Primo

[wpolly]

Originally Posted by rudy235

Any great candidate >38000 digits there?

I guess R49081 is always there, but I'm not sure if it's feasible now....

[Batalov]

Plenty of accessible unique numbers

Phi(34051,-10)    32033    Ray Chandler    04/2009
Phi(36547,-10)    29832    Ray Chandler    04/2009
Phi(35421,-10)    23613    Ray Chandler    04/2009
Phi(32481,10)     21600    Matthew Peets    11/2008
Phi(39855,-10)    21248    Ray Chandler    04/2009

[paulunderwood]

I use a 4x 12 core 2.2 GHz AMD 6174. I imagine a 2x 32 EPYC cores (64 threads each); One of these wIth fast chips and fast DDR4 would really make Primo jet along.

[a1call]

Yes, the 16 core does not seem to exist:

https://en.wikipedia.org/wiki/List_o...icroprocessors

[paulunderwood]

Originally Posted by a1call

Yes, the 16 core does not seem to exist:

https://en.wikipedia.org/wiki/List_o...icroprocessors

Mine is the old Opteron type 6*** series. The new 7*** series are so much better.

https://en.wikipedia.org/wiki/Epyc#Epyc

[a1call]

They still seem to max out at 64 threads.
Probably not worth an upgrade to what you already have.

[paulunderwood]

Originally Posted by a1call

They still seem to max out at 64 threads.
Probably not worth an upgrade to what you already have.

My 6174 chip costs ~£10, whereas a top of the range EPYC is ~£2500 and you would want two of them. It would make a Primo proof less time-consuming.

[DukeBG]

In February I did a 5213-digit number on 4.3.0 in about 4.8h (17288s) on my 8c/8t i7-9700K.
Just now finished a 5331-digit number with 4.3.1 in about 3.8h (13746s). A 20% increase in time and a slightly larger number too. I might also have other stuff running simultaniously in february, so it's far from a clean test, but still.

P.S. also finished a 14k dd number on 4.3.1 that I started on 4.3.0, about 490h total wall-clock time and that felt good.

[rudy235]

Originally Posted by wpolly

I guess R49081 is always there, but I'm not sure if it's feasible now....

A pair of numbers that should be totally doable would be 10³⁵⁶³⁰+239239
or 10³⁵⁹²⁵-9

If one wishes for a more difficult task this pair could be it 10³⁹⁹²²-57 and 10⁴⁰⁰⁰⁰+14253

[vasyannyasha]

Originally Posted by rudy235

A pair of numbers that should be totally doable would be 10³⁵⁶³⁰+239239
or 10³⁵⁹²⁵-9

If one wishes for a more difficult task this pair could be it 10³⁹⁹²²-57 and 10⁴⁰⁰⁰⁰+14253

(2¹²⁷⁰³¹+1)/3 is just above 38000 digits.

[Chara34122]

Originally Posted by vasyannyasha

(2¹²⁷⁰³¹+1)/3 is just above 38000 digits.

For now even (2^95369+1)/3 is not tested allthough it's much smaller than many other ECPP-tested numbers.