Apex Compute has released a beta version of its Apex Neural Rendering (ANR) suite, a toolset designed to provide frame generation and image reconstruction across a broad range of hardware.
The new technology claims to offer up to 2.4 times faster performance than NVIDIA’s DLSS Frame Generation, alongside superior image quality, according to Esport1.
The ANR suite is intended to be an open-source solution that includes super resolution, ray reconstruction, and frame generation. Currently, only the frame generation module is available for public testing in its beta phase. Performance data provided by Apex Compute indicates that on an RTX 5090 GPU at 4K resolution, the technology generated a frame in 1.01 milliseconds in 2x mode, compared to 1.26 milliseconds for NVIDIA DLSS.
In 4x mode, the ANR required 0.61 milliseconds against 0.82 milliseconds for the DLSS equivalent, as reported by Notebookcheck Magyarország. The developer also said that the ANR technology utilizes 214 MB of GPU memory, significantly less than the 562 MB required by DLSS.
Beyond these metrics, internal testing by the company showed that ANR outperformed AMD’s FSR 3.1 on the RX 7800 XT, Arc B580, and Arc A770 graphics cards. The company also reported that its solution maintained lower resource requirements across 2x, 3x, and 4x modes compared to NVIDIA’s technology during 4K testing.
The technology features broad hardware compatibility, supporting NVIDIA RTX 20-series through 50-series cards, GTX 16-series cards, AMD Radeon RX 5000 through 9000 series, and Intel Arc A and B series GPUs. It is also compatible with the Steam Deck and Apple M1 or newer chips.
The software supports Windows 10 and 11, as well as Linux and SteamOS via Proton. Users can install the beta as a standalone tool or as an extension for OptiScaler. Because the installation requires modifying individual game directories, compatibility may vary by title, and users can activate the features directly within the game settings.
While the company has showcased image quality comparisons using The Last of Us Part II Remastered, independent testing has yet to be conducted to verify these performance claims. Apex Compute is still in the process of training its super resolution and ray reconstruction models, and the full source code is scheduled for release at a later date.
