Hello fellow Nvidia users,
We use Tesla T4 running on Ubuntu for video transcoding. In order to avoid overloading the GPUs we monitor the encode and decode utilizations. We’ve noticed that the readings returned by ‘nvidia-smi dmon’ and ‘nvidia-smi stats’ are inconsistent. The latter provides readings that are approx half of the former. See below where one returns 6% and the other 13-14%.
Why this inconsistency?
Note: We also use Tesla M60 and the numbers there are concistent across the 2 commands.
$ nvidia-smi stats -d encUtil -c 1
0, encUtil , 1593541909318963, 6
0, encUtil , 1593541909519612, 6
0, encUtil , 1593541909720366, 6
0, encUtil , 1593541909921131, 6
0, encUtil , 1593541910121965, 6
$ nvidia-smi dmon
# gpu pwr gtemp mtemp sm mem enc dec mclk pclk
# Idx W C C % % % % MHz MHz
0 27 47 - 3 1 14 2 5000 585
0 27 47 - 3 1 14 2 5000 585
0 28 46 - 3 1 13 2 5000 585
$ nvidia-smi
Tue Jun 30 13:32:24 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.64 Driver Version: 430.64 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 Tesla T4 Off | 00000000:00:1E.0 Off | 0 |
| N/A 46C P0 27W / 70W | 216MiB / 15109MiB | 3% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 2313 C ...ocal/WowzaStreamingEngine/java/bin/java 205MiB |
+-----------------------------------------------------------------------------+