Switch 2’s DLSS upscaling might not be as powerful as hoped, it’s claimed

videogameschronicle.com:

The successor to the Nintendo Switch may not include Deep Learning Accelerator tech for its upscaling, it’s claimed.

Numerous reports (including VGC’s own) have suggested that Nintendo‘s next console will use Nvidia‘s DLSS upscaling technology.

DLSS (deep learning super sampling) is a feature exclusive to Nvidia graphics cards. On PC, it works by using AI to upscale the resolution of games, effectively allowing developers to achieve higher graphical settings and better frame rates on weaker hardware.

However, some of Nvidia’s DLSS solutions also use extra tech called Deep Learning Accelerator (DLA), which uses a standalone engine designed to speed up AI calculations.

Because deep learning calculations can be computationally ‘expensive’ – meaning they can use up a lot of processing power – DLA is designed to take on most of the heavy lifting to free up that power so it can be used elsewhere.

In the latest episode of Digital Foundry’s DF Direct show, founder Richard Leadbetter stated that, according to his sources, the Switch’s successor will feature DLSS, but won’t have DLA accompanying it.

“There was the question mark over the cost of DLSS and whether Nintendo is going to include a Deep Learning Accelerator similar to the one that was in the T234 [GPU], which would effectively make DLSS ‘free’, or at least a lot less computationally expensive,” Leadbetter said.

Referring to the T239 GPU which is rumoured to be in the next console, he added: “I’ve had sources come forward saying there’s no DLA in the T239, which would limit the viability of DLSS quite significantly.”

Leadbetter estimated that if the Switch’s successor has DLSS without DLA, it would be unlikely that the console would be able to upscale to 4K, and that more