DLSS is a new and improved deep learning neural network that boosts frame rates and generates beautiful, sharp images for your games. We’re joined by Richard Cowgill from NVIDIA who will demonstrate how the technology can be used in Unreal. We’ll also dive into RTXGI which provides infinite bounce lighting in real time, removing the need for offline lightmap baking. Plus, a conversation around the future possibilities of Caustics!
Thursday, September 3 @ 2:00PM ET - Countdown
It was a great stream but I think it was misleading about how easily we can actually get access to the DLSS part. You have to apply and their website says its ‘available to select developers through an early access program’. The application involves giving the name of your app/game and providing urls to info about the product. Thats not applicable to me at this early stage where I dont have a project in particular nailed down, I just want to evaluate the tech in general.
The RTXGI bit was much easier to get access to, I got that working last night and I am quite pleased with it so far.
I was under the impression that DLSS 2.0 no longer requires game-specific training on Nvidia’s servers (as required with 1.0), and works out of the box for any game (as long as DLSS support is compiled into the game)?
Have you tried enabling it in the NvRTX build as shown in the stream and here in the docs:
I’ve just downloaded the code over lunch – I’ll be building and experimenting with this over the weekend
It doesnt require game-specific training these days but if you read the readme for that branch on github it goes on about the required registry files, and that was also mentioned in the stream, albeit I believe the stream downplayed the complications of this aspect.
For that reason I built that branch but only tried the RTXGI bit so far, because I dont expect DLSS to work. If their instructions are just out of date then I will be wrong about that, but sadly I dont expect to be.