Thanks again for the input MeeLee
I am waiting for delivery of an extra psu so i dont have to worry about the power consumption. I the mean time I have been attempting to get a systemd service running at boot but upto now it fails on reboot but will load from the cli with systemctl start nvpower.service. Here is a copy of the bash script:
#!/bin/bash
DATE=`date '+%Y-%m-%d %H:%M:%S'`
echo "Titan XP Pascal Power Service started at ${DATE}" | systemd-cat -p info
sudo nvidia-smi -i 1 -pl 170
sudo nvidia-smi -pm 1
exit 0
And here is a copy of the nvpower service script:
[Unit]
Description=nVidia Geforce Titan Power Reduced to 170W.
[Service]
Type=simple
ExecStart=/bin/bash /usr/bin/nvpower.sh
[Install]
WantedBy=multi-user.target
As mentioned... this script works from the command prompt and when I nvidia-smi its shows the power capped at 170W. I've reloaded the systemd daemon, enabled the .service to create a symlink and rebooted... it doesn't load.
Any idea what i'm doing wrong?
K
HPZ600 & Geforce Titan XP Pascal - Linux
Moderators: Site Moderators, FAHC Science Team
-
Kevin_Burton
- Posts: 8
- Joined: Sun May 24, 2020 2:30 pm
Re: HPZ600 & Geforce Titan XP Pascal - Linux
If the script works just executing in terminal, did you do:
If you did, there must be different procedures for modern operating systems preventing shell scripts to execute on boot.
Side note:
pm 1 isn't necessary.
It's persistence mode, and more than likely keeps the GPU in high boost, and is like putting the GPU in 'performance mode' in x server.
It doesn't affect folding.
It's better for burst loads (where the GPU frequently goes from near idle, to near full load) in a matter of seconds, so it gains performance speeds by not having to spend time ramping up to full clock speed.
The GPU will remain in boost anyway between WUs. Enabling this feature won't allow the GPU to cool as quickly between WUs (if it takes longer than a few seconds).
The cooling properties may be more beneficial to folding, than keeping it in high boost.
For FAH, your GPU already is running at full frequency when folding.
What's slightly worse, is that when a GPU WU hangs, the GPU will continue to use more power, until the problem is fixed, than it would with PM disabled.
Code: Select all
sudo update-rc.d /etc/init.d/nameofscript.sh defaultsSide note:
pm 1 isn't necessary.
It's persistence mode, and more than likely keeps the GPU in high boost, and is like putting the GPU in 'performance mode' in x server.
It doesn't affect folding.
It's better for burst loads (where the GPU frequently goes from near idle, to near full load) in a matter of seconds, so it gains performance speeds by not having to spend time ramping up to full clock speed.
The GPU will remain in boost anyway between WUs. Enabling this feature won't allow the GPU to cool as quickly between WUs (if it takes longer than a few seconds).
The cooling properties may be more beneficial to folding, than keeping it in high boost.
For FAH, your GPU already is running at full frequency when folding.
What's slightly worse, is that when a GPU WU hangs, the GPU will continue to use more power, until the problem is fixed, than it would with PM disabled.