Topic: [OS] nvOC easy-to-use Linux Nvidia Mining - page 405. (Read 418251 times)

This xorg.conf is missing large amounts of needed settings.  I think I know what the problem is / happened.  The image I made needs to NOT be using integrated graphics.

You have integrated graphics enabled in the bios or have attached a monitor to the video output on the mobo and forced an auto configuration on the mobo to start the integrated graphics.  This caused the system to create a new xorg.conf file which lacks almost all the needed changes.  Technically the one needed should be stored as the backup, but fixing this image will probably take longer than the following:

How is your monitor connected to your rig?  

If it is not attached to the second GPU please attach it there.

Then shutdown your rig.  

Unplug the usb key.  

Power on the rig, bios should post.  

Restore defaults to the bios, it should need to restart to do this.  

When it has rebooted with default configurations, make these three bios changes here and no others:

ensure 'Above 4G Decoding' is enabled in the bios.

ensure PTP aware OS: is set to 'Not PTP Aware' in the bios.

ensure you 'Clear Secure Boot Keys' in the bios

save and reboot

Now reimage your USB key with the ASUS PRIME Z270-A image.

when it is done, attach to rig and press ctrl + alt + delete to reboot

When ubuntu starts up watch the gnome terminal and see if OC occurs correctly.  I think it will if you follow the previous steps.

If it doesn't: open the guake terminal and enter:

lspci | grep VGA

and post what it says.

If it does work also let me know.  


Edit:  note when 7 cards are connected the monitor will display all black while booting; but will display normally once ubuntu starts.

It seems like there is a powerlimit of 125 with those results.  377 Sol/s is way too low for stock power, even without OC.

Note: once you have used a powerlimit at all since boot it will not turn off, until you reboot or logout: it can only be modified. Thus to return to stock TDP you would use a powerlimit of 150, or reboot/logout/in.

In EWBF the pec output will accurately tell you how much power the cards are currently using, look that and see if they likely are using a powerlimit.

Currently with the good ZEC returns; I am using cc+200 mc+900 powerlimit 135 with my zotac 1070 mini rigs.

With my 1080tis I am currently using cc+100 mc+900 powerlimit 225.

Edit: also using manual fans with 75% speed.

anyone has good settings for core/mem in onebash for 1070 and 1080ti

I have tried many settings with various results but tdp seems to be inconsistent.

3 x Zotac 1070 Mini around 439w from the wall, ZEC 1132sols at stock settings, no changes in onebash settings

is this expected?

It's around 1.5Mhs lower in linux per card (22.5Mhs in windows vs 21Mhs in linux)

If you set the linux power limit so that the actual power-at-the-wall matches what it is with windows; how do the hashrates compare?

58da134925d64f01ecd9a9cbbd803936a01361a38ee23770542215815eaea690

I have the rig connected to smart plug which shows power usage and all my cards have 120W TDP. In AB I set all cards to 70% which gives me 84W per card + efficiency of PSU and setting 84W in linux produces ~10W more for entire rig power consumption.

When comparing power consumption; even within the same OS the software is not accurate.  With some brands the information from Nvidia-smi is almost correct, with others it is very off. 

When comparing Afterburner settings from windows with linux powerlimits; you need to know the TDP of the card you are using.  This varies by manufacturer and the specific card model.  Afterburner uses percent TDP while linux powerlimits are implemented in watts.  The 1060 3gb I have have a TDP of 120 watts.  Thus 84 % of 120 is 100.8 watts: this would be the linux powerlimit equivalent of a windows afterburner powerlimit setting of 84.

To have accurate comparisons you really need to have a device such as a kill-a-watt, that will measure the power-at-the-wall.  Tell me if a 101 watt powerlimit in linux produces similar hashrates.

I plan on making a lan management monitor / push / update app.  It is second on my member requested list after improving amdOC beta.

Currently stand alone; but is 16.04 based so you can setup remote access using software or application layer protocol of your choice.

Am thinking of switching a couple rigs to nvoc. Is there a web based remote management set up yet? Or is it still set up so you have to be there to make any changes or reset? Thanks

Reflash did the trick

One last thing

When setting the fan it seems to only affect gpu0? The other gpu's dont spin as fast

press f12 to open the guake terminal

If cpuminer-opt is running rightclick over the guake terminal and select new tab or press ctrl + c to stop it.

then enter:

lspci | grep VGA

this will list your gpus and their pcie addresses

then enter:

gksu gedit '/etc/X11/xorg.conf'

and enter the password:

miner1

when prompted

gedit should now be open with root access to the xorg.conf file

Find the Device section and alter the

BusID          "PCI:1:0:0"

on each Device to match the addressing from lspci (only change the numbers)

save

close all mining processes if open

logout

login, and all cards (up to 6x) should now OC if you did this correctly.

I have 378 driver already installed and the command provided returns 60W for minimum power limit and 140W for max
here is the screen from nvidia-smi command showing all gpus with low power limit set:
https://ibb.co/hUHMLv

nvidia-smi -i 0 --query-gpu=power.min_limit,power.max_limit --format=csv
where -i "number" is the gpu you want to query
regarding second question, I am also seeing discrepancies, try higher overclock or higher PL
also, go for 378 driver!

I've switched from win10 to nvOC as I was heaving issues with windows hanging from time to time and I have couple questions:

1. I have 3 1060 from different manufacturers and when I set power limit to anything from 70W to 84W it works on 2 of them but the third one, which is palit, is running always at 94W and it won't get lower, is this something limited by cards bios?
2. In windows with latest drivers I'm getting 22.5-22.8 Mhs in eth per gtx1060 using Claymores miner but in nvOS I'm getting 21Mhs with exactly same settings (core +120, memory +700, PL 84W) Is this typical for eth because of how Claymore wrote his miner for linux or simply I should look for different settings because of OS/driver differences?

Greetings fullzero and Linux aficionados,
I have been in love-hate relationship with Linux for two years, he is not perfect and I am demanding but also inexperienced  lover, therefore the hate part.
In the long run there may even be a wedding
So, to all of you having problems and thinking of giving up: persevere, be tolerant and learn! We need to emancipate ourselves from Microsoft.

Anyway, back to topic.
What is the correct procedure for adding gpus? Not on nvOC, but generally on Ubuntu?
Should I connect all of them, or one-by-one?
Because I have a problem with nvidia-settings seeing 3 cards and lspci showing all four.
Thanks.

press f12 to open the guake terminal

If cpuminer-opt is running rightclick over the guake terminal and select new tab or press ctrl + c to stop it.

then enter:

lspci | grep VGA

this will list your gpus and their pcie addresses

then enter:

gksu gedit '/etc/X11/xorg.conf'

and enter the password:

miner1

when prompted

gedit should now be open with root access to the xorg.conf file

Find the Device section and alter the

BusID          "PCI:1:0:0"

on each Device to match the addressing from lspci (only change the numbers)

save

close all mining processes if open

logout

login, and all cards (up to 6x) should now OC if you did this correctly.