Hey master81,

I saw your message and will reply to it when I get a second.

-z

Great...but xorg self populates for all the distros. That's the problem with >=16 GPUs. If you let the screen section populate with 19, X will not start and no OC will happen. This is why I had to make it so it doesn't auto-populate, resulting in 16 OC and 3 SC. So have you tested on >=19?

Hey poisonxa,

Thanks for your fork. I'm looking forward to trying it.

Currently on nvOC v0019-v1.3 mining ETH only with Claymore 10.0
I did a lot of work getting nvOC to run 19 OC'd P106-100s headless on a ME mobo but only managed to get 16 OC stable at 18.1 MH/s and 3 stock clocks at 17.5 MH/s.
Yes I know for P106 that's slow but they are Gigabyte with Hynix. ;-(

My apologies if this has been discussed already, but how does your fork handle >16 cards OC? And if that works, I'd love to hear how you did it as I ran into limits with X and thus had to customize xorg.conf device and screen sections to list only 16 screens but have 19 devices.

Thanks and hope to hear from you.
-z

It's REMOTE.

If there is one question that might help solve this that I can't seem to find a straight answer for is:

- Where is nvidia-settings called from which results in the creation of xorg.conf?

I'm back!

So after all that, I don't get miner, watchdog, temp screens. So I tried to kick start it with ./2unix and got this lovely response:


Any ideas?

Hey damNmad,

I am running a 19 GPU rig and can't rely on the auto-creation of xorg.conf because it will detect all 19 cards and will list all 19 in sections Server Layout, Screens, and Devices. If the Server Layout section has more than 16 screens listed in it, X will fail to start (see xorg.0.log) and without X starting, nvidia-settings cannot run and no overclocking will occur for any of the 19 cards.

The trick is to make it so that only 16 screens appear in Server Layout, but allow all 19 to be listed in both Screens and Devices. The result of this is that 16 GPUs will be OC'd and 3 won't. But that's better than none getting OC'd.

I *was* running 1.3 when it tanked. I worked on it until things just didn't seem right so I reflashed 1.3 back on my USB stick and I got it running again, but it was unstable and would either slow way down OR the miner process would totally quit. So I decided to move to the community version 1.4v2. This also has not solved my issue so far.

My only solution thusfar has been to force my version of xorg.conf on every boot, and for whatever reason, something changed and I'm getting some sort of default configuration written into xorg.conf that only has a few cards and they aren't detected as nvidia. So the miner never starts cuz there are no GPUs detected, I guess.

So where I am now is that I have to:

1. Edit /etc/init.d/gpu and commenting out the sections that mv xorg.conf to xorg.conf.backup and create a new one based on nvidia-smi queries.
2. Place my custom xorg.conf in /etc/X11
3. Take that same xorg.conf file, rename it to 52-nvidia.conf, and put it in /usr/share/X11/xorg.conf.d/

Any thoughts?

-z

Hey guys

It's been a while since I've posted cuz my rig has been humming along pretty well but yesterday the flash drive on it died and I had customized nvOC a bit and am trying to get it working again at OC speeds. I need just a little pointer from you:

- What script kicks off nvidia-settings? I have a custom xorg.conf file that need to not be overwritten. On SMOS, it's /root/utils/rclocal.sh, and with that, I can edit the script to NOT make a new xorg.conf file, but instead copy my working xorg.conf file from ~ to /etc/X11/. I can't figure out where I did this on nvOC, or how to prevent xorg.conf from being overwritten.

Can somebody remind me?

Thanks,
z

You can do it all with one PSU if it’s strong enough, but you have to power the risers and GPUs directly from the PSU. The three PSU connectors are only to sync on/off the 2nd and 3rd ones, and the three molex mobo attachments are only if you are attempting to power the risers via the mobo.

1.3

Hey Everyone,

If you have more than 16 GPUs, the key is to have no more than 16 cards listed in the ServerLayout part of your xorg.conf. Those are the ones that will OC. You will have all 19 listed in the Device and Screen sections and the three that remain will run but won't OC. If you add a 17th under ServerLayout, the miner will not be able to OC any because X will crash when it starts when it hits the 17th card.

If you look at nvidia-smi output, you *should* see, at the bottom an instance of X AND a miner process for each GPU. The ones that only have a miner process attached cannot be OC'd. For me this is GPUs 17, 18, 19.

I've had to edit my rclocal.sh to copy a custom xorg.conf from ~ into /etc/X11 to keep it correct. The problem (for me) is that the script that auto creates the xorg.conf won't stop at 16 in the ServerLayout section and thus makes it so that X won't start and thus none of them OC.

Also, for my Gigabyte P106 cards (I'm running 19) for those first 16 (which will OC), I've not been able to get beyond 100 core and 250 mem, producing 18-18.1 MH/s for ETH only.

