[Hardware] NVIDIA RTX 3000 series graphics.

[Dark]

NVIDIA has updated information on the video formats that its graphics cards can hardware decode. It has done so by updating the table where it gives information about the video standards that are supported by the different NVIDIA GPUs, adding the new GeForce Ampere, the architecture used by the NVIDIA RTX 3000 series graphics. If you want to know what codecs your brand new RTX 3000 supports then read on.

One of the functions that a GPU can do is not only render graphics but also play video and generate it. It is therefore important to know what formats each family of graphics cards supports in case they suit our needs in that regard.

What video codecs does your NVIDIA RTX 3000 graphics support?
RTX 3000

From the data published by NVIDIA itself we can know the codecs supported for both encoding and decoding the latest GPUs of the brand, which will be useful not only for those who dedicate yourself to video editing and you are interested in one of these GPUs but also for those who want to dedicate themselves to streaming.

We leave you below the video formats that the RTX 3000 can encode and therefore those that can be generated by hardware:

It is very striking that the codec for streaming content par excellence at the moment because it has the best image quality / bit rate ratio, the AV1, can only be reproduced and video cannot be generated with it only with the GPU. It is one of the most missing things in current NVIDIA Ampere-architecture graphics as it is ideal for creating cloud gaming servers.

On the other hand, the good thing that AV1 can now be played is that video streaming services that make use of this format will consume less power on your PCs with an RTX 3000.

How does a GPU encode and decode video codecs?
NVENC NVDEC

As you may have observed, the entire RTX 3000 range has support for the same video encoding and decoding formats, and this is because it is not the part of the GPU responsible for rendering the graphics in real time that is responsible for decode, but uses specialized accelerators within the GPU itself, which are a type of domain specific processor that take care of that task.

These types of accelerators are found in all GPUs today and for several years regardless of the brand we are talking about; Sometimes new iterations of a graphics architecture may appear on the market whose only change is in the video encoder and decoder, in order to support new formats.

When a format is not fully supported (or is only partially supported) what is done is that the CUDA cores, within the SM or Compute Units of the GPU, are responsible for carrying out the decoding work. Both the NVENC and the NVDEC and equivalents of other brands occupy an area and a miniscule consumption compared to what is the rendering engine that encodes or decodes video when the hardware codecs do not support a format, and in some cases it even ends up being CPU involvement required.