenge, which obviously stresses the system but would not interrupt casual use of the PC.
Separating the two would possibly cause less confusion and/or conflict, i.e. if we had a 10-day GPU challenge then people could prepare for it and then commit to it; a CPU challenge doesn't really affect a PC, unless of course you're wanting to number-crunch on other BOINC projects at the same time.
I don't cruch other projects, so for me it's largely irrelevant (although GFN plus SoB at the same time would cause me to pause for a few hours a day), but that's how I see the "argument". | |
|
|
|
Therefore, it would be helpful if you explained why you like or dislike any idea, rather than merely a +1 or -1.
With the new AVX app, my CPU is probably capable of completing almost as many GFN short work units as my GPUs. By having two challenges run concurrently, you would force me to do one of two things.
1. Run only GFN work units in which case I am able to get a higher score in the GFN challenge.
2. Run both GFN and SoB in which case I am able to do as well as I would have in SoB but suffer in the GFN standings.
[quote]Since this isn't a democracy and you're not voting, this decision will be made based on merit, and not the number of people who like or don't like a proposal. In other words, we want to choose the best ideas, not merely the most popular.[quote]
If you run both challenges at the same time, it will be a democracy since we will all be forced to decide which challenge we want to participate in.
____________
| |
|
|
|
Therefore, it would be helpful if you explained why you like or dislike any idea, rather than merely a +1 or -1.
I'd like to see a Cullen challenge sometime in 2014 because I don't think there's ever been one (if there was, it was before I joined). I'd like to see another SR5 challenge as it's a conjecture project (which I find more compelling), it's relatively new, and IMHO the WU runtimes are "just right".
I think I explained my rationale for running GFN+GFNWR separately from SoB in my prior post, which substantially agrees with what Michael Millerick said in that regard. In fact, I'll go further and say that I hope to not ever see two overlapping challenges on any projects, including PRPnet.
I'll add to my prior proposal a winter solstice pps sieve challenge (tradition), and a 1-day challenge on one of the very short subprojects (whenever, whichever, up to you; it's a brief entertaining test of one's boinc-management skills).
--Gary | |
|
|
|
|
Hi
all the comments about running long GPU tasks in January 2014.
Spare a little thought about us in the southern hemisphere, mid summer, 30deg C outside and in parts of Australia over 40C.
While I set mine up with plenty of fans, I have come home to room temps over 40c.
For most crunchers a week is a long time, and our partners start looking at the power bill.
I will be in for most types of tasks but something new would help.
and congratulations on the new openCL apps for ATI.
Cheers
Ross*
____________
| |
|
|
|
|
I'm opposed to running any BOINC sub projects concurrently. Especially GFN or PPSsieve with any other LLR projects. When running PPSsieve it uses some CPU resources which takes away from LLR resources, as well as GFN if a host needs to lower the block size to avoid screen lag (like my main computer which then uses 50% of a CPU core). This would put those in a disadvantage of having to use part of their LLR (CPU) resources for feeding their GPUs.
____________
Largest Primes to Date:
As Double Checker: SR5 109208*5^1816285+1 Dgts-1,269,534
As Initial Finder: SR5 243944*5^1258576-1 Dgts-879,713
 | |
|
|
|
|
First I'll say I don't really care either way about running a 2 task challenge, I can see good and bad points but it seems that most of the comments are to say it is too much of a challenge!!
That is the idea!! If it was meant to be easy it would not be a challenge. If a screen lag is unacceptable then don't do it, but I would be sorry to see less challengers for that reason.
I am against a GFN WR challenge as I cannot complete a GFN WR in under 8 weeks, I would be absent for such a challenge, as I think would a great many others. It is not just the old computers that would be out of it, so would 2 year old i5/i3 machines with 630 GPUs. Such a challenge would be for the elite, recently upgraded or the rich and we have, so far, been having challenges that try to be as inclusive as possible.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 | |
|
|
|
First I'll say I don't really care either way about running a 2 task challenge, I can see good and bad points but it seems that most of the comments are to say it is too much of a challenge!!
That is the idea!! If it was meant to be easy it would not be a challenge. If a screen lag is unacceptable then don't do it, but I would be sorry to see less challengers for that reason.
Well I think it's everyone's mutual goal to come to some sort of agreement, isn't it? i.e. the greatest participation plus the greatest amount of important work completed, which the mods consider to be priorities.
I am against a GFN WR challenge as I cannot complete a GFN WR in under 8 weeks
If I could just quote you again:
First I'll say I don't really care either way about running a 2 task challenge, I can see good and bad points but it seems that most of the comments are to say it is too much of a challenge!!
That is the idea!! If it was meant to be easy it would not be a challenge.
;) It seems you're being rather selective with your arguments, relating to your personal preferences?
I'm half-joking of course, happy crunching, whatever the outcome of this discussion may be! | |
|
|
|
|
;) It seems you're being rather selective with your arguments, relating to your personal preferences?
I'm half-joking of course, happy crunching, whatever the outcome of this discussion may be!
I see your point, but that is not really right. I was not too clear, I meant the 2 task challenge as GFN/SOB, which I do not have any preference. It is GFN WR without anything else I think is the bad idea.
I see no problem with a GFN/GFN WR or a GFN/SOB or even a GFN/GFN WR/SOB challenge so long as the many who cannot do GFN WR in time are not left out in the cold.
Hopefully a bit clearer,
Happy crunching to you as well.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
...so long as the many who cannot do GFN WR in time are not left out in the cold.
It's not just "in time". Many people can't do GFN-WR at all. Some people don't have a GPU, and even though AMD/ATI GPUs can now be used, not all of them have double precision (which is also true of older Nvidia GPUs.)
____________
My lucky number is 75898524288+1 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2781
ID: 130544
Credit: 934,442,218
RAC: 29,067
              
|
|
I don't see a problem doing WR on GPU & SoB on CPU. PCs are designed to multitask at the end of the day anyway. & we will fit in more into the year. Try it. | |
|
|
|
|
- Yes to SR5, seems to be a sweet spot right now.
- Yes to a 1-day challenge. As mentioned above, exciting (for some) resource-management problem, moreso than usual.
- I like the idea of a Cullen challenge getting in the rotation. I like the idea of exposing people to work they don't normally do, at least for short bursts.