Let me know if this helps!
-z

```
# xorg.conf for 19 P106 on Asus ME that gets me 16 OC and 3 non-OC

Section "ServerLayout"
    Identifier "layout"
    Screen 0 "nvidiascreen-00"
    Screen 1 "nvidiascreen-01"
    Screen 2 "nvidiascreen-02"
    Screen 3 "nvidiascreen-03"
    Screen 4 "nvidiascreen-04"
    Screen 5 "nvidiascreen-05"
    Screen 6 "nvidiascreen-06"
    Screen 7 "nvidiascreen-07"
    Screen 8 "nvidiascreen-08"
    Screen 9 "nvidiascreen-09"
    Screen 10 "nvidiascreen-10"
    Screen 11 "nvidiascreen-11"
    Screen 12 "nvidiascreen-12"
    Screen 13 "nvidiascreen-13"
    Screen 14 "nvidiascreen-14"
    Screen 15 "nvidiascreen-15"
    Inactive "intel"
EndSection

Section "Device"
    Identifier "intel"
    Driver "modesetting"
    BusID "PCI:0@0:2:0"
    Option "AccelMethod" "None"
EndSection

Section "Screen"
    Identifier "intel"
    Device "intel"
EndSection

Section "Device"
    Identifier "nvidia-00"
    Driver "nvidia"
    BusID "PCI:1@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-00"
    Device "nvidia-00"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-01"
    Driver "nvidia"
    BusID "PCI:2@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-01"
    Device "nvidia-01"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-02"
    Driver "nvidia"
    BusID "PCI:3@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-02"
    Device "nvidia-02"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-03"
    Driver "nvidia"
    BusID "PCI:6@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-03"
    Device "nvidia-03"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-04"
    Driver "nvidia"
    BusID "PCI:7@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-04"
    Device "nvidia-04"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-05"
    Driver "nvidia"
    BusID "PCI:8@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-05"
    Device "nvidia-05"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-06"
    Driver "nvidia"
    BusID "PCI:9@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-06"
    Device "nvidia-06"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-07"
    Driver "nvidia"
    BusID "PCI:10@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-07"
    Device "nvidia-07"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-08"
    Driver "nvidia"
    BusID "PCI:11@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-08"
    Device "nvidia-08"
    Option "Coolbits" "28"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-09"
    Driver "nvidia"
    BusID "PCI:12@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-09"
    Device "nvidia-09"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-10"
    Driver "nvidia"
    BusID "PCI:13@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-10"
    Device "nvidia-10"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-11"
    Driver "nvidia"
    BusID "PCI:14@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-11"
    Device "nvidia-11"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-12"
    Driver "nvidia"
    BusID "PCI:15@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-12"
    Device "nvidia-12"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-13"
    Driver "nvidia"
    BusID "PCI:16@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-13"
    Device "nvidia-13"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-14"
    Driver "nvidia"
    BusID "PCI:17@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-14"
    Device "nvidia-14"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-15"
    Driver "nvidia"
    BusID "PCI:18@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-15"
    Device "nvidia-15"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-16"
    Driver "nvidia"
    BusID "PCI:19@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-16"
    Device "nvidia-16"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-17"
    Driver "nvidia"
    BusID "PCI:20@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-17"
    Device "nvidia-17"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

Section "Device"
    Identifier "nvidia-18"
    Driver "nvidia"
    BusID "PCI:21@0:0:0"
    Option "ConstrainCursor" "off"
EndSection

Section "Screen"
    Identifier "nvidiascreen-18"
    Device "nvidia-18"
    Option "AllowEmptyInitialConfiguration" "on"
    Option "Coolbits" "28"
    Option "IgnoreDisplayDevices" "CRT"
EndSection

```

Thank you very much for your help!
My rig is now running with all those settings: New USB, Headless, no monitor attached, know IP, have SSH and mining. Here's the weird thing: See below for my current xorg.conf that is running and I'm getting all 19 GPUs mining but only the 1st one is OC'd. What's weird about it is that:

When I run I get a new xorg.conf that does two things:
Bus IDs lose the "@" symbol, and mining does not work.

Any ideas why some BusIDs use the @ while most I've seen don't, and why this seems to be allowing mining to run?
Also weird is that if you notice, there are only BusIDs for 0-3 (4 and 5 are missing), then it continues from 6-11 and that's it.

What are your thoughts on this?

Current xorg.conf:

current lspci:

Thanks, trying now. Did you have more than 16 GPUs?
(edit)I tried this and mining would not start. Had to do
to return it to being able to mine at all.
I've read that you can add the coolbits part to either screen or device; the code you gave me I think put it in whichever was the wrong one.
Any other ideas?

Hey Linux Gurus:

Any idea how to allow a 19x P106-100 rig to OC in nvOC? I can't get my OC settings to apply though mining does work for all 19.

Suggestions appreciated!
PM me if you'd prefer.

-z

thanks papampi, good to know. Do you have any idea about my question about where the nvOC OC scripting is?

Hey all, glad to hear you got your rigs back up. Mine never stopped but that might be that I had
AUTO_UPDATE_CHECKS="NO"

On another note, would somebody point me to where/how the OC script is run? I'm trying to get my 19x P106 headless rig beyond stock clocks (OC settings aren't applying).

Thanks in advance.
-z

Please excuse my n00b questions:

- I'd like to read the OC script to understand it; can somebody give me the path to it?
- Is there a log of the OC process so I can see if they get applied (instead of just looking at nvidia-smi)?

Thanks!