Exact amount of NVENC engines in different cards

The NVIDIA VIDEO CODEC SDK APPLICATION NOTE - ENCODER has a blurred sentence: ‘certain
variants of the second generation Maxwell GPUs and Pascal Generation GPUs have
two/three NVENC engines physically present’. Can somebody tell exact number for each available card (or at least for Quadros)?

Now officially published https://developer.nvidia.com/video-encode-decode-gpu-support-matrix.

Thank you! What about GeForce cards?

The count should be the same on card with the same chip (GKxxx, GMxxx, GPxxx) see https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units. But be aware if you buy partially faulty chip (not all SM available) the memory bus may be narrower and the secondary encoder unit can missing too. There is also crippled API if you use GeForce or “low end Quatro?” (this information is missing now) you open only two encoding stream per system (see https://developer.nvidia.com/nvenc-application-note - “3. NVENC LICENSING POLICY”).

Is this an official information? (I’ve seen that you post many replies here, are you work for NVIDIA?)

I was thinking the same, however in the support matrix we have Tesla M6 and M60, both with the same GM204 chip (M6 has one and M60 have two of them), but both have two NVENC chips. Which tells us that main chip type doesn’t define number of NVENC chips.

No. I am not affiliated with NVidia in any sense nor NVidia NDA. NVidia people have NVidia green answer and NVidia icons. They respond to questions very rarely. I will never try to ask for any NVidia answer (see https://gridforums.nvidia.com/default/topic/823/grid-boards/tesla-m10/). If you need official answer ask NVidia directly or you can try wait here.

You looked at bad column. M6 and M60 have 2 NVEnc but M60 have 2 chips. (M6 has enabled less SM and less frequency to achieve less TDP (~75-100W) to be mountable inside blade chassis (MXM to mezzanine cards bridge)).

M.C>

The number of encoders in GPUs varies, and is published at https://developer.nvidia.com/video-encode-decode-gpu-support-matrix.

As a general advice, we discourage users from using this information as it can affect the performance. The NVIDIA driver load-balances among available encoders transparently.