Installed the latest version of xmr-stak on my main pc, dual E5-2667 v3 ES and one Rx Vega 64 frontier edition gpu, and its a little slower than xmrig on the cpu's and about the same on the Rx Vega gpu compaired to xmrig-amd.
But if I configure ""low_power_mode" : true," in the cpu.txt config file it mines about 100 H/s less on the cpu's and runs about 10c-12c cooler and frees up 16 threads.
Next I compiled xmr-stak on one of my Open Compute servers and ran it, its about 5-10 H/s slower than xmrig 2.5.2 using 20 cores, ~850-855 h/s vs 857-860H/s, and thats down from ~876 H/s running the older version of xmrig.
I assume that the dev will continue to tweak it and it should hash higher.
I am just testing so I am not running two instances for peak performance because its easier to make changes just running one instance.
So next I configured xmr-stak on the oc server to ""low_power_mode" : true," and it runs ~812 H/s but it runs about 10-12c cooler on 10 threads vs ""low_power_mode" : false," with 20 threads.
I need to hook up my kilowatt meter to compaire the power usage but I think I will run the low power mode during the summer to save on the power usage with the a/c running less.
I thought about turning hyperthreading off or maybe turbo boost off in the bios or both off to see what the hash rate and power use would be and if it would be worthwhile.
Tomorrow I should have the power usage comparison and also I'll install xmr-stak on my E5-2665 nodes to see how much cooler they will be.
But if I configure ""low_power_mode" : true," in the cpu.txt config file it mines about 100 H/s less on the cpu's and runs about 10c-12c cooler and frees up 16 threads.
Next I compiled xmr-stak on one of my Open Compute servers and ran it, its about 5-10 H/s slower than xmrig 2.5.2 using 20 cores, ~850-855 h/s vs 857-860H/s, and thats down from ~876 H/s running the older version of xmrig.
I assume that the dev will continue to tweak it and it should hash higher.
I am just testing so I am not running two instances for peak performance because its easier to make changes just running one instance.
So next I configured xmr-stak on the oc server to ""low_power_mode" : true," and it runs ~812 H/s but it runs about 10-12c cooler on 10 threads vs ""low_power_mode" : false," with 20 threads.
I need to hook up my kilowatt meter to compaire the power usage but I think I will run the low power mode during the summer to save on the power usage with the a/c running less.
I thought about turning hyperthreading off or maybe turbo boost off in the bios or both off to see what the hash rate and power use would be and if it would be worthwhile.
Tomorrow I should have the power usage comparison and also I'll install xmr-stak on my E5-2665 nodes to see how much cooler they will be.
Last edited:
