Hey guys, total newbie to the forums here. I’d love if you nice people could help answer a question I have.
I’m running Ubuntu 18.04 on a Dell Precision Tower 3620. I recently upgraded the graphics card to a RTX 2070 Super and got a 750W Corsair PSU to power it all.
However, the power limit is set to 215 W. This seems low given my 750W capable PSU. I’m currently taking a deep learning class and running the card up to 215 W training DNNs brings the card’s temperature up to 75 C. Am I right in thinking that there should be room to spare here? Wouldn’t setting a higher power limit allow for better card performance? Am I wasting my time?
Anyway, I tried following a previous thread on here by making nvidia-persistenced and nvidia-smi able to be run as sudo without a password, rebooting, and then running the following command, which unfortunately did not work. This was the command/output:
➜ ~ sudo nvidia-smi -pl 250
Provided power limit 250.00 W is not a valid power limit which should be between 125.00 W and 215.00 W for GPU 00000000:01:00.0
Terminating early due to previous errors.
This is a total bummer! Is there really no way I can override this setting? Or would increasing the power limit not even matter much?
If increasing the power limit isn’t useful, then what else can I do to boost performance?
Oh yeah, I should mention my driver info and stuffs!
Here is nvidia-smi after doing DNN training for 5 minutes:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.50 Driver Version: 430.50 CUDA Version: 10.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce RTX 207... Off | 00000000:01:00.0 On | N/A |
| 61% 73C P2 190W / 215W | 6555MiB / 7973MiB | 100% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1011 G /usr/lib/xorg/Xorg 278MiB |
| 0 1570 G /usr/bin/gnome-shell 159MiB |
| 0 4589 G ...quest-channel-token=6529682849136542766 51MiB |
| 0 28503 C ...e/seth/.virtualenvs/default/bin/python3 6033MiB |
+-----------------------------------------------------------------------------+
It says CUDA 10.1, but I actually have CUDA 10.0 on my system.
nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130