One of the first practical AI functions in Windows that will not rely on the cloud, but will use neural coprocessors directly on the computer, will be called Car SR (Automatic super resolution). Its task will be to scale the image from a low resolution to a higher one in real time and reconstruct the missing details.
We’ve known that something similar is coming since December, thanks to insider information from Zac Bowden of Windows Central magazine. The novelty should be part of the autumn release of Windows 11 24H2, which Microsoft is already publicly testing. And some already have it in the settings System | Screen | Graphic Design new item Auto SR.
It should automatically improve the image with one click, and it is even possible to select only specific applications or games in which the function will be active. We don’t have the feature available yet, so we can’t test it.
However, Microsoft has been offering the related Video Super Resolution function as part of the Edge browser since last year. It can rescale videos in resolutions from 192p to 720p and higher. However, it uses acceleration with the help of a graphics card, which is computationally very demanding, so consumption and noise suddenly skyrocket, and of course the laptop’s endurance will be shortened.
Integrating directly into the system would ensure that upscaling can also work in, say, the VLC player. And thanks to economical NPUs, it should work quickly and without increased consumption. Since last year, neural coprocessors have been part of the Phoenix family of processors from AMD and this year’s Meteor Lacquers from Intel. And they will also be in Qualcomm’s upcoming Snapdragon X Elite. However, the Auto SR function could theoretically use the AI cores that are part of the GPU (Nvidia’s Tensor or Intel’s XMX), but we have not confirmed this.
Graphics drivers from AMD and Nvidia already include functions to increase the resolution from lower to higher. Radeon Super Resolution or Nvidia Image Scaling are still just stupid sharpening filters, no AI with image reconstruction. We already have such technologies (DLSS, XeSS), but they require support from the game, they cannot be launched everywhere by clicking.
The main challenge for Auto SR will be making sure the feature doesn’t add too much latency to gameplay. Every post processing slows down the rendering, and the reaction time from sending the command to the display on the monitor increases. This is not a problem for video, but in games where almost every millisecond counts, such technology may not be desirable.
Source: PhantomofEarth via Videocardz
2024-02-12 12:52:41
#Windows #learns #enhance #lowresolution #video #games #click
%20ES%20QNs%20Year%20of%20the%20Dragon%20CG_128897496_4646)