- In the same idea, perhaps we consider some sort of 2 or 3-year templated challenge rotation. Looks like there's just a bit too many types of work for 1 year to catch 'em all, but not so many that we can't keep a few favorites each year. Not sure what this would look like exactly, but it might come down to alternating things like GFN / SoB, Cullen / Woodall, PPS Sieve / TRP Sieve. And then adjusting the schedule based on challenge length / seasonality / overlap with PRPnet challenges, etc. Of course, force a spot for a new work type (like the SR5 this year). I think 8-12 challenges / year is about right.
- Perhaps we should consider extending the scoring positions beyond 200 places? I don't know how long the challenge system has existed, but I think it's multiple years. Not sure where the 200 spot number came from, but my initial guess is that it was to weed out all the single-task performances and excessively light contributors (and maybe it's as much as the administration wanted to handle). I'd like to imagine participation has grown such that spots 201 thru (some number here, maybe 300/400/500/1000) are actually contributing a reasonable amount, and I'd hope that it's not terribly hard to adjust the scoring algorithm appropriately (perhaps breaking backward-compatibility, but with different challenge lengths for the same work, I doubt they are comparable currently anyway). With more challenges in the year, this might drum up a bit more enthusiasm, with people thinking "Well, I might not be able to crack spot 200 in any given challenge, but I will do them all around a spot-350 pace, and maybe that'd be enough reliability to get me around 200th when all is said and done at the end of the year." If 200 is still a good cut-off, keep it, just throwing out ideas.
- Regarding the proposed long challenge in January, I am on board with that. It does penalize those in the southern hemisphere, but it is the most likely way to guarantee clean-up by the end of the year. I could see moving this to March / April. You might also consider (not very heavily) the CPU / GPU release cycle. Not sure if those first-in-line (or maybe the wave of folks after the guinea pigs) would be eager to see how many of the big WUs they could crank out with new metal.
- If we do GFN, I really like the idea of combining GFN and GFN WR, with an additional bonus to the WR tasks.
These next few points aren't necessarily asking for answers, but they're things I would consider if I had the information that would answer them.
- If we do GFN / GFN WR + SoB in 2014, what do we do in 2015 if it's (subjectively measured as) a success? What if it's not?
- If we do GFN / GFN WR in 2014, would we skip SoB in 2014 and then move it to 2015? By 2015, how long will SoB tasks be / how long will the challenge have to last? (I imagine that the data exists to look at how task length has been changing over time, and this would be easy-ish to project.)
- If we do SoB in 2014, would we skip GFN / GFN WR in 2014 and then move it to 2015? By 2015, how long will GFN tasks be / how long will the challenge have to last? (Probably a harder one to estimate.)
- If we did GFN / GFN WR in 2014 and then SoB separately in 2014 (say, January and March), then what would 2015 look like?
- Are there a "significant" amount of undiagnosable / time-consuming to fix errors going on with any of the above that would cause headaches to new participants with those work types? That is, if someone tried to participate and they were out of commission for, say, 2 days trying to work through an issue, would that be enough to make them unable to get credit? If there are these errors, do we think that they will be solved before a challenge in 2014? 2015?
Hope this helps. | |
|
|
|
I don't see a problem doing WR on GPU & SoB on CPU. PCs are designed to multitask at the end of the day anyway. & we will fit in more into the year. Try it.
That's a fair point. If the Genefer challenge were limited to only World Record tasks, I would agree with you... run that and a SoB challenge in parallel. But if the Genefer challenge counted work from both short and long tasks, I'd want to be able to devote the CPU to doing short GFN tasks without having to neglect a SoB challenge running in parallel.
Cheers,
--Gary | |
|
|
|
|
Thanks for the extensive reply :)
I'm going to reply some of your points, others are simply noted not ignored.
- In the same idea, perhaps we consider some sort of 2 or 3-year templated challenge rotation. Looks like there's just a bit too many types of work for 1 year to catch 'em all, but not so many that we can't keep a few favorites each year. Not sure what this would look like exactly, but it might come down to alternating things like GFN / SoB, Cullen / Woodall, PPS Sieve / TRP Sieve. And then adjusting the schedule based on challenge length / seasonality / overlap with PRPnet challenges, etc. Of course, force a spot for a new work type (like the SR5 this year). I think 8-12 challenges / year is about right.
An X-year template is an option, but one I'm not particularly fond of in this case. New subprojects and both improved hardware and software appear too often to make this a viable option in my mind. Take 2013 alone. We've added a gazillion new Genefer clients both for CPU's and GPU's and we've added SR5. You can't really plan ahead for that. Furthermore, SoB is a conjecture project. I.e. it may end at some point. It's entirely possible, though unlikely, that during the next year the final primes are found.
- Perhaps we should consider extending the scoring positions beyond 200 places? I don't know how long the challenge system has existed, but I think it's multiple years. Not sure where the 200 spot number came from, but my initial guess is that it was to weed out all the single-task performances and excessively light contributors (and maybe it's as much as the administration wanted to handle). I'd like to imagine participation has grown such that spots 201 thru (some number here, maybe 300/400/500/1000) are actually contributing a reasonable amount, and I'd hope that it's not terribly hard to adjust the scoring algorithm appropriately (perhaps breaking backward-compatibility, but with different challenge lengths for the same work, I doubt they are comparable currently anyway). With more challenges in the year, this might drum up a bit more enthusiasm, with people thinking "Well, I might not be able to crack spot 200 in any given challenge, but I will do them all around a spot-350 pace, and maybe that'd be enough reliability to get me around 200th when all is said and done at the end of the year." If 200 is still a good cut-off, keep it, just throwing out ideas.
The "scoring algorithm" is non-existent. I've just taken what it was last year and copy/pasted it to this year with an extension for 13 days. Rejiggling it to allow for more people is no problem at all. We just need to see if it's desirable. I'll look into this further and will get back to you on it.
- If we do GFN / GFN WR + SoB in 2014, what do we do in 2015 if it's (subjectively measured as) a success? What if it's not?
- If we do GFN / GFN WR in 2014, would we skip SoB in 2014 and then move it to 2015? By 2015, how long will SoB tasks be / how long will the challenge have to last? (I imagine that the data exists to look at how task length has been changing over time, and this would be easy-ish to project.)
- If we do SoB in 2014, would we skip GFN / GFN WR in 2014 and then move it to 2015? By 2015, how long will GFN tasks be / how long will the challenge have to last? (Probably a harder one to estimate.)
- If we did GFN / GFN WR in 2014 and then SoB separately in 2014 (say, January and March), then what would 2015 look like?
