I have a PC with 2x RTX2070 plugged in PciEx gen3.0 x16.
On card 0 I run TensorRt and Cuda computing on second card NvEnc. But I am limited to three session of NvEnc even when I have 2 cards. (https://developer.nvidia.com/video-encode-decode-gpu-support-matrix) I thought that each graphic is limited to 3 sessions, but this limitation is apparently per system, not per graphic.
So I do want to buy Nvidia Quadro RTX4000 which has unlimited NvEnc sessions. So the system will be like this :
- PciEx gen3.0 x16 - RTX2070 (Cuda, TRT)
- PciEx gen3.0 x16 - RTX2070 (Cuda, Procession of large amount of operation on matrixes)
- PciEx gen3.0 x4 - Quadro RTX4000 (NvEnc -> 6 sessions of NvEnc)
My question is, Will the limitation of NvEnc disapper, because I will start to use Quadro RTX4000 or the limitation of NvEnv Per System will be 3 until i unplug all RTX2070 and leave just Quadro RTX4000?
My second question is, how big the problem is, plugging Quadro RTX4000 to PCIex gen3.0 x4? Now the streams/encodes take up tu 20% of one RTX2070, so it is not very heavy on the GPU.
My last question is : Is the limitation of NvEnc really per system not per graphic?
Thank you for your time.