: CPUs (832)
This would correlate to the 6344's I tested. A pair does 800h/s. Xmr-stak default config is 9T per CPU with 3T at low power for double mem and consumption of all L2/L3. In dual configuration this is drawing 250w from the wall. I tried setting 24T and 1x on memory usage but that was 40H/s slower and same power draw.
Sundownz don't like me. I just wanna be fwiends and swap knowledge.
Upon further review it appears that Sundownz is on the right track with 1x threads on every core. These Opterons behave differently than the Xeons that I have optimized before.
The 6300 series Opterons have a CC (Core Complex) that consists of 2 integer cores stuck together.
When I used:
1x thread on one of the two cores in the CC results in 65 H/s.
2x thread on one of the two cores in the CC results in 75 H/s.
1x threads on both of the two cores in the CC results in 82 H/s.
Power does go up with more cores running but not as much as I expected.
For some reason the hash rate does not drop much when the thread count exceeds the L2/L3 cache. When running all 12 cores the hash rate goes to 39/40 from 41.
I am now running this configuration on my dual 6348 Opteron system:
"cpu_threads_conf" :
[
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 0 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 2 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 3 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 4 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 5 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 6 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 7 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 8 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 9 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 10 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 11 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 12 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 13 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 14 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 15 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 16 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 17 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 18 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 19 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 20 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 21 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 22 },
{ "low_power_mode" : 1, "no_prefetch" : true, "affine_to_cpu" : 23 },
],
The reason I am not using CPU #0 core 1 is because I have that core affined to the Nvidia GPU's.
My overall hash rate for the CPU's only has jumped from 827 to 947. Power only increased 32 watts so it is profitable to use those extra 120 H/s.
https://www.cryptocompare.com/mining/calculator/xmr?HashingPower=120&HashingUnit=H%2Fs&PowerConsumption=32&CostPerkWh=0.074&MiningPoolFee=1My system now produces 1640 H/s while using 486 watts.
Profit per year now has increased to $510.
https://www.cryptocompare.com/mining/calculator/xmr?HashingPower=1640&HashingUnit=H%2Fs&PowerConsumption=486&CostPerkWh=0.074&MiningPoolFee=1