Apex Neural Rendering: new DLSS rival promises faster frame generation on NVIDIA, AMD and Intel GPUs

Apex Compute announced on Thursday (12) Apex Neural Rendering (ANR), a suite of neural rendering technologies that aims to compete with NVIDIA's DLSS — with one key difference: it works on graphics cards from all three major vendors, NVIDIA, AMD and Intel.

ANR brings together three components: frame generation (ANR FG), super resolution (ANR SR) and ray reconstruction (ANR RR). For now, only frame generation is available in public beta, distributed via GitHub and integrated into games through OptiScaler. The super resolution and ray reconstruction models are still in training, according to the company.

Broad compatibility is the big selling point

According to Apex Compute, ANR works with GeForce GTX 16 and RTX 20, 30, 40 and 50 series, Radeon RX 5000, 6000, 7000 and 9000 series — including the Steam Deck — Intel Arc A- and B-series, and even Apple silicon (M1 or later). The beta runs on Windows 10 and 11, Linux and SteamOS via Proton, and macOS 26 or later.

The numbers come from the company itself — and are not yet validated

It is important to separate announcement from proof: all performance figures released so far were measured by Apex Compute itself and have not yet gone through independent testing.

According to the company, on a GeForce RTX 5090 running at 4K, ANR FG generates a frame in 2x mode in 1.01 ms, versus 1.26 ms for DLSS Frame Generation; in 4x mode, the average drops to 0.61 ms, versus 0.82 ms for DLSS. Video memory usage would be 214 MB, versus 562 MB for DLSS, and the neural network weights would take up just 0.17 MB. The company also claims ANR delivers higher image quality, with demonstrations in The Last of Us Part II Remastered and Hogwarts Legacy.

In its official post on X, Apex Compute said ANR delivers "up to 2.4× faster performance and higher quality rendering than NVIDIA DLSS". That is the company's claim — a STATEMENT, not a verified fact — and only independent testing will confirm whether the advantage holds up in real games.

Another caveat: although part of the code is on GitHub, no clear license has been set yet, so the project cannot be described as open source for now. Installing the beta is also not as simple as toggling an option in a game's settings: it requires modifying each game's directories, and compatibility varies between titles.

If the numbers hold up, ANR could be a relevant alternative, especially for owners of older cards or other brands that are currently left out of DLSS. For now, the way forward is to test it: the beta is already available to download.

Image credit: official Apex Neural Rendering comparison image, published by Apex Compute on its official website.

Sources

Transparency: this article was produced with the help of artificial intelligence, based on cross-referencing public posts on X and sources on the internet, and reviewed against the original sources cited. The performance claims come from Apex Compute and have not yet been validated by independent testing. Check the original sources to confirm the details.

By Muse