Prime95 30.8 (big P-1 changes, see post #551)

[James Heinrich]

Originally Posted by Prime95

I'm not familiar with "Fishhawk Fall". If this is a not-yet-released product...

Fishlake Falls is the platform name for Sapphire Rapids-X Xeon CPUs
A decent overview: https://wccftech.com/intel-fishhawk-...-workstations/

[JuanTutors]

Is there a table or formula for appropriate B2/B1 values as a function of ram usage for P95v30.8b15? Trying to figure out if B2/B1=73 for 27 GB ram is appropriate.

[James Heinrich]

Originally Posted by JuanTutors

Is there a table or formula for appropriate B2/B1 values as a function of ram usage for P95v30.8b15?

It's a complex interplay between RAM, exponent, chance of factor (previous TF, new B1/B2). It's probably safe to say that if Prime95 selected those bounds for that assignment, then it's the appropriate bounds.

[JuanTutors]

Originally Posted by James Heinrich

It's a complex interplay between RAM, exponent, chance of factor (previous TF, new B1/B2). It's probably safe to say that if Prime95 selected those bounds for that assignment, then it's the appropriate bounds.

Got it. 100M digit Mersenne factored to 2^80, B1=1369000 and B2=96916000. Given how fast this is compared to a PRP test, if I happen to be on the same version of P95, I might increase those values manually next time just because it's so fast!

[James Heinrich]

Originally Posted by JuanTutors

I might increase those values manually next time just because it's so fast!

You can, of course, but the auto-selected values are carefully balanced to give an optimal balance of time spent on P-1 vs chance of finding a factor. You can increase the bounds and have a higher chance of finding a factor on this exponent, but your overall rate of factors found per time will decrease (the amount of time spent goes up faster than the chance of finding a factor).

[axn]

Originally Posted by JuanTutors

Got it. 100M digit Mersenne factored to 2^80, B1=1369000 and B2=96916000. Given how fast this is compared to a PRP test, if I happen to be on the same version of P95, I might increase those values manually next time just because it's so fast!

If it is possible to upgrade the RAM in your computer, upgrade it to the maximum capacity and allocate the maximum RAM you can to the program. It will automatically choose _much_ higher bounds.

[JuanTutors]

Originally Posted by axn

If it is possible to upgrade the RAM in your computer, upgrade it to the maximum capacity and allocate the maximum RAM you can to the program. It will automatically choose _much_ higher bounds.

Unfortunately I only have 32GB ram and I allocated 27GB. Maybe I'll get more ram on my next laptop.

[Prime95]

30.8 build 16 released for Windows and Linux 64-bit. A bug fix for Ryzen 7000.

[James Heinrich]

Originally Posted by Prime95

30.8 build 16 released for Windows and Linux 64-bit

Download links for those who don't want to go searching:
https://www.mersenne.org/ftp_root/gi...8b16.win64.zip
https://www.mersenne.org/ftp_root/gi...linux64.tar.gz

@George: you're still not comfortable putting v30.8 on the homepage (even though it's the default download on the download page)?

[Prime95]

Originally Posted by James Heinrich

@George: you're still not comfortable putting v30.8 on the homepage (even though it's the default download on the download page)?

Just being lazy.

[Jwb52z]

So, we don't need build 16 if we're not using the Ryzen 7000?