When the program is running (4x2080Ti), the log says

libEGL warning: DRI2: failed to authenticate
libEGL warning: DRI2: failed to create any config
libEGL warning: DRI2: failed to create any config

here and there. I suppose they exist simply because the fake display does not fully support EGL and thus these warnings can be ignored, right?

First off, thanks for the hack, it works fine on a headless GTX 1070.

The only problem I ran into is that when running the script as-is, when it executes sed, it outputs the following:

sed: couldn't open [xorg.conf full path]: No such file or directory

I "fixed" it by putting the command in a single line, I don't know if there is some shell incompatibility when handling spaces in the sed command or something like that.

It might be something related to my environment, and it's just a minor point, I'm just notifying it if anyone else run into it.

Hi I'm using both Alex and your version. I have GTX1070, the fan is increase to what I want. But during calculation the power (Watt) cannot go beyond 25W, does it happen in your server ?

I'm not super familiar with tcsh, and I translated the tcsh code to bash for my purposes, but I noticed that the stop option may not work properly (it did not in my bash version) because the setting of target to -1 at this line is overwritten by the else statement in the next check, and target gets set to the string "stop".
The fix is to reverse the order of the two if statements:

if ("x$1" == "x" || "x$1" == "xstart" || "x$1" == "x-display") then
    # this means "use heuristic"
    @ target = -3
else
    @ target = $1
endif

if ("x$1" == "xstop") then
    @ target = -1
endif

[1] 20166
Persistence mode is already Enabled for GPU 0000:02:00.0.
Persistence mode is already Enabled for GPU 0000:81:00.0.
All done.

X.Org X Server 1.18.4
Release Date: 2016-07-19
X Protocol Version 11, Revision 0
Build Operating System: Linux 4.4.0-83-generic x86_64 Ubuntu
Current Operating System: Linux WAIR-P8000 4.4.0-83-generic #106-Ubuntu SMP Mon Jun 26 17:54:43 UTC 2017 x86_64
Kernel command line: BOOT_IMAGE=/boot/vmlinuz-4.4.0-83-generic root=UUID=512494ce-8ef7-4b2f-8996-d81639985410 ro text
Build Date: 17 July 2017 05:05:12PM
xorg-server 2:1.18.4-0ubuntu0.3 (For technical support please see http://www.ubuntu.com/support)
Current version of pixman: 0.33.6
Before reporting problems, check http://wiki.x.org
to make sure that you have the latest version.
Markers: (--) probed, (**) from config file, (==) default setting,
(++) from command line, (!!) notice, (II) informational,
(WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(==) Log file: "/var/log/Xorg.0.log", Time: Wed Jul 26 17:16:09 2017
(++) Using config file: "/tmp/xorg-aryyQphn.conf"
(==) Using system config directory "/usr/share/X11/xorg.conf.d"
/opt/set-gpu-fans/nvscmd: 84: /opt/set-gpu-fans/nvscmd: Syntax error: "(" unexpected (expecting "fi")
xinit: connection to X server lost

waiting for X server to shut down (II) Server terminated successfully (0). Closing log file.

I'm wondering if there is a fix for this error?

We are testing the script on a headless system.

NVRM version: NVIDIA UNIX x86_64 Kernel Module 381.22 Thu May 4 00:55:03 PDT 2017

the test setup is a non Titan GTX and the nvidia-smi line used internally by the script
/usr/bin/nvidia-smi dmon -s p -c 1
results in

Not supported on the device(s)
Failed to process command line

The solution would be to simplify the script, eliminate nvidia-smi call and use it to run nvidia-settings without GPU monitoring

/usr/bin/nvidia-settings -a '[gpu:0]/GPUFanControlState=1' -a '[fan:0]/GPUTargetFanSpeed='100'

Steps we followed:

• download into ${HOME}/set-gpu-fans
• create symlink ln -sf ${HOME}/set-gpu-fans /opt/set-gpu-fans
• no X running
• set persistent mode sudo nvidia-smi -pm 1
• execute

cd /opt/set-gpu-fans
sudo tcsh
./cool_gpu >& controller.log &
tail -f controller.log

We get the following output:

        (++) from command line, (!!) notice, (II) informational,
        (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(==) Log file: "/var/log/Xorg.0.log", Time: Sun May 21 20:43:25 2017
(++) Using config file: "/tmp/xorg-HLKicaMP.conf"
(==) Using system config directory "/usr/share/X11/xorg.conf.d"
@: Expression Syntax.
xinit: connection to X server lost

waiting for X server to shut down (II) Server terminated successfully (0). Closing log file.

I have to say this is a wonderful work. I can run it on my system. However, I found new issues. On my computer, I have two cards, gpu0: TITIAN RTX (two fans) and gpu1: RTX 2080ti (one fan). After I run this script to control gpu fans, it seems that only one fan's speed are adjusted for TITAN RTX. leading to high temperature. Besides, two gpus' fans speed are controlled in the same level. Actually, the second gpu's temperature is 15 degrees lower the first gpu. I want to control them seperately. How should I modify this script to realize functions I want.
Thanks.

I replace 'run_forever' with 'stop' in cool_gpu, but it does not work.