- Are there a "significant" amount of undiagnosable / time-consuming to fix errors going on with any of the above that would cause headaches to new participants with those work types? That is, if someone tried to participate and they were out of commission for, say, 2 days trying to work through an issue, would that be enough to make them unable to get credit? If there are these errors, do we think that they will be solved before a challenge in 2014? 2015?
Hope this helps.
SoB currently only increases very slowly if I'm not mistaking, i.e. in a year the units will be a bit longer that will be hours rather than days. GFN WR on the other hand increases a lot more quickly, though Mike can probably throw around some actual numbers.
2015 is too far away for my liking to start planning for already.
And with regard to potential problems for participants. Well there aren't any inherent problems with the clients themselves as far as I am aware. Problems are generally due to something or other on the participants computer and to me that's part of the challenge.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
SoB currently only increases very slowly if I'm not mistaking, i.e. in a year the units will be a bit longer that will be hours rather than days. GFN WR on the other hand increases a lot more quickly, though Mike can probably throw around some actual numbers.
The leading edge for GFN WR is currently 13134. That is predicted to take 263 hours on my GTX 460. Lets say a year from now we've tripled the leading edge and it's at around 39000. That would increase the time to 293 hours.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
Maybe it is an idea to bring one multiple challenge, where we have to use cpu and gpu and everyone is free to do what task he/she want? In my opinion it gives people an understanding on system stability and the option for combine the right tasks for a high score as possible.
All components overclocked and also at full load is good to understand what it does for your system. Perhaps not the most important thing of a challenge, but an addition that might be valuable.
I don't mind what further challenges are coming and when in 2014. Just suprise me. :) | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
Maybe it is an idea to bring one multiple challenge, where we have to use cpu and gpu and everyone is free to do what task he/she want? In my opinion it gives people an understanding on system stability and the option for combine the right tasks for a high score as possible.
All components overclocked and also at full load is good to understand what it does for your system. Perhaps not the most important thing of a challenge, but an addition that might be valuable.
I don't mind what further challenges are coming and when in 2014. Just suprise me. :)
Although the challenges are Pyrus's responsibility, and ultimately his decision is what counts, I must veto this idea as being unfeasible from a server standpoint. There's two problems with it:
1) We use three different challenge scoring systems, depending on the type of work being done: LLR, Genefer, and sieving. No matter how we decide to combine them, it will be unfair to someone.
2) During a challenge AND during the challenge cleanup, we must suspend the purging of completed tasks from the database. As a result, the number of tasks in the database grows dramatically. It's not so bad with the big tasks, because there aren't that many of them even though those challenges run many days, and the short tasks are manageable because those challenges typically run for only a few days, but a challenge for everything would have to be long, and the cleanup would last for months. The database would become hundreds of times larger than it normally is, and the system isn't designed to handle that.
Also, people are pretty good at "finding the shortest path to the cheese". We'll end up with the majority of computers (and possibly the majority of people, too) running the single project that provides the highest credit/hour rate, so we may as well just run a challenge on that project.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
So, I've been doing so bean counting and this is what I found:
Year M. Project Name Participants
2012 1 PPS LLR Year of the Dragon 3242
2012 4 SGS LLR Marie-Sophie Germain 2090
2012 5 TRP LLR Cinco de Mayo 1269
2012 6 GFN Alan Turing Year 430
2012 8 TRP Sieve Olympics 850
2012 10 PPS LLR An Apple a Day 1200
2012 11 321 LLR 321 Blast Off 1651
2012 12 PPS Sieve Winter Solstice 1963
2013 1 GFN Short Year of the Snake 810
2013 2 SGS Valentine's Day 1030
2013 3 PPS LLR Equinox 2542
2013 4 WOO LLR Low countries fest. 826
2013 6 SR5 LLR Father's day 944
2013 8 TRP Sieve Perseid shower 1114
You can clearly see there's always way more than 200 participants. I'm not in favor of the whole getting a trophy just for participating lark, so that definitely is not going to happen. However, I am going to ponder on it a bit more to see what I'm going to do with scoring if anything. As before, you'll hear in due time.
To be thorough I've also checked the teamstats and this is the result of that:
Year M. Project Name Teams
2012 1 PPS LLR Year of the Dragon 449
2012 4 SGS LLR Marie-Sophie Germain 316
2012 5 TRP LLR Cinco de Mayo 224
2012 6 GFN Alan Turing Year 109
2012 8 TRP Sieve Olympics 154
2012 10 PPS LLR An Apple a Day 215
2012 11 321 LLR 321 Blast Off 244
2012 12 PPS Sieve Winter Solstice 297
2013 1 GFN Short Year of the Snake 166
2013 2 SGS Valentine's Day 180
2013 3 PPS LLR Equinox 365
2013 4 WOO LLR Low countries fest. 162
2013 6 TRP LLR Father's day 156
2013 8 TRP Sieve Perseid shower 206
Again, always more than the 100 teams that can get in a scoring position, however there aren't that many more that I'm going to spend any time on (thinking about) reworking the scores for teams. Those will stay the same.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
JaySend message
Joined: 27 Feb 10
Posts: 99
ID: 56067
Credit: 53,213,356
RAC: 840
                
|
|
I'd like to add my request for a SOB challenge. I don't think there has ever been one (at least for as long as I've been here). Also, although the project may run indefinitely if the conjecture doesn't work out, I believe there are only 6 left to find before the project is put to bed. My thinking is that this is a mathematical problem that can be solved (conjecture proven) in a relatively small amount of time with enough attention and effort made. To me that is a bigger deal than finding the next largest prime (of any type) which just ends up moving the goal posts as there will always be a "next largest." | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2781
ID: 130544
Credit: 934,442,218
RAC: 29,067
|
|
Definitely add a SoB challenge please. It's time to give this a go now. | |
|
|
|
|
All subprojects with left k are interesting with a higher chance to eliminate some more k`s. Also if ESP can move to BOINC that could be a big benefit. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
... I believe there are only 6 left to find before the project is put to bed. My thinking is that this is a mathematical problem that can be solved (conjecture proven) in a relatively small amount of time with enough attention and effort made. To me that is a bigger deal than finding the next largest prime (of any type) which just ends up moving the goal posts as there will always be a "next largest."
While I agree completely with the importance of the conjecture projects, I'd like to point out that while there are only 6 k's left of the original 17, in the almost 3 years (I think) that SoB has been running here at PrimeGrid, there have been no primes found either by PrimeGrid or by the Seventeen or Bust project. In fact, it's been nearly 6 years since the last Sierpiniski Problem prime was found, in October of 2007. The largest prime found has an n value of about 13 million (and the next largest was 9 million), and we're currently at 21 million and have not yet found a single prime.
