Apex Compute has launched Apex Neural Rendering, a new upscaling and frame generation suite designed to compete with NVIDIA DLSS and AMD FSR. The system combines frame generation, resolution scaling, and ray tracing denoising into an open framework claiming faster rendering times and lower memory consumption than existing solutions. The software is currently available in beta form on GitHub, with only the frame generation component active at launch.
Apex Compute tested the technology on an NVIDIA GeForce RTX 5090 running at 4K resolution. The ANR frame generation tool produced a frame in 1.01 milliseconds compared to 1.26 milliseconds for DLSS Frame Generation. Video memory allocation measured 214 MB for ANR, whereas the NVIDIA solution required 562 MB, representing an approximate 2.5 times reduction in memory overhead. The neural network weights for the frame generator occupy 0.17 MB of storage. Visual comparisons released by the developers in Hogwarts Legacy and The Last of Us Part II Remastered show sharper detail retention than DLSS in specific rendering scenes.
Hardware support spans multiple graphics architectures. The software runs on NVIDIA GeForce RTX 20 series cards and newer, the GTX 16 series, AMD Radeon RX 5000 series and newer, Intel Arc A and B graphics, and the Steam Deck. Apple computers running M1 chips and newer are also supported. Compatible operating systems include Windows 10, Windows 11, Linux, SteamOS via Proton, and macOS 26+.
The current release operates through an injector based on OptiScaler. This setup enables players to activate the technology in titles that already support DLSS Frame Generation. Apex Compute published the full source code and system files on GitHub. Resolution scaling and ray tracing noise reduction tools will arrive in future updates.
