View Single Post
Old 2021-05-30, 14:22   #4
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

22·3·13·37 Posts
Default

Quote:
Originally Posted by mersenneNoob View Post
Ok, I am thinking of starting an operation trillion digits. I am finding factors form range 3321928094941 to 3321928095989 currently.
Please consider helping with 100M-1G factoring, or up to 10G, instead. There's too little factoring being done to finish those in our lifetimes. I see no point in committing resources to TF that won't be needed done for over 50 years, probably over 150 years, unless quantum computing delivers bigly and soon.

There is to my knowledge no server or database to cover PRP or LL above 1G, or TF or P-1 above 10G. http://www.mersenneforum.org/showpos...11&postcount=9

A single PRP test at OBD takes too long for current hardware, short of some serious supercomputer time. There's no P-1 factoring software suitable for OBD yet with completion of one factoring attempt to suitable bounds in a year. There's only one OBD candidate with TF done to adequate depth. A trillion digit Mersenne is ~2,000,000 times slower to primality test or P-1 test than an OBD, and so since it currently would take much longer than the usual lifetime of human civilizations, there is no software to attempt it and no point at this time to create software for a futile attempt.
https://www.mersenneforum.org/showpo...6&postcount=14
If Moore's law persisted at 2 year doubling, it would take 42 years for a trillion digit Mersenne to become "only" as much a challenge as OBD are today, requiring years on the fastest software and consumer-market GPU for one PRP test. That 42 years is most of or beyond the remaining life expectancy of most GIMPS members.

There comes a point in the number line where it's not even a matter of waiting for Moore's law to make it more tractable, since feature shrink will stall out before reaching atomic size limit and already stopped providing clock rate increase ~15 years ago, & where there's not enough available mass on which to store the interim residues for P-1 factoring or PRP even at 10 bits per particle. (Attempting LL without some yet to be demonstrated bullet-, cannonball-, and nuke-proof error detection and correction would be insane. It's also pretty sketchy at 100Mdigit, and almost certain of error at 1G.)

Last fiddled with by kriesel on 2021-05-30 at 14:47
kriesel is offline   Reply With Quote