That being said, challenges DO bring a huge amount of processing power to the table and this is certainly a project that would benefit from that. The logistical problems, however, are significant.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
17 or bust isn't going to be "put to bed" any time soon. It could go on for centuries.
Don't get me wrong. I love 17 or bust. It's a mathematical scavenger hunt! However, realistically, the last n value could be in the billions or even trillions. We just don't know.
It would be nice to get it down to 5 k values though.
At the Riesel Problem is moving along again... Two primes from the recent challenge! | |
|
JaySend message
Joined: 27 Feb 10
Posts: 99
ID: 56067
Credit: 53,213,356
RAC: 840
|
17 or bust isn't going to be "put to bed" any time soon. It could go on for centuries.
I certainly don't expect it to be finished "soon" as in months, or with 2-3 challenges dedicated to it or anything like that. I hope though that with advances in computing power, maybe new mathematical theories or approaches, it can be "solved" within a decade.
Don't get me wrong. I love 17 or bust. It's a mathematical scavenger hunt! However, realistically, the last n value could be in the billions or even trillions. We just don't know.
I agree completely, especially with "We just don't know." Realistically, the last 6 we need could be the next 6 that the server hands out. Or, like you said, "in the billions or even trillions."
It would be nice to get it down to 5 k values though.
I'd describe it as better than "nice." Have any of the conjecture projects had a challenge that focused on a single remaining k? Maybe I'm mistaken, but shouldn't concentrating on one of the 6 remaining (instead of equally going after all 6 at once) mean finding a prime about 83% faster (statistically)? Would it be at all beneficial to concentrate on eliminating one of the pesky remaining, almost as a proof of concept, to show that the method is working? I'll admit to sometimes wondering if the method is sound and without holes given the amount of time with no success.
At the Riesel Problem is moving along again... Two primes from the recent challenge!
That's fantastic! Congratulations to all involved!
| |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2781
ID: 130544
Credit: 934,442,218
RAC: 29,067
|
Maybe I'm mistaken, but shouldn't concentrating on one of the 6 remaining (instead of equally going after all 6 at once) mean finding a prime about 83% faster (statistically)? Would it be at all beneficial to concentrate on eliminating one of the pesky remaining, almost as a proof of concept, to show that the method is working?
I like this idea. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
Maybe I'm mistaken, but shouldn't concentrating on one of the 6 remaining (instead of equally going after all 6 at once) mean finding a prime about 83% faster (statistically)? Would it be at all beneficial to concentrate on eliminating one of the pesky remaining, almost as a proof of concept, to show that the method is working?
The opposite is true. The fastest way to find primes, statistically, is to crunch all the k's simultaneously with the lowest n values.
Since processing time is proportional to the square of the size of the number, and the primes also get rarer as the number gets bigger, we're more likely to find *a* prime within a given time period by searching them in parallel at the smallest n values.
However, taking the long view, since the goal is to find ALL the primes, it doesn't matter what order we search in. There's X amount of processing to be done no matter which order we do it in. That's true even if the conjecture is false and X is infinite.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
Just a question: IF a PSP challenge was scheduled, and let's say that two K values were eliminated, would that not A) Be good in itself, but B) un-suspend the PSP sieve project? I'm not sure there's ever been a PSP challenge, and since I joined PG 2 and a half years ago, I've not had the chance to participate in PSP sieving. A challenge would tick both of those boxes. Just an idea - PSP WUs would be more friendly to slower CPUs than SoB WUs, too... And I don't think updated sieve files for TRP or PSP have been condidered yet in this thread. They seem like very important factors (pun intended) though, and should be taken into account surely?
And for the record, I am still up for an SoB challenge. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
Just a question: IF a PSP challenge was scheduled, and let's say that two K values were eliminated, would that not A) Be good in itself, but B) un-suspend the PSP sieve project? I'm not sure there's ever been a PSP challenge, and since I joined PG 2 and a half years ago, I've not had the chance to participate in PSP sieving. A challenge would tick both of those boxes. Just an idea - PSP WUs would be more friendly to slower CPUs than SoB WUs, too... And I don't think updated sieve files for TRP or PSP have been condidered yet in this thread. They seem like very important factors (pun intended) though, and should be taken into account surely?
And for the record, I am still up for an SoB challenge.
I'm not sure if finding more k's would mean the sieving would be reopened. However, I do know you're wrong about there never having been a PSP challenge. We had one in September a few years ago. I remember it because the cleanup didn't finish until January or February, which meant the final challenge results for that year's entire challenge series weren't available until a month or two into the next year. This is why we now run any long-task challenges at the beginning of the year.
____________
My lucky number is 75898524288+1 | |
|
|
|
Just a question: IF a PSP challenge was scheduled, and let's say that two K values were eliminated, would that not A) Be good in itself, but B) un-suspend the PSP sieve project? I'm not sure there's ever been a PSP challenge, and since I joined PG 2 and a half years ago, I've not had the chance to participate in PSP sieving. A challenge would tick both of those boxes. Just an idea - PSP WUs would be more friendly to slower CPUs than SoB WUs, too... And I don't think updated sieve files for TRP or PSP have been condidered yet in this thread. They seem like very important factors (pun intended) though, and should be taken into account surely?
And for the record, I am still up for an SoB challenge.
Answer for A - Of course it would be good in itself, but it could also mean that we're in for a long n-range where no primes excists. Primes tend to come in pairs, so it is not unlikely that the next 2 PSP primes will be relatively close to each other (n-wise)
Answer for B - If we find 2 primes, it also means that the SOB/PSP combined sieve-file will have only 12 k's in stead of 14 k's in the sievefile. The reduced amount of k's will in itself mean that we will see a speed-increase, but the reduced amount of candidates in the sieve-file, will unfortunantly mean that the amount of possible factors will go down. So finding 2 primes will not make a significant difference to weather it is usefull or not to resume sieving.
