Runs Prime95 on Raspberry Pi

[primeawesome]

Hello everyone,

Anyone tested prime95 on raspberry pi? I intend use my Raspberry Pi 3 Model B to run this program, but I don't know if this is possible. If can share review I will appreciate.

Thanks

[Nick]

Try this thread.

[VictordeHolland]

A doubleCheck takes 1.5-2 months on a quadcore A53 running 24/7, so know what you start. You could run ECM or P-1 tasks, they're much shorter and should only take a few days.

[Uncwilly]

Originally Posted by VictordeHolland

You could run ECM or P-1 tasks, they're much shorter and should only take a few days.

Do they come with enough memory to make P-1 worthwhile on it?

[VictordeHolland]

Originally Posted by Uncwilly

Do they come with enough memory to make P-1 worthwhile on it?

Good point, but I think it should still be possible.

Rasp Pi 3 B has 1 GB of DDR2, so it is indeed a bit tight.

If you limit the memory usage to 500MB for the high memory worker (and limit high memory workers to 1) you might still be able to do a P-1 in the current wavefront:

With Prime95 limiting the memory from 4500MB to 500MB just reduces the B1 and B2 bounds:

[Feb 13 16:38] Worker starting
[Feb 13 16:38] Setting affinity to run worker on any logical CPU.
[Feb 13 16:38] Optimal P-1 factoring of M86143697 using up to 4500MB of memory.
[Feb 13 16:38] Assuming no factors below 2^76 and 2 primality tests saved if a factor is found.
[Feb 13 16:38] Optimal bounds are B1=695000, B2=13205000
[Feb 13 16:38] Chance of finding a factor is an estimated 3.05%
[Feb 13 16:38] Using AVX FFT length 4608K, Pass1=512, Pass2=9K
[Feb 13 16:38] Worker stopped.
[Feb 13 16:39] Worker starting
[Feb 13 16:39] Setting affinity to run worker on any logical CPU.
[Feb 13 16:39] Optimal P-1 factoring of M86143697 using up to 500MB of memory.
[Feb 13 16:39] Assuming no factors below 2^76 and 2 primality tests saved if a factor is found.
[Feb 13 16:39] Optimal bounds are B1=650000, B2=8937500
[Feb 13 16:39] Chance of finding a factor is an estimated 2.8%
[Feb 13 16:39] Using AVX FFT length 4608K, Pass1=512, Pass2=9K
[Feb 13 16:39] M86143697 stage 1 is 0.00% complete.
[Feb 13 16:39] Worker stopped.

[ewmayer]

But which p-1 client? Mlucas doesn't (yet) support p-1, though it's been something on my to-do list for years and years it keeps getting pushed down by more important 'core mission' (i.e. LL)-related to-dos.

For smallish moduli, GMP-ecm/p-1 is of course fine, but your mention of 1GB implies GIMPS-wavefront-sized stuff.

[VictordeHolland]

Whoops