Primes/s is a completely useless metric...
From your screenshots:
5-chains seems to be less than what an I7 on HP-builds gives
9-chains not displayed in your screenshot, but this is what chains/day tries to estimate in the CPU-miner. You can get >=1.0 with an I7, which means an estimated 1 block per week at diff >= 9.8. Based on the 8-chains performance, this doesn't look too good.
Reaper updates every second and every second there be a couple of 5-6-7-8 chains!
jhPrimeminer is much slower.
My i5 gives "primespersec" : 2695 in HP10
what you mean "5-chains seems to be less than what an I7 on HP-builds gives"?
Do you mean in HP-builds displaing 5-chains in primespersec command?
Na, the numbers for the chains on the screenshots are per HOUR and not per second. You certainly don't have 8-chains every second, otherwise you would find a block every 5 minutes at least.
Since the reaper sourcecode leaked, it's easy to proof:
in App.cpp:
for(uint i=2; i<15; ++i)
{
if (chainspersec > 0)
{
double num = 3600.0*chainspersec/((ticker()-starttime)*0.001);
string str = SIize(num);
cout << str << " " << i << "-chains ";
}
}
If you run the hp builds with "-printmining -printtoconsole" you'll see primemeter every now and then, which also gives 5-chains per hour estimates. I think I get around 600 5-chains / h on an older Intel I7.
Also:
My i5 gives "primespersec" : 2695 in HP10
Means 2695*3600 = 9702000 primes per hour. So the 15 million primes per hour (and not second!) is actually not even 2x faster than your i5. Primes per hour is still a pretty useless metric to begin with, since you can tweak the settings so that you get more primes but less chains. Estimated blocks per day/week/month or whatever is the only thing that counts.