Why does the GPU memory used by process not add up to memory used according to nvidia-smi?

chase@ubuntu2:~$ nvidia-smi
Mon Dec  9 23:36:35 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.70       Driver Version: 418.70       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  GRID T4-2Q          On   | 00000000:00:0B.0  On |                  N/A |
| N/A   N/A    P0    N/A /  N/A |   1597MiB /  2048MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0     57230      G   /usr/lib/xorg/Xorg                           423MiB |
|    0     57309      G   /usr/bin/gnome-shell                         145MiB |
|    0     63736      G   ...request-channel-token=42824295405078622   383MiB |
|    0     73857      G   ...se/.nvm/versions/node/v12.10.0/bin/node     3MiB |
|    0     73870    C+G   ...se/.nvm/versions/node/v12.10.0/bin/node   236MiB |
|    0     73872    C+G   ...se/.nvm/versions/node/v12.10.0/bin/node   133MiB |
+-----------------------------------------------------------------------------+

423+145+383+3+236+133 = 1,323 MB

Because the GPU has overheads associated with certain types of activity that are not directly associated with a particular user process. The major activity overheads here would involve CUDA (C) and display (G).