Bases 251-500 reservations/statuses/primes

[KEP]

Originally Posted by em99010pepe

It's BOINC guilt because the people who run it don't understand nothing about what they are doing. They don't know they need to have a stable machine to test numbers, they just run BOINC for the stats, to help their teams climb in the stats. It's intrinsic. Of course with the amount of CPU power they have they can easily doublecheck 3, 4, 5 times but that's ridiculous. By 3 or 4 clicks you can easily change BOINC projects even when you are an ignorant on the matter. That should not be like that. People should understand what they are doing, study a little bit of LLR and prime stuff.
Do you want to know the lastes problem of BOINC? It's possible to hijack the teams...true, check here.

I admit, BOINC has its issues. However if bad residues or faulty results gets validated as "Valid" in stead of "Invalid" as they should be if the result doesn't match, then it still has nothing to do with the user, then it has something to do with the validator and the software running the validator. BOINC can never be underestimated for its power, and the need thing about BOINC is that you can always find somewhere to help out even if you have a faulty machine. Actually it appears that it is only LLR based projects, which can be hard to help if the machine gets faulty, however such projects, i.e. PrimeGrid and RieselSieve are supported by a great community where one gets several heads up, if one starts producing faulty results. I've tried it before myself, just before I lost my last machine, it started producing bad results because of a faulty PowerSupply and later it crashed the HDD. However due to the support among the community (and the halfing of results per day when one bad result is returned) I rather shortly stopped producing bad results. So I guess if people keep on eye on their results list and start asking into the reasons why red results show up, then a lot of mess can be avoided. But overall BOINC is no matter what skills one has a great and powerfull tool for science, give and take the few hick ups in security, though I must admit that since the past 5 years (since the beginning of BOINC) it has improved its safety a lot... but well whos perfect, I know I'm not and I guess no one else on this planet actually is perfect, its just a matter of being able to admit to it or not

Kenneth!

[gd_barnes]

9675*256^41822-1 is prime
7788*256^42163-1 is prime

Riesel base 256 is currently at n=44K; still going to n=75K. 40 k's are still remaining.


One month ago and the larger prime would have been top-5000.

3 months ago and they both would have been top-5000.

I should have started sooner. lol


Gary

[gd_barnes]

Riesel base 256 is now complete to n=50K. 2 primes for n=40K-50K previously reported.

Continuing on to n=75K after pausing for 2-3 weeks to continue Sierp base 12 from n=167K.

[gd_barnes]

Originally Posted by gd_barnes

Riesel base 256 is now complete to n=50K. 2 primes for n=40K-50K previously reported.

Continuing on to n=75K after pausing for 2-3 weeks to continue Sierp base 12 from n=167K.

I've now restarted Riesel base 256 from n=50K...still going to n=75K. With some extra fire power now, I'll keep on running base 12 and 256 and possibly reserve some base 16 stuff within the next few weeks.


Gary

[gd_barnes]

Riesel base 256 is now at n=54K. No new primes to report.

[gd_barnes]

Originally Posted by KEP

I've no idea exactly how long it is going to take, regarding the Sierp. base 19 but I guess around 2 weeks from now. Regarding Sierp. base 252 it is most likely only 2 days of work left on the Quad, after that it will most likely have progressed (if not finished) very far with the remaining ranges for Riesel base 3 k<=500M.

KEP!

Sierp base 252 for n=51K-100K is FAR more than 2 days work on a quad (i.e. 8 CPU days). What is your testing time per candidate at n=51K and how many candidates are remaining to be tested?

Keep in mind that a test at n=100K will take FOUR times as long as a test at n=50K.

If you give me the above info., I can, in effect, use the equivalent of compound interest formulas to give you a fairly accurate estimate of how long base 252 should take for n=51K-100K.


Gary

[KEP]

Originally Posted by gd_barnes

OK, for now I'll reserve sieving only to you on k=100M-200M for Sierp base 3.

Sierp base 252 for n=51K-100K is FAR more than 2 days work on a quad (i.e. 8 CPU days). What is your testing time per candidate at n=51K and how many candidates are remaining to be tested?

Keep in mind that a test at n=100K will take FOUR times as long as a test at n=50K.

If you give me the above info., I can, in effect, use the equivalent of compound interest formulas to give you a fairly accurate estimate of how long base 252 should take for n=51K-100K.


Gary

At n=52000 it takes 1045 sec per test. A total of ~2200 tests is left to test.

KEP

[gd_barnes]

Originally Posted by KEP

At n=52000 it takes 1045 sec per test. A total of ~2200 tests is left to test.

KEP

Expected time is 5.074M CPU secs. or ~58.7 CPU days. Running all 4 cores of a quad 24 hours a day 7 days a week will take ~14.7 calendar days.

A bit longer than 2 days.

