what is the best config for xmr-stak to mine monero? - configuration

Question

My tech spec is dual intel gold 6140 (36 cores 2.3ghz) 96gb ram, 2 800gb ssd (raid) and 2 nvidia v100 32gb cards. (i don;t have access to bios to overclock)
I was wondering what would be the best config and setup to get the most hashes? currently I'm getting 3000-4000 h/s thats both gpu and cpu combined. I allowed the large pages and increase page size to 64gb (not sure if that was necessery or not) also I installed latest cuda.
this is my cpu config not sure if i'm getting the max amount of threads out of it, also i get some error that it can't go to 86 only 63, not sure what it means.
"cpu_threads_conf" :
[
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 0 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 2 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 4 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 6 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 8 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 10 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 12 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 14 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 16 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 18 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 20 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 22 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 24 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 26 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 28 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 30 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 64 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 66 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 68 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 70 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 72 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 74 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 76 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 78 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 80 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 82 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 84 },
{ "low_power_mode" : false, "no_prefetch" : true, "asm" : "auto", "affine_to_cpu" : 86 },
],
this is my gpu config (default)
"gpu_threads_conf" :
[
// gpu: Tesla V100-PCIE-32GB architecture: 70
// memory: 32127/32642 MiB
// smx: 80
{ "index" : 0,
"threads" : 4, "blocks" : 640,
"bfactor" : 6, "bsleep" : 25,
"affine_to_cpu" : true, "sync_mode" : 1,
"mem_mode" : 1,
},
// gpu: Tesla V100-PCIE-32GB architecture: 70
// memory: 32127/32642 MiB
// smx: 80
{ "index" : 1,
"threads" : 4, "blocks" : 640,
"bfactor" : 6, "bsleep" : 25,
"affine_to_cpu" : true, "sync_mode" : 1,
"mem_mode" : 1,
},
],

Answer 1

I believe the correct option would be to test out how much mining your rig gets done when slightly changing your specs then try to compare it to the energy cost/ damage your CPUs and GPUs are accumulating and decide what is the best settings based on that. I believe using even numbered values will most likely give best results and making sure each thread has at least enough cache memory for holding a hash seems intuitive. Also check to make sure your cards do not require a dummy video output or video output short in-order to run at max capacity. Some miners i've see need hacked video shorts for max performance.