However, something that I have come up with, may actually give the possibility to sieve "a lot further" as soon as Rogue adds the function I've suggested to him, so sr2sieve. I haven't done any significant testing, but initial study shows that many factorcandidates can be excluded, simply because they doesn't meat the criteria of the most (90 % +) factors. Unfortunantly the amount of candidates that can be excluded seems to decrease untill we reaches 0% and by then we will have to jump to the next sievelevel, i.e. if we sieved from p=10^18 we would then have to jump to p=10^19 and sieve untill optimal sievedepth :)
Well let's see if sieving is such productive as I hope, once Rogue adds the new method to sr2sieve. Unfortunantly I doesn't have a clue as to the status of developing and adding of my sievemethod to sr2sieve. But as meantioned, there may be light at the end of the tunnel when it comes to the possibility of resuming the SOB/PSP combined sieve :)
Regards
KEP | |
|
|
|
|
TENTATIVE schedule
There are some notes to the whole story, some challenges even have special notes all of their own.
General notes:
[*] challenge credits will be reworked. I'll do this when the schedule has been finalized.
[*] it's tentative, which means I can (and most likely will) change a bunch of stuff.
January 3 - 18 GFN Long* -- 15 days
January 3 - 18 GFN Short* -- 15 days
February 14 - March 1 SOB -- 15 days
April 1 - 4 PPS LLR -- 3 days
May** - Cullen -- 8 days
June** - SR5 LLR -- 3 days
July/August/September** - TRP Sieve -- 2 days
October 4 - 7 - TRP LLR -- 3 days
November** - 321 LLR -- 5 days
December 4-5 - SGS LLR -- 1 day
December 18-21 -- PPS Sieve -- 3 days
Special notes:
* - These will be two separate challenges, each with their own score board. As from a project point of view it is much more interesting to increase the power on the long ones, I'm contemplating awarding a higher challenge score for the same position in that challenge.
** Exact dates still need to be figured out. If there's any interesting festivity in your country that you'd like to have as a theme please feel free to suggest it. Any and all suggestions will be taken into consideration, though that is no guarantee I will actually use it.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
|
Thanks for the proposed list! I'd still like to see something for Shakespeare's 450th birthday (April 23) and Mother's Day (May 11th, at least in the USA), but I'm sure the PRPnet challenge schedule can accommodate that if BOINC doesn't.
BTW, would it be better to run the TRP Sieve challenge *before* the TRP LLR challenge, so as to remove more candidates beforehand? Or does it not work like that?
Good to see SR5 and Cullen (finally!) in there.
--Gary
____________
"I am he as you are he as you are me and we are all together"
87*2^3496188+1 is prime! (1052460 digits)
4 is not prime! (1 digit) | |
|
|
|
|
Makes sense. (But part of your post won't after my edit ;)) I'll switch around SR5 and TRP LLR.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
pschoeferVolunteer developer
Volunteer tester
Send message
Joined: 20 Sep 05
Posts: 661
ID: 845
Credit: 2,161,397,878
RAC: 110,554
           
|
|
Well, I liked 13 days as maximum length, but we had even longer challenges in the past, so I can live with 15 days. I like the idea of running two challenges simultaneously, as it's something completely new after 6 years of challenges.
I'd like to see the May challenge taking place near the end of the month, so that it doesn't collide with the 5th BOINC Pentathlon (May 5 to May 19), otherwise you might not get that much contribution by some of the major BOINC teams. As there is new moon on May 28 (19:43 UTC), one idea would be a "Race to New Moon" challenge starting on May 20.
Although we don't search for Mersenne primes, we could celebrate Marin Mersenne's 525th birthday on September 8.
November challenge should be at the beginning of the month, so that we have some break before the two December challenges.
____________
  | |
|
|
|
|
Just a couple of cents from me.
I think the minimum challenge duration should be at least 3 days. Since a start of challenge is at different times in different parts of the world, we don't give a chance to get a good position to those people who start late just because they were at work/sleeping late at night and so on.
Also challenges give a great push to subproject so it's worth making it a little bit longer to the most out of it.
I would prefer a minimum 5-day challenge.
TRP LLR are getting longer, so I think 3 days are not sufficient. My Core2Duo laptop will only be able to finish 1 WU per core in the challenge, if 'n' won't have grown really high by that time.
Cheers | |
|
JaySend message
Joined: 27 Feb 10
Posts: 99
ID: 56067
Credit: 53,213,356
RAC: 840
|
|
Thank you for including SoB in the tentative schedule. I hope that it remains in through whatever process happens before the schedule is finalized. A fair number of people in this thread expressed support for a SoB challenge and I appreciate that it was taken into account and included. Thanks again.
Jay | |
|
|
|
|
After lots of fiddling, here's the schedule for next year.
Individual scoring positions will be increased to 300 positions, team scoring positions will remain the same. Please stay tuned for the exact numbers, as I am now going to figure out how to display a 8x300 table nicely here.
2014 Challenge Series
Date start Date end Time UTC Project Challenge Name Duration
1 3-Jan-2014 18-Jan-2014 18:00:00 GFN Long Year of the Horse - Stallion Edition 15 days
3-Jan-2014 18-Jan-2014 18:00:00 GFN Short Year of the Horse - Pony Express
2 1-Mar-2014 16-Mar-2014 18:00:00 SOB --PENDING-- 15 days
3 20-Apr-2014 23-Apr-2014 16:16:00 PPS LLR Shakespeare's birthday 3 days
4 20-May-2014 28-May-2014 19:43:00 CUL LLR Race to the New Moon 8 days
5 25-Jun-2014 28-Jun-2014 18:00:00 SR5 Sovereign Challenge 3 days
6 12-Aug-2014 14-Aug-2014 18:00:00 TRP Sieve Perseid shower 2 days
7 4-Oct-2014 7-Oct-2014 18:00:00 TRP LLR World animal day 3 days
8 14-Nov-2014 19-Nov-2014 18:00:00 321 LLR Leonids 5 days
9 5-Dec-2014 5-Dec-2014 0:00:00 SGS Saint Nicholas' Challenge 24 hours
10 18-Dec-2014 21-Dec-2014 17:11:00 PPS Sieve Winter Solstice 3 days
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
|
Conclusion: there's no good way to display a table like that on the boards, so here's a link to them. Individual scores and Team scores
Edit: most important to note: the GFN WR challenge will give 50% extra challenge credit and the first 300 rather 200 people will score on all challenges.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
|
Since the GFN WR challenge is going to happen, shouldn't we focus on GFN WR sieving before the challenge to increase our chances of finding a prime? | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
Since the GFN WR challenge is going to happen, shouldn't we focus on GFN WR sieving before the challenge to increase our chances of finding a prime?
It's not necessary. Although we're not at the optimal sieving point yet (in fact, no where even close to it), we're still deep enough into the sieve so that sieving, at this point, doesn't eliminate candidates very frequently. The chance of eliminating a candidate we'll be testing during the challenge is very small.
