The giant advantage of GeForce systems is support for DLSS technology, which significantly increases their computing power that can be practically used. However, competition is coming from AMD i Microsoft. Accessible and open to everyone.
Nvidia DLSS is a unique and amazingly working image reconstruction algorithm. The operation of its first version left a lot to be desired, but the second generation DLSS is almost a show of magic. And the amazing opportunities unlocked by properly trained artificial intelligence.
DLSS is incredibly effective in its image reconstruction. More than once showed during testingthat the image rendered natively in 1440p and reconstructed with DLSS up to 4K has better quality and higher detail than that rendered natively and with classic methods in Ultra HD. It uses much less graphics system resources than full resolution rendering.
DLSS it’s a huge advantage of systems GeForce. The counterpart, however, tends to Radeons. I Xboxa Series. And actually, also for GeForce.
We learn that the image reconstruction technology will appear in Radeon systems from the post about the Xbox Series S and X consoles. The graphical layouts of the RDNA 2 line contain computational blocks for accelerating the calculations related to machine learning. The interface for their operation was developed by Microsoft – it is called DirectML.
DirectML is to be available to game developers who create games for the Xbox Series and for personal computers. I would like to point out that it is not about PCs with Radeons, but simply personal computers. DirectML, including DirectML Super Resolution, is part of the API set from the family DirectX.
This means that any GPU with compute blocks to accelerate machine learning algorithms – including the new GeForce – will be able to take advantage of this interface. It is unclear what this means for PlayStation 5. After all, this console also uses the RDNA 2 chip, however, most likely DirectX is not one of the developer tools compatible with it. However, it is highly probable that AMD – as is customary – will provide an open source version of the interface as part of the GPUOpen / FidelityFX initiative.
It is unclear how effective DirectML Super Resolution will be. However, it already has the advantage of multiplatformity.
As noted above, it was only DLSS 2.0 that turned out to be a remarkable technology, but the first version of this mechanism did not impress. We don’t know how effective the AMD and Microsoft algorithms are. However, we know that these can be updated programmatically. Just like DirectX is constantly being improved.
DLSS is a mechanism that works only on PCs with a GeForce chip. Meanwhile, DirectML Super Resolution is an API present on the Xbox Series S | X, PC with Radeonem RX 6000, GeForce’em RTX and it is possible that also on PlayStation 5. The attractiveness of this solution is obvious for developers. This tool, if effective, will benefit more players.
However, hopes are high. DirectML is not a recent Microsoft project – but something the company has been working on for years. The above screenshot is taken from the 2018 presentation. It shows Forza Horizon 3 rendered in 1080p (right) and image reconstruction via DirectML to 4K (left).
There is something to fight for. The benefits in computing power are huge.
4K resolution at 60 fps is a rendering of almost 8.3 million pixels every 16.7 milliseconds. 8K increases this number to over 33.1 million pixels every 16.7 milliseconds. In turn, Full HD 1080p is only slightly over 2 million pixels. Each new resolution standard means a huge increase in the demand for computing power.
Rendering in 1080p with image reconstruction up to 4K is much less effort for the graphics system than rendering in 4K. After all, the freed computing power can be used for other purposes. To aid in ray tracing, for more precise shadows, textures, lighting, or anything else. And above all, for smooth animation while maintaining full Ultra HD resolution in practice.
So there is much to hope for. DLSS is a wonderful technology. However, it is its direct competitor that is likely to become the standard. For example, due to its openness and expected multi-platform nature.