Apex releases frame-generation beta with support for older GPUs
The company claims lower processing overhead and memory use than Nvidia DLSS, while plans to publish its source code remain unfulfilled.
Apex Compute is offering a public beta of Apex Neural Rendering, a frame-generation alternative that supports older graphics cards than Nvidia’s competing technology. The company claims performance and image-quality advantages over DLSS, though it has not published its source code or provided a full technical explanation.
Tests conducted internally by Apex put the processing cost per generated frame at 1.01 milliseconds in 2x mode on an RTX 5090 running at 4K. The corresponding DLSS figure was 1.26ms. Apex also reports approximately 62% lower VRAM consumption: 214MB for its system compared with 562MB for DLSS.
Compatibility reaches back to Nvidia’s GTX 16 generation and AMD’s Radeon RX 5000 family. Supported hardware also includes Intel Arc A- and B-series cards and the Steam Deck. A separate Metal test kit provides support for Apple M1 and later hardware. Nvidia says its AI-based frame generation depends on hardware first introduced in the RTX 40 series.
In videos featuring Hogwarts Legacy and The Last of Us Part II, Apex presents comparisons showing fewer ghosting artifacts than DLSS. Work on super resolution and ray reconstruction remains unfinished.
The beta can be downloaded from GitHub or Apex’s website and uses OptiScaler. Apex intends to open-source the technology. Its GitHub repository also offers a comparison tool for capturing frame-generation inputs losslessly and replaying them through rival systems.