Or, to put it another way, if we were considering just the range we're likely to be testing, the optimal point to stop sieving would be in the neighborhood of about P=10E. We've currently sieved to a depth of P=104.2E. (The optimal sieve depth for the whole WR search is about P=1.7Z).
____________
My lucky number is 75898524288+1 | |
|
|
|
|
After some discussion I've moved the SoB challenge so that it doesn't overlap with the Tour de Primes. It now starts on the first of March, rather then ending at that date. The theme is still pending. Suggestions are welcome.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
The theme is still pending. Suggestions are welcome.
Finishing one day earlier, it could be the Ides of March challenge. One day later, would be St. Patrick's day challenge. Finishing on the 17th but also moving the start earlier a day to 1800 UTC on Feb. 28th (which wouldn't technically conflict with prpnet) would make it a Seventeen day challenge on Seventeen or Bust which finished on the Seventeenth of the month. Just a thought.
--Gary | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2781
ID: 130544
Credit: 934,442,218
RAC: 29,067
|
|
I like it. 17 Days or Bust! | |
|
|
|
|
We have no yet Challenge Points Schemas for 17 or even 15 days challenges.
Only 1, 2, 3, 5, 8, 13 days durations are available.
1, 2, 3, 5, 8, 13 are the first 6th members of Fibonacci sequence.
(Fibonacci began the sequence not with 0, 1, 1, 2, as modern mathematicians do but with 1, 2, etc.)
I guess you guess what I mean. ;)
____________
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
We have no yet Challenge Points Schemas for 17 or even 15 days challenges.
Pyrus posted the schedule for 15 day challenge points in a post just a few days ago (about 6 posts before yours). Furthermore, it's relatively trivial to invent challenge point charts for any length challenge. That will never be an impediment to running a challenge of whatever length we deem to be appropriate.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
Oh, excuse me, I missed them.
Anyway, Fibonacci resents!!! ))
____________
| |
|
pschoeferVolunteer developer
Volunteer tester
Send message
Joined: 20 Sep 05
Posts: 661
ID: 845
Credit: 2,161,397,878
RAC: 110,554
|
|
Thanks for the schedule.
Shouldn't the start/end time for next year's Winter Solstice Challenge be 23:03 UTC instead of this year's 17:11? :)
____________
| |
|
Sysadm@NbgVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1181
ID: 18646
Credit: 477,508,055
RAC: 9,031
            
|
...
2014 Challenge Series
Date start Date end Time UTC Project Challenge Name Duration
2 1-Mar-2014 16-Mar-2014 18:00:00 SOB ...
On our own behalf, please do not call it "Sysadm@Nbg's Obvious Birthday" ;-)
____________
Sysadm@Nbg
my current lucky number: 3749*2^1555697+1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
Thanks for the schedule.
Shouldn't the start/end time for next year's Winter Solstice Challenge be 23:03 UTC instead of this year's 17:11? :)
So it should. Consider it changed.
____________
My lucky number is 75898524288+1 | |
|
Ken_g6Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 908
ID: 3110
Credit: 176,867,532
RAC: 2,248
         
|
|
For that unnamed SoB race, if you're really doing it, what about calling it the Caesar's Bust challenge? Because it ends near the Ides of March.
But can an un-overclocked Core 2 Quad really do SoB work in 15 days? I'm having to suspend one of four cores to get GFN work done in 15 days.
____________
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
For that unnamed SoB race, if you're really doing it, what about calling it the Caesar's Bust challenge? Because it ends near the Ides of March.
But can an un-overclocked Core 2 Quad really do SoB work in 15 days? I'm having to suspend one of four cores to get GFN work done in 15 days.
That's a good question, especially since my computer is a Q6600.
I guess a good plan would be to start one task, see how long it will run for, start a second, see how long that one will run, and so forth, until you get to four. If needed, abort one of more if running multiple tasks pushes them over 15 days.
This will be very computer specific, and some systems with Q6600s may have no trouble with running LLR on all 4 cores in 15 days, while others might not. It's the entire system performance, including the main memory, that's important.
The question is really, can YOUR computer run it in 15 days?
For what it's worth, I find that right now, my system, running one core, is maybe 10% slower than the PrimeGrid-wide average CPU run time for most tasks. The average run time for SoB is 225 hours, so I would expect my system to run it in about 250. 15 days is 360 hours. The big question is how much will the system slow down with 4 cores running, and there's only one way to find out.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
Once again, I'm asking for your opinions :)
We're currently nearing the end of our current set of Seventeen or Bust work units. This will have a significant effect on the Double Top Secret Mystery challenge. It looks like we'll be getting a new set of numbers, but these will be about 65% longer than the current ones. That means about 16 days on average per work unit, for the average system. This leaves us with several options:
1) Leave everything as is, which means excluding a lot of people a lot of people.
2) Make the challenge a lot longer (28 days). This of course comes with an increase in challenge points.
3) Switch challenge project over to PSP and leave the length at 15 days.
I have my own opinion of course, but I'm curious to hear what you all prefer. Please also include why you prefer the option you prefer
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
2) Make the challenge a lot longer (28 days). This of course comes with an increase in challenge points.
I like this one. I think a longer challenge would be a nice change of pace! It might favor those who have a large number of machines, but then again if some of those only procure the large numbers on a temporary basis to compete in challenges then it could have interesting results :) | |
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1884
ID: 352
Credit: 3,037,812,339
RAC: 83,518
|
|
Charley, if 1) means load new range and leave challendge for 15 days, I'm not in favour of this. Since BOINC is not good in estimating time and some participants will miss that SoB would take 2/3 longer, they will miss deadlines...which is unfortunate after 15 days of crunching.
2) means that GFN and SoB challenges will mostly determine 2014 positions. PG 2013 Challenge was 46 days in Total, 2014 was designed to be 60 days now possibly 73 days. It is considered too much for some teams.
That sayd, I'm inclined to option 3 - keep the dates, switch to PSP.
Personally, I'll stick with SoB for the moment - still a bit to go for Sapphire. On the other hand - PSP can be upgraded as well :-)
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 | |
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2125
ID: 1178
Credit: 8,404,284,720
RAC: 540,645
|
|
I prefer option #3.
Option #1 excludes too many people, and we already did that with one-half of the first challenge of the year.
Option #2 is too long for a single project. It will slow our progress for the year on other projects. I also agree with Honza that it pretty much would mean that GFN and SoB would determine the challenge champion for the year (which might reduce interests in the remaining challenges).