Even if all tests took the same amount of time, it would be 1045*2200/86400 = 26.6 CPU days or 6.65 days on all 4 cores of a quad. But clearly all tests don't take the same amount of time.

To get to the original calculation, I assumed a constant rate of change in the n-values being tested and a constant SQUARED rate of change in the time each test would take.

So if n=52K takes 1045 secs., then n=52K*2=104K would take 1045*4=4180 secs. Also:
n = 52K * sqrt(2) = 73.5K would take 1045*2=2090 secs.

Obviously LLR time goes up in fits and spurts with fftlen changes so this can only be said to be a rough estimate. But it should be in the ball park since in the long run, LLR times varies with the square of the exponent.


Gary

[KEP]

Reserving Riesel base 255 for a future "Riesel base 255 attack". I expect it to overtake the "Riesel base 3 attack" website. I've already begun some initial testings and it seems very prime dense, so I will have no problem running this conjecture up to PG level, if Rytis wanna help out. Expect by no terms any further work from me beyound k<=500M to n<=25K for Riesel base 3 conjecture. It turns out to be to much of a handfull for me to handle, I still like the conjectures, so therefor I'm doing the initial preparations to launch a fullscale attack on the Riesel base 255 conjecture

Hope no one mind. Also I'm gonna need something to keep my computers busy as they slowly runs out of work for previously reservations

Regards

KEP!

[gd_barnes]

Originally Posted by KEP

Reserving Riesel base 255 for a future "Riesel base 255 attack". I expect it to overtake the "Riesel base 3 attack" website. I've already begun some initial testings and it seems very prime dense, so I will have no problem running this conjecture up to PG level, if Rytis wanna help out. Expect by no terms any further work from me beyound k<=500M to n<=25K for Riesel base 3 conjecture. It turns out to be to much of a handfull for me to handle, I still like the conjectures, so therefor I'm doing the initial preparations to launch a fullscale attack on the Riesel base 255 conjecture

Hope no one mind. Also I'm gonna need something to keep my computers busy as they slowly runs out of work for previously reservations

Regards

KEP!

This is a multi CPU-year effort to test such a high base with a high conjecture up to PrimeGrid level, regardless of how prime the base is. I'm assuming that PrimeGrid level would be defined as tests that would yield top-5000 primes. For base 255, that would be at n=42K-43K now; higher when you get up that far later.

Didn't you test Sierp base 255 up to n=2500? Do you remember how long it took for you to get that tested? Now, you're going to test the Riesel side with a conjecture that is twice as high?

You are correct, it seems that all bases (b) where b=2^q-1, i.e. 3, 7, 15, 31, etc. are prime dense but because they are so primeful, they also have very high conjectures vs. their neighbor bases, which makes them more difficult to prove then a large majority of other bases.

I have 3 alternatives for you if you want to test base 255 up to PrimeGrid level in a reasonable amount of time:
(a) Do Sierp base 255 instead. It's already at n=2500 and the conjecture is half as high.
(b) Start a new thread here at CRUS and attempt to get some help testing Riesel base 255.
(c) Buy several new quads. (lol)

Even if you do (a), you'll likely still need to enlist some help here at some point.


Gary

[KEP]

Originally Posted by gd_barnes

This is a multi CPU-year effort to test such a high base with a high conjecture up to PrimeGrid level, regardless of how prime the base is. I'm assuming that PrimeGrid level would be defined as tests that would yield top-5000 primes. For base 255, that would be at n=42K-43K now; higher when you get up that far later.

Didn't you test Sierp base 255 up to n=2500? Do you remember how long it took for you to get that tested? Now, you're going to test the Riesel side with a conjecture that is twice as high?

You are correct, it seems that all bases (b) where b=2^q-1, i.e. 3, 7, 15, 31, etc. are prime dense but because they are so primeful, they also have very high conjectures vs. their neighbor bases, which makes them more difficult to prove then a large majority of other bases.

I have 3 alternatives for you if you want to test base 255 up to PrimeGrid level in a reasonable amount of time:
(a) Do Sierp base 255 instead. It's already at n=2500 and the conjecture is half as high.
(b) Start a new thread here at CRUS and attempt to get some help testing Riesel base 255.
(c) Buy several new quads. (lol)

Even if you do (a), you'll likely still need to enlist some help here at some point.


Gary

I'm not going to reserve any of these bases anyway. Even taking it to n<=25000 is more work than I really feels like doing anymore... so I'm wrapping my final works (8 weeks to go at least) amd then I'll decide what (if any) to reserve when that time comes. Sorry for yabbing again, but I made a mistake by simply forgetting how many more weeks of work I've left. Hope you understand and accept my appology... Anyway base 19 Sierp is going to be wrapped somewhere close to next weekend (maybe already the coming weekend)

Take care everyone and happy crunching

KEP