Some updates:
The question was about K2200 (maxwell gen 1) not K4200 (eg.K4000 described in OLD nvenc SDK pages), but Ok. I did not received any official answer from nvidia (eg. > 2 month).
BUT:
There is NEW nvidia video codec SDK version 6 and nvidia changes 2+ encoder sessions support - [url]https://developer.nvidia.com/nvidia-video-codec-sdk[/url].
New SDK (by webpage, not verified) supports 2+ encoder sessions on more cards (including >K2000 also maxwell gen 1 K2200) and new maxwells M* too. There is also removed licensing bug if you combine non-low-end and low-end/geforce card in one system (<=2 encoder session enforced on low-end/geforce card only, eg. N (unlimited for non-low-end) + 2 (for all low/geforce) encoder sessions allowed per system). Now I am slightly happier.
Thanks for this thread, lots of information useful to me that I can’t seem to find anywhere else. Can anyone find any information on single engine NVENC performance benchmarking across the NVIDIA card line? Does a single thread on the maxwell gen 1 Quadro K2200 beat a single thread on an entry level maxwell gen 2 GeForce 960?
My project requires me to transcode many short video clips (6s) very fast, and I’d like to run two simultaneous threads at the fastest speed with about a $600-800 video card budget. I’m thinking about just running a single thread on each of two GeForce cards.