S385 tested to n=2.5k + sieved to 1G (2.5-10k)

7608 remain

Results emailed - Base released

S394 is complete to n=300K; no primes were found for n=200K-300K; 2 k's still remain; base released.

Trying new base R400; planning to take it to 10k.

[QUOTE=VBCurtis;559372]Trying new base R400; planning to take it to 10k.[/QUOTE]


Several CPU years worth of work.


I'm sure you're aware to remove k's that are squares. After removing k's with trivial factors there are 953 squared k's that need to be removed. There is a list of them in a link on the main page.

[QUOTE=gd_barnes;559393]Several CPU years worth of work.


I'm sure you're aware to remove k's that are squares. After removing k's with trivial factors there are 953 squared k's that need to be removed. There is a list of them in a link on the main page.[/QUOTE]
Thanks for the tips- I'll let you know pretty soon if I decide not to complete it to at least 2500 like other new bases. I was prepared for 1-2 thread-years, so your warning is well heeded that this is maybe too big for me.

Update: R327 complete to 625k. Still no primes, still running on 3 cores (one per k).

[QUOTE=VBCurtis;559372]Trying new base R400; planning to take it to 10k.[/QUOTE]

Are you using srbsieve or starting new bases script to start up R400?

It is a lot faster, to start a base, using srbsieve, compared to using the starting new bases script :smile:

You can find the latest version of srbsieve: [URL="https://mersenneforum.org/showpost.php?p=474389&postcount=273"]here[/URL]

You can find the nescessary .ini files: [URL="https://mersenneforum.org/showpost.php?p=422905&postcount=1"]here[/URL]

And the link in the last post states, reach out, if there is anything unclear. You are an experienced and skilled user, so I reckon that you will have an easy time understanding how to proceed. Must add, that you need to have srsieve in your file folder, aswell as pfgw.

Hi KEP! Thanks for your support and suggestions.

I'm using srbsieve for R400. I'm running the newpgen pre-sieves now, and then will run srbsieve on 1% of the k-range to get a sense of how lengthy this project is. I'll report back in a week or so with an un-reservation, or a projection of when I think I'll make it to n=2500 checkpoint.

[QUOTE=VBCurtis;559491]Hi KEP! Thanks for your support and suggestions.

I'm using srbsieve for R400. I'm running the newpgen pre-sieves now, and then will run srbsieve on 1% of the k-range to get a sense of how lengthy this project is. I'll report back in a week or so with an un-reservation, or a projection of when I think I'll make it to n=2500 checkpoint.[/QUOTE]

Good to hear. I'm not sure how srbsieve and srsieve handles the squared k's, but other than that, I can see that you have to go through 15 phases and the highest you have to sieve is: p=78,267,000 for (at most) 31458 k's. I'm not sure how fast your core is, but the sieving aswell as the running towards n=2500 could complete very fast. Sure am looking forward to see you estimation of completion :smile:

On a final sidenote, wich you may very well be aware, set the stop at p= value to 1e18 or maximum value (don't recall anymore) such that NewPGen stops once there is no more composite pairs in the sievefile - else you will produce bad results.

S354 is complete to n=300K; no primes were found for n=200K-300K; 3 k's still remain; base released.

[QUOTE=KEP;559554]Good to hear. I'm not sure how srbsieve and srsieve handles the squared k's, but other than that, I can see that you have to go through 15 phases and the highest you have to sieve is: p=78,267,000 for (at most) 31458 k's. I'm not sure how fast your core is, but the sieving aswell as the running towards n=2500 could complete very fast. Sure am looking forward to see you estimation of completion :smile:[/QUOTE]

I completed k's from 1 to 1M to n=2500 in 100 thread-hours on an old laptop, so I'm comfortable I have the patience to get R400 to 10k. I'll run srbsieve on the laptop in blocks of a few milllion k's, with surviving k's moved to a desktop to sieve and test n=2500-10000. I'm playing with srbsieve sieve & phase settings, so I may find some more speed; I'm trying 5 phases this time rather than 4 last time, for instance. 15 is far too many! Also, I sieve deeper than 78e6 on the last two phases; perhaps I'll try a bit less sieving to see if overall speed improves.

I expect 100-120 days to reach 2500, and less than twice that to reach 10,000.
I'll update when I get halfway to 2500.

There are about 5000 k's left from the first 1 million; with the conjecture just over 20M, I expect just under 100,000 k's left at n=2500 for the entire base.

[QUOTE=VBCurtis;559923]I expect 100-120 days to reach 2500, and less than twice that to reach 10,000.[/QUOTE]

I hope you're planning on adding a whole lot of resources for n=2500-10K. :-)

n=2500-10K is 3 times the n-range of n=1-2500.

A test at half the n-range for n=2500-10K (n=6250) would take 25 times as long as a test at half the n-range for n=1-2500 (n=1250) since the n-value is 5 times as big.

Extrapolating: 3 times the n-range x 25 times the testing time = 75 times as long for n=2500-10K vs. n=1-2500.

Allowing for a reduction in k's remaining on the higher range perhaps cut that by one-half to two-thirds so maybe 25-50 times as long.

I realize that this is overly simplified but I've found it to be a reasonable predictor of estimating a higher n-range after completing a lower n-range.

Perhaps I'm off base since you're going from 20 million k's at the beginning to 100,000 k's remaining at n=2500. But most of those 20 million k's are very quickly eliminated so I feel it's still a valid analysis.

My rough guess: If it takes you 4 months to complete to n=2500, with the same amount of resources it will take you (4x30* months= 120 months or) 10 years to complete to n=10K plus or minus a few years.

Feel free to stop at n=2500 if you'd like. :-)


*I chose 30 since it's on the lower end of the 25-50 times as long estimate.