Option #3 avoids all of this, and thus, is in my opinion the best choice.
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
|
I have my own opinions, which I'll keep to myself, but I'd like to point out that all of the arguments so far for not choosing option 2 (which reduce down to "28 days is just too long") imply that we'll never run an SoB challenge since that situation is not likely to change in the future.
____________
My lucky number is 75898524288+1 | |
|
tngSend message
Joined: 29 Aug 10
Posts: 384
ID: 66603
Credit: 21,651,663,290
RAC: 760,526
                        
|
I have my own opinions, which I'll keep to myself, but I'd like to point out that all of the arguments so far for not choosing option 2 (which reduce down to "28 days is just too long") imply that we'll never run an SoB challenge since that situation is not likely to change in the future.
Agreed, and I would like to see an SoB challenge. I would be OK with options 1 or 2, but option 1 will probably impact a lot of systems. I say go with option 2 so that we can still have an SoB challenge and let most of those interested have a chance to participate.
____________
 | |
|
|
|
|
Let's go for option 2, it gives a great push to the most advanced conjecture and it will propably leave the last SoB conjecture with a fair and great impact, that we can always look back on. Who knows, we may be able to push it hundreds of thousands over the 28 days rather than "just" 120000 n's. Also we might be able to find an SoB prime, because it's been a while and we are propably very close to finding an SoB prime, so we might just be lucky and knock down one of the remaining 6 k's :)
So let's extend to 28 days, I'll be able to do 8-12 WU on my 1 year old I5, dependent on the k size. If we just focus on the lowest k, I'll maybe be able to do 16 WU, however I think with the thousands of users joining the challenge, that the best option is to test by incrementing n and not just focusing on a specifik k :)
Also to those who might think and feel that this is too long a challenge, make no mistake, no one is forcing you to stay any longer than you feel like, but of course any support is greatly appreciated. Just my 2 cents, thank you :)
Take care
Kenneth | |
|
|
|
|
I favor option 3 (switch to PSP, keep the dates the same). There's only been one PSP challenge in my recollection, and it was late in the year which caused some problems with the clean-up and year-end scoring. But if we do it in March, that shouldn't be an issue. Plus PSP is also a "conjecture" project, which I favor (not to say that SoB isn't one, of course). But SoB WUs take something like 3.5 days on the fastest cores now, and it sounds like that will jump to 5.5 to 6 days. The timeout deadline will have to be increased accordingly, so the clean-up will take a *really* long time... I'd bet longer than the GFN-WR one will take.
I think a month-long traditional challenge is too long. You can accuse me of wanting "instant gratification" and I will plead guilty as charged. I used to enjoy the 1-day challenges on the PRPnet "low" port, where work units took something like 40 seconds each. :-)
I would not mind if there is never a SoB challenge. I like the project and run it on occasion. As an incentive to users to run it (if such incentive is desired), maybe grant bonus challenge credit or a special badge based on the number of valid SoB WUs returned during the entire year? (that idea is only half-baked, I know, but I thought I'd put it out there)
--Gary
____________
"I am he as you are he as you are me and we are all together"
87*2^3496188+1 is prime! (1052460 digits)
4 is not prime! (1 digit) | |
|
|
|
|
I favor 2.
Many teams choose Projects of the Month and stick on it the whole month, I am betting a lot would then make PG the PotM and make the challenge fun. It also allows us to get the new numbers a little boost in percentage done.
I agree, 1 excludes too many.
3 is my second choice, but I may not stick to it personally. I would stick to the SoB one.
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1 | |
|
|
|
|
I favor #2 - move SOB to 28 days
#1 - removes too many people/computers. I think I would only have two or three computers (at reduced cores) capable of completing these new longer wus during the challenge. Many others would not have any computers capable so can't participate at all.
#3 - While this would keep the challenge at 15 days we have had a PSP challenge more recently and the task length is not growing as fast so we could do a 15 day PSP next year.
#2 - While 28 days is a very long challenge and we will most likely be running cleanup work into October I think this may be the best chance to run an SOB challenge and many people expressed interest in seeing a SOB challenge. For scoring it puts some more CPU points out to offset the GPU heavy GFN-WR + GFN.
____________
 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2781
ID: 130544
Credit: 934,442,218
RAC: 29,067
|
|
Definitely 2, though how much I participate personally may depend on personal circumstances. | |
|
|
|
Once again, I'm asking for your opinions :)
1) Leave everything as is, which means excluding a lot of people a lot of people.
2) Make the challenge a lot longer (28 days). This of course comes with an increase in challenge points.
3) Switch challenge project over to PSP and leave the length at 15 days.
I prefer 3.
Option 1 pretty much limits it to Intel-AVX only for any chance of more than 1 WU / core.
Option 2 is too darn long. (IMHO)
| |
|
Ken_g6Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 908
ID: 3110
Credit: 176,867,532
RAC: 2,248
|
I have my own opinions, which I'll keep to myself, but I'd like to point out that all of the arguments so far for not choosing option 2 (which reduce down to "28 days is just too long") imply that we'll never run an SoB challenge since that situation is not likely to change in the future.
Michael, just for a sense of scale, how long would your C2Q 6600 take to complete four of these new SoB WUs? How long would it take to complete one on its own?
Also, once upon a time, there was discussion of using llrCUDA for SoB. I believe v0.60 (Win64 binary) is the best version that could do it. Note that it never worked for 32-bit binaries. I could never get it working as a BOINC binary, but it seems like someone did with the LLR wrapper. So:
1. What are the chances of getting an official app using llrCUDA 0.60, for Windows, Linux, and Mac, for SoB? (I assume very small.)
2. What are the chances of getting a semi-official set of llrCUDA 0.60 binaries and app_info files, for Windows, Linux, and Mac, for SoB? (I'm hoping better?)
My opinion is that 28 days is too long - 15 days was too long - but I would like to be able to run a SoB race. So I'm grasping a straws for ways to make option 1 work.
____________
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13398
ID: 53948
Credit: 229,092,028
RAC: 11,702
|
[quote]I have my own opinions, which I'll keep to myself, but I'd like to point out that all of the arguments so far for not choosing option 2 (which reduce down to "28 days is just too long") imply that we'll never run an SoB challenge since that situation is not likely to change in the future.
Michael, just for a sense of scale, how long would your C2Q 6600 take to complete four of these new SoB WUs? How long would it take to complete one on its own?
