Nvidia Teases Gaming Future, Gamers Hold Breath

One use for this is neural texture compression, something that Nvidia has been working on for a while, but fancy APIs aren’t strictly necessary to pull this off, as Ubisoft has shown with Assassin’s Creed Mirage. But a strong suspicion remains that if Huang does focus on this, it will actually be all about making path tracing better and faster, as this is exactly what Nvidia was promoting at the GDC event last week.

Can the average gamer even tell the difference between path tracing and regular ray tracing when they are too busy fighting stutters? Had the RAMpocalypse not come to pass, the industry almost certainly would have seen the Super refresh of the RTX 50-series cards by now, and while it’s still distantly possible that these cards do get announced, it now seems extremely unlikely given just how bad the DRAM situation is (neatly making one tech pundit the world’s worst tech prophet in the process).

New RTX cards usually appear alongside some kind of new DLSS or RTX software feature, but the market has already seen those, in the form of RTX Mega Geometry foliage system and DLSS Dynamic Multi Frame Generation. Nvidia was probably going to keep these for the Super launch, but their appearance suggests that no new gaming GPUs will be coming from Team Green for a long time now. Does anyone actually need a $1,500 graphics card to play the same indie pixel-art games they enjoyed five years ago?

DLSS Ray Full Construction Generates Fake Photons

All things considered, Huang is probably going to just reiterate what was said at the GDC, which is fine because anything that can be done to improve the performance of rendering, without the loss of visual fidelity, is certainly a good thing. Gamers just want their frames, and if Nvidia can deliver them without melting the power grid, everyone wins. But wait, the speculation hasn’t touched on anything truly pie-in-the-sky yet.

Okay then, how about DLSS Ray Full Construction, an AI system that doesn’t just denoise a ray-traced scene but actually uses machine learning to generate thousands of additional fake rays? It sounds like science fiction, but if any company has the silicon to brute-force that kind of nonsense, it is Nvidia. Imagine a world where the GPU renders one out of every ten photons and just guesses the rest.

Jensen Talks, Gamers Dream, Reality Disappoints

Ah no, a better prediction exists. Huang will hop onto the stage with a new GeForce RTX graphics card. It will have 10,000 CUDA cores and 4 GB of VRAM, but it will be the first to use RTX Ultra Memory, an AI-powered system that neurally compresses everything automatically and does it so quickly and so well that it effectively quadruples the VRAM and its bandwidth on the graphics card. Yours is only $1,299.

It is a classic Nvidia move: sell a solution to a problem that only exists because of their own hardware limitations. The keynote will likely be entertaining, the slides will be shiny, and the audience will clap on cue. But until actual hardware arrives on shelves that doesn’t require a second mortgage, the future of real-time rendering will remain firmly in the realm of keynote fantasy for most PC gamers.