AMD can create something like Nvidia’s DLSS

AMD has been in a rather uncomfortable position for many years, in which, instead of revolutionizing technology on the graphics market, it has to keep up with the competition, offering replacements for what it has. This is mainly seen with FreeSync (responses to G-Sync), which AMD approached with a much better strategy. Now the company is thinking of its own competitive solution to Nvidia’s DLSS rendering.

It’s all about the DirectML API, which Microsoft introduced right after DirectX Raytracing in 2018. The issue of this DirectX 12 add-on was raised by Adam Kozak from AMD, who revealed that Radeon VII is doing great in DirectML. In short, this API allows developers to implement algorithms in their applications that use the field of artificial intelligence, namely machine learning. This allows AMD to present “its DLSS technology” in the future, but with a completely different approach.

An example of this is the Radeon VII result higher by 62% compared to the RTX 2080 in the Luxmark test. This one uses a derivative of OpenCL’s real-time ray-tracing rendering, but relying on something like GPGPU. And this is perfectly explained by our wonderful Wikipedia, from which I allowed myself to copy the most important point:

GPGPU – A technique whereby the GPU, usually only concerned with computing related to computer graphics, allows general-purpose computing to be performed just like a CPU. As a result, many calculations, mainly parallel calculations, can be performed much faster.

Of course, you can do much more with DirectML, but it requires both a lot of work and producers willing to use novelties (e.g. games), who are often unwilling to cooperate, relying on proven solutions, while minimizing the risk of budget depletion. See – Nvidia Hairworks.

Also read: AMD is working on a technology that competes with Nvidia RTX

Source: Wccftech