A new competitor is entering the neural rendering market, with Apex Compute introducing its Apex Neural Rendering (ANR) framework as an alternative to NVIDIA’s DLSS technology. The company claims its solution can deliver faster frame generation, lower GPU memory consumption, and image quality that matches or exceeds DLSS in certain scenarios. The framework combines three technologies: ANR Frame Generation (ANR FG), ANR Super Resolution (ANR SR), and ANR Ray Reconstruction (ANR RR).
According to Apex Compute‘s initial benchmarks, ANR Frame Generation achieves a latency of just 1.01 ms per generated frame at 4K resolution using the 2x setting on NVIDIA’s GeForce RTX 5090. For comparison, NVIDIA DLSS Frame Generation reportedly takes 1.26 ms under the same conditions. The company has not disclosed which DLSS version was used during testing, making it difficult to establish a direct comparison with NVIDIA’s latest implementation.
Apex Compute also claims a significant reduction in GPU memory usage. ANR FG reportedly consumes only 214 MB of VRAM, compared to 562 MB for NVIDIA DLSS Frame Generation. This represents approximately 62% lower memory consumption. The company further states that its ANR FG neural network weights occupy just 0.17 MB, allowing the technology to operate with a relatively small memory footprint.
Image quality is another area Apex Compute is highlighting. Samples published on the company’s website show comparisons between ANR and DLSS, with the company claiming that its technology can deliver comparable results and better visual fidelity in some situations. Demonstrations featuring Hogwarts Legacy and The Last of Us Part II Remastered also showcase ANR Frame Generation running in supported games.
The current ANR FG implementation is available as a beta injector built around OptiScaler. It is designed to work with games that already support NVIDIA DLSS Frame Generation, potentially allowing users to experiment with the technology without waiting for individual game developers to introduce native ANR support.
Hardware compatibility is also extensive. NVIDIA support includes GeForce RTX 20 series graphics cards and newer models, as well as the GeForce GTX 16 series. AMD Radeon support starts with the RX 5000 series and extends to the Steam Deck. Intel Arc A-series and B-series graphics cards are supported, while Apple M1 processors and newer are also listed as compatible.
Apex Compute has additionally tested the framework across Windows 10, Windows 11, Linux, SteamOS through Proton, and macOS 26 or later. This makes ANR a potentially interesting option for users across different GPU platforms and operating systems.
The company plans to open-source the Apex Neural Rendering framework, including its individual components, with the code expected to arrive on GitHub soon. However, the published performance and image quality results currently come from Apex Compute, so independent testing will be important to establish how ANR compares with NVIDIA’s latest DLSS technologies across different games and hardware configurations.
