Switch Pro, Switch 2, Super Switch, Switch Model S: it’s not entirely clear what it will actually call or when it arrives, but the rumors do not stop and an ongoing topic is the use of Nvidia’s DLSS – super deep learning -sampling – the way to combine “traditional” temporal super-sampling with a neural network to dramatically increase resolution. It has been put to good use on PC and its job on a new Switch would be to take the 720p image designed for the mobile screen and then scale it up to 2160p for a nice presentation on 4K displays. today. That’s the theory, but in practice, is DLSS really viable for a portable device? We decided to run the numbers.
First, we need to look at the plausibility of a true DLSS compatible processor ending up in a mobile console. There at least there are positive responses. Nvidia’s latest Tegra (system on chip) SoC is codenamed Orin, it is based on the latest Ampere architecture and it is primarily designed for the automotive industry which has our first problem – it has a power budget of 45 W, in a world where the first Switch peaked at 15W, with 10-11W probably being used by the main processor itself (and probably half of it in mobile mode). There is a workaround, however: There is a very low power render of Orin that runs at just 5W by default, but should easily scale with more power consumption and more performance when connected.
Could a Next Generation Switch Really Provide 4K DLSS? Here’s Alex Battaglia with our video breakdown.
Even so, despite incorporating the tensor cores needed to make DLSS a reality, we’re still looking at a small power budget – and scaling AI isn’t ‘free’, the next step. of our experiments therefore consists in measuring the computational load of DLSS. You will see in the video how we did it, but the calculation method is quite simple. Using Doom Eternal as a basis, we used an RTX 2060 to measure the time it takes to process DLSS render versus native resolution and came to the conclusion that the process render time cost is 1.9ms. . The RTX 2060 has about 5.5 times the machine learning capability of the Orin chip operating at 10W, so assuming linear scaling, the PC iteration of DLSS would require a substantial processing time of 10. 5 ms on a future Switch Pro. In a world where Doom Eternal targets 16.7ms per frame, that’s just too high. However, for a 30fps game with a 33.3ms render budget, this is very, very viable.
It’s important to point out that our measures on the cost of DLSS are based on comparing inputs and outputs and we don’t have access to all the inner workings, so the calculations are approximate, but this gives you a idea of the viability of the technology for a mobile platform. And there are a whole bunch of additional variables that we need to take into account. For starters, we assume 4K output. Nothing prevents a developer from using DLSS to turn that 720p mobile image into a DLSS 1440p output, and then using the GPU scaler to deliver the final 4K output. In this scenario, the 10.5ms cost of DLSS in a mobile Nvidia chip drops to just 5.2ms. There would be some loss of quality, of course, but it may be suitable for some games or some visual content.
Of course, we’re basing all of our testing here on a PC implementation of DLSS and it’s certainly not impossible for Nvidia to optimize the technology for a console experience – remember the company even created its own low-level graphics API. just for Switch. It is the commitment. But beyond these variables and our testing, there is actually something just as important, if not more, can DLSS actually look decent by transferring a 720p image to 2160p?
Hopefully the screenshots on this page illustrate precisely why DLSS is a potential game changer for a future next-gen Switch, and how an image rendered for a 720p mobile screen can actually turn into a perfectly viable living room presentation. DLSS not only increases scaling, but it anti-alias the image in the process – a two-in-one solution, if you will. By combining information from previous images with motion vectors that tell the algorithm where pixels are likely to move and informing all of this with deep learning, the effect is transformative.
Does a DLSS 4K image compare favorably to a native 4K presentation? In the PC space, performance mode uses a native 1080p input for scaling, switching to quality mode at 1440p. Scaling from 720p means a lot less data to work with, which means more inaccuracies and artifacts in the output image – but the point is that a future next-gen switch won’t does not need to provide native 4K quality. The criteria for success are very different when comparing a PC experience up close by users demanding a high-end experience to a more mainstream console gamer watching the action unfold from a distance from a living room flat screen. . The new Switch doesn’t need to deliver 4K precision, it just needs to deliver an image that looks good on a typical modern living room TV.
Beyond DLSS itself, there are obviously a lot of theories in this article. For starters, why would we assume Nintendo would target a 720p display for a next-gen Switch? From our perspective, that’s the best balance between GPU power and pixel density – and it’s no mistake that Valve has targeted an equally dense display with its own 800p display for Steam Deck. Going from a 720p target to 1080p essentially means that much of Switch’s generational leap to its successor would be spent on more pixels, as opposed to better quality pixels.
DLSS 2.2 is the latest version of Nvidia’s machine learning-based scaling technology. This video shows its application in Lego Builder’s Journey, then takes a look at the latest DLSS 2.2 innovations in other titles.
The next big guess is that Nintendo will stick with Nvidia for the next-gen Switch (which is basically a no-brainer at this point – if only for compatibility reasons) and the company will actually use the chip. Orin, or a variant of it. On this last point, one of the most precise leakers of the moment – kopite7kimi seems pretty convinced. Despite that, our calculations for DLSS in a new Switch are based on a power budget of around 10W – and there’s nothing stopping Nintendo from going further than that, with adequate cooling.
But the purpose of this test was basically twofold: First, to determine if a mobile Nvidia chip based on the latest architecture could actually run DLSS – and the answer is yes. It’s just that we have to be aware that while DLSS is based on hardware acceleration through core tensile, it is not a “free” upscaler – there is a cost to using it. The question is whether this cost can be offset by increasing the power and frequency when docked, but the bottom line is that, theoretically, DLSS is viable.
The second question concerns quality. Even with a mere 720p resolution to manage, DLSS still produces good 4K results – stunning and remarkable when put side-by-side with the input, even if it is “not as good as native”. The upcoming apps are also enticing – and not just for Nintendo consoles. Machine learning apps – including supersampling – are also a way forward for “Pro” renderings of PS5 and Xbox Series consoles, if that’s the direction Sony and Microsoft want to take.
Article source https://www.eurogamer.net/articles/digitalfoundry-2021-in-theory-can-switch-pro-handle-4k-dlss