Funny you should ask that. I did a benchmark test of SoB on my 1st gen Core i3 laptop (the Q6600 suffered a hardware failure) and this is the result:
67607*2^21388571+1: 971,662 seconds
67607*2^27000001+1: 1,623,349 seconds
That's a 67% increase. (The Q6600 was somewhat faster.)
Also, once upon a time, there was discussion of using llrCUDA for SoB. I believe v0.60 (Win64 binary) is the best version that could do it. Note that it never worked for 32-bit binaries. I could never get it working as a BOINC binary, but it seems like someone did with the LLR wrapper.
The last time I looked at llrCUDA it was only about twice as fast as a non-AVX CPU core and had reliability problems (likely the same overclocking problem that GeneferCUDA has.) I'm not aware of any progress since then.
I'm pretty sure it works with the same LLR wrapper since I recall testing llrCUDA with app_info. While I don't consider llrCUDA worth using, there's nothing stopping anyone from running it with app_info if they desire. It's possible the wrapper needed minor modifications, but I don't really remember.
So:
1. What are the chances of getting an official app using llrCUDA 0.60, for Windows, Linux, and Mac, for SoB? (I assume very small.)
2. What are the chances of getting a semi-official set of llrCUDA 0.60 binaries and app_info files, for Windows, Linux, and Mac, for SoB? (I'm hoping better?)
1) Zero, unless there's significant improvement in llrCUDA. If you're talking about in time for the SoB challenge, it's zero. 2) From me? Zero. As you know, I don't have a development machine right now. There's nothing stopping anyone else from making the binaries and app_info available if they so desire. None of this is relevant to the challenge, however.
My opinion is that 28 days is too long - 15 days was too long - but I would like to be able to run a SoB race. So I'm grasping a straws for ways to make option 1 work.
What we could use is a multi-core version of LLR. (Ditto for Genefer and PFGW). It wouldn't be as efficient, but it might be worth it in order to crunch these large tasks.
____________
My lucky number is 75898524288+1 | |
|
|
|
... there's nothing stopping anyone from running it with app_info if they desire ...
[Rant]
Snap goes the lock - creak go the hinges.
Ever hear of Pandora's box?
In TheDawgz "not even close to humble" opinion - the two worst features of BOINC ... Adaptive Replication and app_info.
[/Rant]
____________
There's someone in our head but it's not us. | |
|
|
|
|
I'm in favor of (3) because I still need to push PSP to Amethyst while SoB is already there.
____________
| |
|
|
|
|
Thanks for all your input people, it was really valuable :) It tells me a lot of people are really interested in running a SoB challenge as well as a PSP challenge, as equal amounts of people prefered those. This also means I am going to disappoint people, no matter which way I go. After careful deliberation I've decided to go with PSP and leave it at 15 days this year. I will however give serious consideration to a long SoB challenge when drafting next years schedule.
Now, let me explain why I have decided this as there are a bunch of good reasons, other than the "It's too damn long!" argument many of you (including me) would give.
We are going to run larger numbers through LLR than anyone has ever done before to the best of our knowledge. Having a challenge on the switching point is a great way to increase both awareness and throughput, but as we are entering uncharted teritory we also might stumble upon unforeseen problems. Although they are undesirable at any point, having those during a challenge is the worst as it will leave a bad impression with the many visitors we have at that point. With what little time we have left before the challenge, we will not get enough units out and back again for a good testrun so we can't guarantee a smooth challenge.
Furthermore, we would be dealing out a mix of long and short units. You can see this as part of the challenge, but it could also completely mess up peoples chances of coming into a scoring position. In the time left we also can't clean out all the short units to prevent this.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
| |
|
|
|
|
As always, very sound reasoning from PrimeGrid's administrators.
I hadn't expressed my preference, since I didn't really have one, but I was thinking: we had a long challenge with GFN this year (and this would probably be the only GFN challenge ever), so maybe the long challenge for next year could be SoB. But again: I bow to the superior knowledge and insight of PrimeGrid's administrators and we'll see what happens.
So this year: PSP it is! | |
|
|
|
|
Well, this decision at least leaves it an open challenge for more participants, since all levels of computer (at least those <5 years old) will now have a fair chance of complete a couple of workunits. Also PSP could really use a boost, not only to find a new prime, but also to catch up with SoB, so bring it on with the PSP challenge.
Anyone here that thinks we can complete 25000+ results? maybe even 50000+? With the shorter units and shorter challenge, I expect to be able to complete at least 10 results (and I only have one computer) :) | |
|
|
|
|
I have to say, the recent trend of people bringing a significant number of CPUs online only during challenge periods really sucks some of the fun out of the challenges.
It would be nice if there were some longer marathon challenges - say six month or even a year - that could be run in parallel with the challenges that happen now. This would encourage the folks who are in PrimeGrid for the long haul and not just spiking up for a short duration challenge to keep participating and also reward them too.
I'm not saying we need to get rid of the current challenges, but as I look over the results from the various challenges over the last couple of years, you can sure see a trend of pure "challenge crunchers" emerging :p
It could also be a great way to get some projects that I know the team would like to see more crunching on some significant attention. | |
|
|
|
|
I have a view. Seems to me like the current series are a bit of a war of attrition. How would it be if we reduced the number of challenges by say 50 or 75% and went at the remainders <expletive deleted> hard?
____________
Oh Bondage? Up Yours.
http://www.youtube.com/watch?v=ogypBUCb7DA
 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 2781
ID: 130544
Credit: 934,442,218
RAC: 29,067
|
|
I think what we have now is a good balance, plenty of interest & exercises strategic skills necessary for people fitting their processing priorities round challenge participation if required.
Certainly not appropriate for having challenges that are months long as that would be at the detriment to other projects. It has its appeal in a certain way - I think it would work as a team of people all wanting to hammer their subproject of choice, creating a 'virtual' challenge. Maybe custom teams could spring up. | |
|
|
|
|
If there are a lot of "challenge crunchers", it seems to me the administrators are reaching their goal: getting people to crunch for PrimeGrid, and especially for targeted subprojects.
If some users (sometimes including myself) have a feeling that the challenge standings are not a true reflection of the overall efforts, maybe we should realise there already is a triple form of reward for 'the long haul': credits, badges and, ultimately, primes!
Coming back to an older discussion in another thread: it would be a form of recognition if (existing) badges reflected the number of primes and/or their size. | |
|
Message boards :
Number crunching :
2014 PrimeGrid Challenge Series ideas |
