In brief: Nvidia DLSS is generally considered the gold standard in frame generation, but the feature is officially limited to RTX 40 and 50 series GPUs. A new competitor boldly claims that its solution, which it plans to open-source, is not only faster and better-looking, but also supports every graphics architecture since the GeForce GTX 16 and Radeon RX 5000 series.
Users can now download an open beta to test Apex Compute's claim that its neural rendering suite outperforms and outshines Nvidia DLSS frame generation. Super resolution and ray reconstruction (denoising) components are still in development.
Apex's YouTube clips show its frame generation producing less ghosting than DLSS in The Last of Us Part II and Hogwarts Legacy. Internal benchmarks also show that 2x frame generation on an RTX 5090 at 4K costs only 1.01 milliseconds per generated frame, compared to 1.26ms with DLSS. Apex also claims to cut VRAM usage by about 62%, from DLSS's 562MB to 214MB.
Hardware support is also substantially broader with Apex. Nvidia says its AI-driven frame generation requires hardware that debuted with the RTX 40 series. Although Apex hasn't yet released its source code or fully explained its technology, the solution supports Nvidia GTX 16 and newer, AMD Radeon RX 5000 and newer, Intel Arc A- and B-series, Apple M1 and later (through a separate Metal test kit), and the Steam Deck.
Available via GitHub and the company's website, Apex Neural Rendering runs through OptiScaler, with separate packages for games that already have it installed and those that don't.
Source link







