After we find a twin, our next exponent will be...

[MooooMoo]

I made this poll in response to the growing number of posts relating to this topic.

The options are above, so feel free to vote. The options are public, because I'm only counting the votes of people who've contributed something to the project. The final votes will be counted after we find a twin.

Keep in mind that:

n= 250,000 takes 2.7 times as much computing power to find a twin than the current n.
n= 300,000 takes 5.6 times as much computing power
n= 350,000 takes 10.4 times as much computing power
n= 400,000 takes 17.7 times as much computing power

Personally, I'm against any n > 350K or n < 250K. Any n < 250K will mean that the project will take too long to move on, while any n > 350K will take too long to complete. I don't want this to resemble the RC5-72 project, which is only 0.4% done with its keyspace after 4 years.

I'll make my decision on whether to vote for n=250K-300K or n=300K-350K after a few weeks, when I've a better, more stable estimate of how much primegrid users are contributing to the project.

[axn]

Disclaimer:- I haven't crunched for this project.

I would imagine that looking for a 100,000 digit twin has a great psychological value. I especially like the "333333" proposed by biwema in the other thread.



PS:- Is there any way you can conduct a similar poll among the PrimeGrid user community (since I don't believe most of them are members of this forum)?

[Rytis]

Most people in PG don't have an idea what they are crunching. Besides, TPS project is not public in PG yet (only beta testing), so I think we'll choose the next n here.

[jmblazek]

Originally Posted by MooooMoo

The final votes will be counted after we find a twin..

Won't that be a bit too late...no presieved ranges to start the next twin prime search effort???

I will wait for a couple of weeks to vote so I can see the throughput of the current PG effort. As Rytis states it's only in beta stage right now...and as far as I can tell the 90% first pass/10% double check has not gone into effect yet.

Currently I'm leaning towards n= 333,333 suggested by biwema...100,000 digit, FFT change, etc.

MooooMoo mentions computing power. Something that would be interesting to know is what the computing power of TPS was before the beta PG effort was added (# users, # hosts, CPU yr/day, GFLOP/s). That way a more accurate assessment can be made.

For example, if the previous TPS effort was doing 100 GFLOP/s and now with beta PG it's doing 270 GFLOP/s then n= 250,000 would be like the previous effort.

I would suspect that the TPS effort now with beta PG has increased it's computing power by more than a factor of 10. When PG TPS goes active then who knows...maybe Rytis can give us a better view as to how much work is currently being done and by how many users and hosts.

I, too, don't want this to resemble the RC5-72 project...so....what's the computing power needed to make the next next n a reasonable effort.

p.s. would a distributed sieve effort make any impact on what the next n should be?

[Skligmund]

Figured since I dedicated all of my computer efforts to the TPS PG beta as of yesterday, I'd drop in and say I like the 333,333 number myself.

Gives me that warm, fuzzy feeling.

[MooooMoo]

Originally Posted by jmblazek

Won't that be a bit too late...no presieved ranges to start the next twin prime search effort???

Good point. The votes will be counted by the end of this year.

25 days to go...

[MooooMoo]

I got a PM yesterday suggesting that this estimated progress chart be posted:

Apr 13 Total 0M Monthly Average N/A
May 13 Total ~130M Monthly Average 130M
Jun 13 Total ~265M Monthly Average 135M
Jul 12 Total ~400M Monthly Average 135M
Aug 11 Total ~475M Monthly Average 75M
Sep 13 Total ~500M Monthly Average 25M
Oct 14 Total ~550M Monthly Average 50M
Nov 12 Total ~800M Monthly Average 250M
Dec 13 Total ~1450M Monthly Average 650M

Even if the current rate of 650M were doubled, it would still take almost another year to get to 15G. Finding a 100,000 digit twin takes ~8.5 times as much computing power as finding a 58,700 digit twin, so I'm not sure if n=333333 is a good idea.

[pacionet]

In my opinion we must wait that PrimeGrid finishes the beta testing and go to full )maybe Rytis decides to run only LLR).; at that time we know the real RATE.

[KEP]

Originally Posted by pacionet

In my opinion we must wait that PrimeGrid finishes the beta testing and go to full )maybe Rytis decides to run only LLR).; at that time we know the real RATE.

Actually I hope not, that he decides to do only LLR search, since I hope to see that we will some day have the biggest sequential prime database (which we according to my knowledge already have), going higher than 1,000,000,000,000,000. This will also take many years, but since that is the real no. 1 project, it could be nice to see that happen to. And also since cell technology is going to bring computers up to a pace of 250 GFlops, even to an affordeable price, I don't think that the goal is really that bad. Maybe also all the LLR users running Riesel Sieve, might actually discover our TPS when the Riesel Sieve prove is wrapped, and that might just bring us a further 1,000 + hosts, which will also be another strength on the servers as well as a huge boost to our current pool. So keep up the faith and spirit, and look deep into your heart, and you will see that "nothing is imposible for those carrying the will within their heart"

Regards!

[Skligmund]

Yeah, with all the dual-core this, and quad-core that, and the future looking to 8 cores (2008), I don't see a big issue. My next big computer upgrade will probably consist of a 4 core processor. BOOM! Like that, I have 4+ times the computing power than I have now with this one computer. I think multi-core is a big plus for this type of work, as long as it is managed correctly. I works out well with BOINC, as each processor has its own work unit. My old dual PIII 1.0 almost keeps up with my AthlonXP @ 2150, just because it does 2 at the same time.

Just a thought.

[jmblazek]

I'd like to see TPS find a record twin prime within 6 months...and then move on to the 100,000 digit twin prime. In order for this to happen I suggest we need the computing power to process 1G a week...(6 months is 26 weeks so that's 26G...which puts us in the ~85% probability range; in reality it's a bit higher because we've already completed ~1.45G).

Let's say the "average" host can compute 1/2M a day (erroring on the safe side). 1G a week is ~143M a day. We'll need 286 "average" hosts computing full time each day. Is that possible??? Absolutely! Is it probable? I don't know yet. What do y'all think?

Looking at the top 20, briefly glancing at their computers, it appears only 2 participants are running TPS right now (17th ranked Skligmund and 19th ranked Virus-X). Congrats to Skligmund for being the TOP TPS producer! However, it's a bit odd that KEP keeps finding all those primes on his single computer.

If just half of the top 20 switched to TPS we'd have our 286 "average" hosts of computing power...even if it's the 2nd half.

Right now it looks like PG TPS is doing 150M a week...we have a ways to go to get to a sustained 1G a week. Who knows, maybe PG TPS will become AA5...someone contact those crazy fun Aussies and invite them over here...they could use some extra work - they're ranked 31st. Boinc@Australia is always good at stirring up some competition. AA4 just finished today so they'll be looking for another project to assault. PG was actually in the voting for AA4 but Seti won out. At the time, they didn't know about TPS...however, they were concerned whether PG servers could handle the assault.

There are 38 participants with primes. I presume most people working on TPS now have found a prime which means we don't have too many more participants...let's say about 50. How do we get that number up?

Without taking too much away from PG's primary project, can we add enough computing power to TPS to reach 1G a week??? It will make selecting n=333333 a lot easier. (1G/wk at n=333333 will take 4 years to reach 85% probability of finding a twin prime - 2G/wk 2 years - 4G/wk 1 year). Now we need those quad & 8 cores!!! Break out the solar panels and wind turbines, we need more power!

p.s.

Maybe also all the LLR users running Riesel Sieve, might actually discover our TPS when the Riesel Sieve prove is wrapped

Riesel Sieve has 68 more k so they won't be wrapped up for several years.