Apex Compute has unveiled its Apex Neural Rendering (ANR) rendering toolkit, which is said to deliver higher quality and performance than Nvidia’s DLSS solutions while being compatible with older Nvidia GeForce, AMD Radeon and Intel Arc cards.
Image source: TechPowerUp
Apex Neural Rendering includes the ANR Super-Resolution (ANR SR) upscaler, the ANR Frame Generation (ANR FG) frame generator, and the ANR Ray Reconstruction (ANR RR) ray reconstructor. The frame generation component is already available in beta, while the image upscaling and ray reconstruction models are still in training and will be available later.
When using the Frame Generator (ANR FG) on an Nvidia GeForce RTX 5090 at 4K (2x upscaling), latency is reported to be 1.01ms per frame generated, compared to 1.26ms for Nvidia DLSS Frame Generation. Although the version of DLSS FG involved in testing is not specified, the resource TechPowerUp believes that we are talking about the most current version.
And if ANR FG uses only 214 MB of video memory, then DLSS FG requires more than twice as much – 562 MB.
It is reported that the weights of the ANR FG neural network occupy only 0.17 MB, which indicates high efficiency. The examples provided on the company’s website confirm that the image quality provided by ANR is not inferior to DLSS, and in some cases even surpasses it.
Apex Compute also demonstrated the technology in the games Hogwarts Legacy and The Last of Us Part II Remastered: compared to DLSS, its solution provides higher visual fidelity when generating frames. ANR FG is a beta version of a software module (injector) based on OptiScaler. It can be used in any game that supports Nvidia DLSS Frame Generation technology.
The list of those receiving ANR support includes Nvidia GeForce RTX 20 series and newer video cards, as well as the GeForce GTX 16 series; AMD video cards, starting with the Radeon RX 5000 series; Compatibility with the Steam Deck portable console is also stated. In addition, Intel Arc A and B series graphics cards are supported, as well as Apple devices based on M1 processors and newer chips.
Operating systems tested and supported include Windows 10, Windows 11, Linux, SteamOS with Proton, and macOS 26 and later.
Apex Compute also announced plans to open source the ANR framework (including all its components) and host it on the GitHub platform.
If you notice an error, select it with the mouse and press CTRL+ENTER.
