Thank you for answer it was very useful. but those statements was for Primegrid. I use PPS Cuda sieve only for specific K-s ( there is 72 K ) and for specific N range. Ok, range is not problem, but specific K is tricky part. When I run in datles sieving mode ( version is 0.2.3b) then I got 5% slower speed and all factors of all K in range. And that is huge amount of useless data for me. But, as you say in that case PPSCuda sieve uses only 2% of CPU core, and 12 KB of memory, and that is for today nothing.
But, when I run it on my range ( from abcd file) I got 5% faster, and got output for only K that are in sieve, however memory usage is around 440 MB and CPU usage is 5%.
So , this was reason why I ask will I gain in speed if I stop a sieving process once a day, remove found factors, and start again or just start from begging to end .
92*10^1439761-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
314187728^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie!