Monero Mining Performance
- Thread starter Patrick
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Yea the issue is the config file. Needs to use environment variables to make silly easy to swarm.
For even systems as small/old as 12 threads, I have swapped to hugepages=number of threads, and have picked up ~10% in performance (with Wolf's Miner) - give it a shot when you've got a chance.
ok, docker with Stak works great, +9% roughly.
Hugepages seems to do nothing at all for wolf, but I dont know if I need to change my command:
nice -19 ./minerd -a cryptonight -o stratum+tcp://xmr.pool.minergate.com:45560 -u email@address.com -p x -t 8
I tried as root too, no change.
As for Stak Docker, the config doesnt bother me at all, edited once and I'm good, but I don't know how to "nice" a docker container, no how to change the container to nice the miner. I would have to make some sort of re-nice sleeper script.
Hugepages seems to do nothing at all for wolf, but I dont know if I need to change my command:
nice -19 ./minerd -a cryptonight -o stratum+tcp://xmr.pool.minergate.com:45560 -u email@address.com -p x -t 8
I tried as root too, no change.
As for Stak Docker, the config doesnt bother me at all, edited once and I'm good, but I don't know how to "nice" a docker container, no how to change the container to nice the miner. I would have to make some sort of re-nice sleeper script.
Well, don't mind the instant reply... I spent an hour trying to fix it and figured it out, then realized I never posted my problem
so i use -v to substitute local fs files overtop of docker container ones.
I replaced the miner with a script that called the miner under a new name, and added a nice command in front. I also had to add back in the miner binary under a new name.
So basically im running a docker container to mine, and I substituted out the miner, the execution script, and the config file... all I have left is the environment and libraries.
Its retarded, but i'll use this until I learn how to do it better.
so i use -v to substitute local fs files overtop of docker container ones.
I replaced the miner with a script that called the miner under a new name, and added a nice command in front. I also had to add back in the miner binary under a new name.
So basically im running a docker container to mine, and I substituted out the miner, the execution script, and the config file... all I have left is the environment and libraries.
Its retarded, but i'll use this until I learn how to do it better.
@Spartus how are you doing threads? Are you hand editing the config?
Need to get:
To be the right number of threads and then:
Matching number of lines one per CPU.
For one machine it is easy to edit by hand. Need it to be automated to be useful for docker. Yet another case where my bash text manipulation skills are not up to snuff.
Need to get:
Code:
"cpu_thread_num" : 2,
Code:
"cpu_threads_conf" : [
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 0 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 1 },
],For one machine it is easy to edit by hand. Need it to be automated to be useful for docker. Yet another case where my bash text manipulation skills are not up to snuff.
My plan was to just leave every machines appropriate config on it locally and then just reference it, but yeah if you need to dynamically edit the config based on system resource commands I can help with that.
I don't have a dual socket linux box, can you give me sample output from lscpu and nproc from a couple example systems. I am assuming that is all I need to figure out the logic.
I could really use help with cgroups or nice or something as I am still struggling to have docker use 100% when the cpu is idle and <1% when the cpu is loaded. best I get is 50/50 even with nice level set (and autogroups are off).
I don't have a dual socket linux box, can you give me sample output from lscpu and nproc from a couple example systems. I am assuming that is all I need to figure out the logic.
I could really use help with cgroups or nice or something as I am still struggling to have docker use 100% when the cpu is idle and <1% when the cpu is loaded. best I get is 50/50 even with nice level set (and autogroups are off).
yeah I was going to help do the config file editing. I assume you want to use the default config as a starting input and modify it in place right?
Yes. The other thing I would want to do is have the username get pushed into the script from an environment variable.
Here is what the launch script is for the wolf's miner +
Code:
./minerd -a cryptonight -o stratum+tcp://xmr.pool.minergate.com:45560 -u $username -p x -t $numthreads
50K in 8 days? That doesn't seem right. There's only 15k miners on minergate.
Hey has anyone gotten AMD RX470 or RX480 mining on Ubuntu working? I've cycled through Trusty and Xenial and I can't get it to work. I've got a RX470 4GB.
Hey has anyone gotten AMD RX470 or RX480 mining on Ubuntu working? I've cycled through Trusty and Xenial and I can't get it to work. I've got a RX470 4GB.
I am using this thread to document my setup
After seeing @cafcwest comment, I re-run Wolf minerd ( had bee using non-docker version for previous few days ).
Enabled hugepages for both wolf and xmr-stak-cpu
sudo sysctl -w vm.nr_hugepages=128
Single E5-2650 v3 QS QFSB (10 core , 25MB cache ) CPU running minerd
-t 10 272 H/s
-t11 402H/s
-t12 421 H/s
-t13 400 H/s
-t14 370 H/s
In my setup, xmr-stak-cpu (465 H/s ) still faster than wolf (421 H/s).
After seeing @cafcwest comment, I re-run Wolf minerd ( had bee using non-docker version for previous few days ).
Enabled hugepages for both wolf and xmr-stak-cpu
sudo sysctl -w vm.nr_hugepages=128
Single E5-2650 v3 QS QFSB (10 core , 25MB cache ) CPU running minerd
-t 10 272 H/s
-t11 402H/s
-t12 421 H/s
-t13 400 H/s
-t14 370 H/s
In my setup, xmr-stak-cpu (465 H/s ) still faster than wolf (421 H/s).
@Patrick
You mentioned your E5-2650L v3 has the same / better performance.
I just checked Intel Ark , they are two different in terms of Core and Cache , not TDP rating
E5-2650L v3 = 12 cores , 30MB cache, base 1.8 ghz , turbo 2.5 ghz
E5-2650 v3 = 10 cores , 25 MB cache , base 2.3 ghz , turbo 3.0 ghz
I have a E5-2650L v3 as well, will try to compare the performance , slow vs fast clockrate vs more cores vs cache size.
You mentioned your E5-2650L v3 has the same / better performance.
I just checked Intel Ark , they are two different in terms of Core and Cache , not TDP rating
E5-2650L v3 = 12 cores , 30MB cache, base 1.8 ghz , turbo 2.5 ghz
E5-2650 v3 = 10 cores , 25 MB cache , base 2.3 ghz , turbo 3.0 ghz
I have a E5-2650L v3 as well, will try to compare the performance , slow vs fast clockrate vs more cores vs cache size.
I tried both , not making difference.
sudo sysctl -w vm.nr_hugepages=128 ( instruction from xmr-stak-cpu readme.txt )
and
sudo sysctl -w vm.nr_hugepages=30 ( instruction from minerd readme.txt 3 x 10 cores )
then I monitored with
grep Huge /proc/meminfo
sudo sysctl -w vm.nr_hugepages=128 ( instruction from xmr-stak-cpu readme.txt )
and
sudo sysctl -w vm.nr_hugepages=30 ( instruction from minerd readme.txt 3 x 10 cores )
then I monitored with
grep Huge /proc/meminfo
running R9 290 crossfire on Arch, It works, but its worse than windows purely because I can overclock in windows and don't want to reflash my GPU bios clocks (yet...)
performance is basically identical before overclock.
performance is basically identical before overclock.