Apex Compute has opened beta access to its frame generation tech, and older GPUs are part of the deal. Its Apex Neural Rendering software aims to bring AI-made frames to cards that lack native support for NVIDIA DLSS Frame Generation. GeForce, Radeon, and Intel Arc users all get a shot, though game support may vary.
Hardware support is quite broad. NVIDIA GTX 16 cards and RTX 20 models make the cut, along with newer GeForce chips. AMD users need at least a Radeon RX 5000, while Intel Arc A and B cards are on the list as well. Even the Steam Deck gets a mention.
On the software side, Windows 10 and 11 work with the beta. Linux and SteamOS users can run it through Proton. OptiScaler handles the setup, with different files for games that have it installed and those that don't. But each game may act differently, and a supported GPU alone won't ensure it works.
Here's where the numbers get neat. On an RTX 5090 at 4K, Apex measured just 1.01 ms to make a frame at 2× mode. DLSS took 1.26 ms in the same test. That's a clear lead in this one task, at least by the firm's own data.
Memory use shows a wider gap. Apex lists 214 MB for its frame tool, while DLSS used 562 MB. Nice if those figures hold up. But that's just the memory needed for this step, not the full game. Textures, other buffers, and assets still eat VRAM.
Actual play is another matter. A quick frame generation step won't fix a slow base frame rate or a busy CPU. Input lag can still be an issue, and drivers may affect how well it runs. Apex has shared visual scores and game shots, yet outside tests haven't checked those claims.
As for other features, super resolution and ray reconstruction aren't ready for public use. Both remain in the plans. The beta only handles extra frames, while full source code has yet to land. GitHub has the install files and guides, but the core tech isn't open for a full code review.
The real test will come from players trying it on old cards, new GPUs, and a range of games. Frame quality, crashes, lag, and smooth play all need a fair check. For the moment, Apex Neural Rendering is a neat beta option, not a proven swap for DLSS, FSR, or XeSS.
Hardware support is quite broad. NVIDIA GTX 16 cards and RTX 20 models make the cut, along with newer GeForce chips. AMD users need at least a Radeon RX 5000, while Intel Arc A and B cards are on the list as well. Even the Steam Deck gets a mention.
On the software side, Windows 10 and 11 work with the beta. Linux and SteamOS users can run it through Proton. OptiScaler handles the setup, with different files for games that have it installed and those that don't. But each game may act differently, and a supported GPU alone won't ensure it works.
Here's where the numbers get neat. On an RTX 5090 at 4K, Apex measured just 1.01 ms to make a frame at 2× mode. DLSS took 1.26 ms in the same test. That's a clear lead in this one task, at least by the firm's own data.
Memory use shows a wider gap. Apex lists 214 MB for its frame tool, while DLSS used 562 MB. Nice if those figures hold up. But that's just the memory needed for this step, not the full game. Textures, other buffers, and assets still eat VRAM.
Actual play is another matter. A quick frame generation step won't fix a slow base frame rate or a busy CPU. Input lag can still be an issue, and drivers may affect how well it runs. Apex has shared visual scores and game shots, yet outside tests haven't checked those claims.
As for other features, super resolution and ray reconstruction aren't ready for public use. Both remain in the plans. The beta only handles extra frames, while full source code has yet to land. GitHub has the install files and guides, but the core tech isn't open for a full code review.
The real test will come from players trying it on old cards, new GPUs, and a range of games. Frame quality, crashes, lag, and smooth play all need a fair check. For the moment, Apex Neural Rendering is a neat beta option, not a proven swap for DLSS, FSR, or XeSS.