Re: [OS] nvOC easy-to-use Linux Nvidia Mining v0019

Have you set cpu mining to off ?

yep, for sure

looks like 3main select only one GPU for a reason I don't know

this is due to a problem with the 3main implementation wi$em@n found; but I haven't fixed yet. You can do this manually for now by finding this area in 3main:

Code:

if [ $COIN == "XMR" ]
then
HCD='/home/m1/xmr/stakGPU/bin/xmr-stak-nvidia'
ADDR="$XMR_ADDRESS.$XMR_WORKER"

cat </home/m1/xmr/stakGPU/bin/config.txt

"gpu_threads_conf" : [
  { "index" : 0,
   "threads" : 32, "blocks" : 18,
   "bfactor" : 8, "bsleep" : 10,
   "affine_to_cpu" : false,
  },
],

"use_tls" : false,
"tls_secure_algo" : true,
"tls_fingerprint" : "",

"pool_address" : "$XMR_POOL",
"wallet_address" : "$ADDR",
"pool_password" : "x",

"call_timeout" : 10,
"retry_time" : 10,
"giveup_limit" : 0,

"verbose_level" : 4,

"h_print_time" : 60,

"output_file" : "",

"httpd_port" : 0,

"prefer_ipv4" : true
EOF

cd /home/m1/xmr/stakGPU/bin

screen -dmS miner $HCD

if [ $LOCALorREMOTE == "LOCAL" ]
then
screen -r miner
fi

and adding an additional:

gpu_threads_conf" : [
  { "index" : 0,
   "threads" : 32, "blocks" : 18,
   "bfactor" : 8, "bsleep" : 10,
   "affine_to_cpu" : false,
  },
  { "index" : 1,
   "threads" : 32, "blocks" : 18,
   "bfactor" : 8, "bsleep" : 10,
   "affine_to_cpu" : false,
  },
  { "index" : 2,
   "threads" : 32, "blocks" : 18,
   "bfactor" : 8, "bsleep" : 10,
   "affine_to_cpu" : false,
  },
  { "index" : 3,
   "threads" : 32, "blocks" : 18,
   "bfactor" : 8, "bsleep" : 10,
   "affine_to_cpu" : false,
  },
],

index block per GPU to the "gpu_threads_conf"

as the above would be for 4x gpus.

thanks a lot Fullzero

I'm now able to use my 7x GPU with your modification

unfortunately watchdog is not working anymore as it saying that GPU utilisation is to low

not a big deal

thanks

I will look into a watchdog conflict with multiple gpus while using xmr-stak.

I noticed that xmr-stack takes a while to load up all the gpu's with work - on 13 card systems this can take up to 3-4 minutes (it also depends on the thread/block count in xmr-stak config.txt - the settings from fullzero load under 2G of data, but you can tweak those numbers higher getting, resulting higher hash rates, but also increasing the initialization time).
Once stak loads all the cards, and assuming your OC settings are stable - it will behave itself for days. I meant to modify watchdog to recognize stak is running and increase the initialization time appropriately, but I still need to migrate to v0019. for now if you want to run watchdog, change the initial timeout to 360 seconds it should do the trick.
Once all the cards are loaded stak will spit out a bunch of submits to the pool in one shot.

I've seen issue with WATCHDOG while using keccak algo (for MAX COIN) on MPH; the GPU utilization for this coin is algo is always around 90 and WATCHDOG always restarts the mining eventually the whole RIG based on that usage!

I remember one of the user complained about keccak; asked about how to disable it (thinking its probably because of this utilization low issue).