|

Nvidia Decreases Limitations on Consumer GPUs

Nvidia consumer GPUs just got a big performance increase.

Nvidia silently removed some of the encoding limitations placed on consumer GPU’s – increasing the number of simultaneous encoding streams from 3 to 5. This will increase the performance of consumer graphics cards for video encoding.

Multiple video encoding streams allow the graphics card to serve multiple simultaneous streams from captured video sources. This is most useful for making a single media source available to many destinations or even for immediate consumption. A good example of this is live streaming – you may wish to live stream at 720P 30FPS and have a 1080P 60FPS video encoded at the same time for on demand streaming that would be uploaded later.

This will result in better performance and make the lives of video enthusiasts easer, but won’t allow consumer cards to have the edge on professional and data centre grade graphics cards. It is also important to bear in mind that with more video encode streams running simultaneously, the speed of encoding can start to suffer.

Nvidia’s increase on the number of concurrent NVENC (Nvidia Encoder) on consumer GPUs from 3 to 5 is effective for products featuring Maxwell 2nd Gen, Pascal, Turning, Ampere and ADA Lovelace architectures (this excludes some of the MX-series products usually seen in laptops). This covers most of the products released in the last 8 years.

The number of simultaneous NVDEC (Nvidia Decoder) decodes isn’t restricted, but there are limits to how many real time decode streams can be run concurrently.

Historically, Nvidia has restricted the number of concurrent encoding sessions on all their graphics cards. Especially the consumer level hardware. With consumer GPUs having a maximum number of simultaneous encoding streams increased from 3 to 5, whereas Nvidia workstation cards  running the same silicone as consumer cards could support 11-17 concurrent encoding sessions depending on the quality of the hardware. A few years ago, these limitations could be removed by a relatively simple hack, but this has since been patched out.

Nvidia’s stance on encoding and decoding limitations has loosened a bit for the consumer level, but the real video processing horsepower remains with workstation and data centre-grade GPUs; there are no limits on these GPUs, with the number of concurrent sessions being limited by the hardware capabilities, codec choice and the quality of streams being encoded or decoded.

Similar Posts