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

First, thank you so much fullzero Grin

for this. I've rolled my own in the past, so this was a nice find and saved me a lot of trouble to do my own.

Have a couple of questions
1) I'm using 4 x PNY GTX 1060 3GB cards on a Gigabyte Z97 SOC Force motherboard. Under Win10 with power 70, cclock 150, and mclock 850 I consistently get ~23 Mh/s. With nvOC with the same settings I get 21.4 Mh/s. Increasing the mclock doesn't seem to increase the hashrate. cclock at 100 doesn't seem to make a difference either.

My oneBash looks like this:

Code:

COIN="ETH"

POWERLIMIT="70" # YES NO

INDIVIDUAL_POWERLIMIT="NO" # YES NO

__CORE_OVERCLOCK=150
MEMORY_OVERCLOCK=850

INDIVIDUAL_CLOCKS="NO" # YES NO

When the bash file starts it shows that the attributes are set, but it appears that the clock settings aren't taking effect. I don't see any errors thrown when Claymore starts up except that the cuda library doesn't have a version number.

2) I added a couple of parameters to the Eth command line, but it doesn't seem to pick them up. For instance, when try to interactively change the dcri it tells me that i need to set -asm 1, but that is already set in the bash file.

Code:

if [ $COIN == "ETH" ]
then
HCD='/home/m1/9.0/ethdcrminer64'
ETHADDR="$ETH_ADDRESS/$ETH_WORKER"

if [ $ETHERMINEdotORG == "YES" ]
then
ETHADDR="$ETH_ADDRESS.$ETH_WORKER"
fi

until $HCD -epool $ETH_POOL -ewal $ETHADDR -epsw x -mode 1 -esm 0 -estale 0 -asm 1 -dcri 12
do
echo "FAILURE; reinit in 5" >&2
sleep 5
done
fi

Any recommendations would be much appreciated. Windows 10 sucks for mining - so much babysitting, so hoping to get this tweaked and running well.

Powerlimits in windows are in percent TDP; in linux they are in watts. Thus a powerlimit of 70 in windows is .7 * 120 watts or whatever the TDP is (84 watts if the card TDP is 120); while 70 in linux is 70 watts. I would recommend comparing Windows and linux powerlimits with a kill-a-watt to get them exactly equal; but most likely a 3gb 1060 has a 120 or 125 watt TDP.

Also linux OC offsets are scaled differently than windows; you will need to use higher offsets to get the same results in linux.

In regards to Claymore ETH settings; the Claymore readme says -asm is for AMD only. I haven't tested this but -dcri 30 (or any valid number) is only supposed to work with mode 1 (ETH only) if -asm mode is used.

fullzero, thank you for the quick reply. This was hugely helpful.

So strange that Windows is percentage of TDP and linux is watts. The PNY 1060-3's are 120W TDP, so I've set it at 84 in linux. I'll check it out with my power meter.

I increased the mclock to 1500, which is getting me about 22.7 Mh/s now. I'll keep playing with it, but I see that the nvidia control panel has a max of 2000 for memory, whereas it's 1000 in afterburner on Windows. It seems like the offset might be 2:1 on linux, so my 850 would be 1700 on linux.

Finally, with the dcri, you're right. I have a mix of AMD and nvidia cards, so just assumed it worked on nvidia. Tried it on a Windows nvidia only rig, and it gives the same error, so pretty sure dcri tuning doesn't work with nvidia cards.

Thanks again, this rocks!

powerlimits for different cards and in general:

if using the same powerlimit for each card set:

Code:

POWERLIMIT="YES"

and in this block of code:

Code:

if [ $POWERLIMIT == "YES" ]
then
sleep 6
#change powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6
fi

find the line:

Code:

expect -c 'spawn sudo nvidia-smi -pl 125

and replace 125 with the desired powerlimit in watts

if using the individual powerlimits for each card set:

Code:

INDIVIDUAL_POWERLIMIT="YES"

and in this block of code:

Code:

if [ $INDIVIDUAL_POWERLIMIT == "YES" ]
then
sleep 6
#change GPU 0 powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -i 0 -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6

#change GPU 1 powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -i 1 -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6

#change GPU 2 powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -i 2 -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6

#change GPU 3 powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -i 3 -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6

#change GPU 4 powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -i 4 -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6

#change GPU 5 powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -i 5 -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6

#change GPU 6 powerlimit by changing the number after -pl to the desired wattage
expect -c 'spawn sudo nvidia-smi -i 6 -pl 125
expect "*password*:"
send "miner1\r" '
sleep 6

........

change each line:

expect -c 'spawn sudo nvidia-smi -i 0 -pl 125 ( this is for GPU 0 )

expect -c 'spawn sudo nvidia-smi -i 1 -pl 125 ( this is for GPU 1 )

expect -c 'spawn sudo nvidia-smi -i 2 -pl 125

expect -c 'spawn sudo nvidia-smi -i 3 -pl 125

expect -c 'spawn sudo nvidia-smi -i 4 -pl 125

expect -c 'spawn sudo nvidia-smi -i 5 -pl 125 ( this is for GPU 5 )

to the desired powerlimit.

Note: the indexing of the GPUs might be different than you expect.

If you open the guake terminal; press f12

and enter:

Code:

nvidia-smi

It should list your cards in the correct order.