Apex Neural Rendering Promises DLSS-Beating Frame Generation On NVIDIA, AMD, And Intel GPUs
Apex Neural Rendering is here, and it wants to break NVIDIA’s grip on frame generation. Apex Compute has announced Apex Neural Rendering (ANR), a cross-platform neural rendering suite that claims up to 2.4x faster frame generation than NVIDIA DLSS, with better image quality and far lower VRAM use. Best of all for PC gamers on older hardware, it runs on NVIDIA, AMD, and Intel GPUs alike, and the frame generation part is already available as a free beta.
The announcement came through an official post from Apex Compute on X on November 11, 2026. If the claims hold up under independent testing, ANR could become the most interesting graphics tool of the year for PC players.
What Is Apex Neural Rendering?
ANR is a three-part neural rendering suite built by Apex Compute. It covers the same three jobs as NVIDIA’s DLSS feature set:
- Frame Generation (FG): creates extra frames to smooth motion. Available now in beta.
- Super Resolution (SR): upscales a lower resolution render to a sharper image. Still in development.
- Ray Reconstruction (RR): cleans up ray traced lighting with a neural model. Still in development.
Only frame generation is ready to try today. The super resolution and ray reconstruction models are still in training, according to Apex Compute, so the full suite is not here yet. Even so, a working frame generation beta that runs across vendors is a big deal on its own.
The Performance Claims: Faster Frames, Less VRAM
Apex Compute published internal benchmarks to back up its claims, and the numbers look strong on paper. All figures below come from Apex Compute’s own testing, not independent reviews, so treat them as the developer’s claims until third parties verify them.
On a GeForce RTX 5090 running at 4K, Apex Compute reports the following frame generation times:
| Frame Generation Mode | Apex Neural Rendering | NVIDIA DLSS Frame Generation |
|---|---|---|
| 2X | 1.01 ms per generated frame | 1.26 ms per generated frame |
| 4X | 0.61 ms per generated frame | 0.82 ms per generated frame |
Figures: Apex Compute internal benchmarks on GeForce RTX 5090 at 3840×2160, as reported by Wccftech.
Two things matter here. First, these numbers measure the cost of generating each extra frame, not a straight FPS gain in a game. Faster generation time is good news, but it is not the same as a 2.4x higher frame rate. Second, the headline up to 2.4x faster figure comes from Apex Compute’s own tests in certain scenarios, while the published RTX 5090 numbers show a smaller gap. Wccftech flagged the same point in its coverage.
VRAM is the other headline claim. Apex Compute says ANR frame generation uses 214 MB of video memory at 4K, against 562 MB for NVIDIA DLSS frame generation, a reduction of about 62 percent. The neural network weights are tiny too, at just 0.17 MB. For players on cards with limited VRAM, that difference could matter as much as speed.
Image quality is the third claim. Apex Compute’s internal quality measurements favor ANR over DLSS frame generation across CGVQM-5, ColorVideoVDP, and PSNR metrics, from 2X to 6X frame generation. The company also published side-by-side comparisons covering particles, moving objects, water, vegetation, and disocclusion artifacts, plus video showcases in The Last of Us Part II Remastered and Hogwarts Legacy.
DLSS itself has been under scrutiny lately, after the DLSS 5 October update drew complaints about shadow and reflection quality. That makes the arrival of a rival worth watching even more closely.
How frame generation actually works, in plain terms
Frame generation does not make your GPU render faster. It renders real frames at the normal rate, then uses a neural model to invent extra frames in between them. In 2X mode, one generated frame sits between every two real frames. In 4X mode, three generated frames fill the gap. The screen shows more frames per second, and motion looks smoother, but the generated frames are predictions, not renders.
That is why the generation time in the table matters. Every extra frame costs GPU time, and the cheaper each generated frame is, the more of the GPU’s budget is left for the real frames. ANR’s claimed 1.01 ms per frame at 2X, against DLSS’s 1.26 ms, is a real saving if it holds up, but it is a saving of a fraction of a millisecond per frame. It is not the same as a 2.4x jump in frame rate, and the headline figure should be read with that in mind.
There is also a trade-off no frame generation tool can escape: latency. A generated frame is built from frames that already exist, so what you see is always slightly behind your input. The smoother the picture, the more the game can feel delayed, especially in fast competitive games. That is true of DLSS, FSR, and ANR alike. It is the reason frame generation shines in slow, cinematic single-player games and matters much less in shooters where every millisecond counts.
Which GPUs Can Run Apex Neural Rendering?
This is where ANR stands out. DLSS frame generation is locked to NVIDIA RTX cards, but ANR is built to run on nearly every modern GPU from the last five years:
- NVIDIA: GeForce GTX 16 series and RTX 20, 30, 40, and 50 series
- AMD: Radeon RX 5000, 6000, 7000, and 9000 series, plus Steam Deck
- Intel: Arc A series and B series
- Apple silicon: M1 and later, through a native Metal test kit for macOS 26 and newer
The beta supports Windows 10 and 11, plus Linux and SteamOS through Proton. That Steam Deck support is a notable point: handheld players could get frame generation in games that never offered it on the Deck.
How To Try The Beta
The frame generation beta is free to download from the Apex Compute GitHub page. Installation runs through OptiScaler, the open source upscaling tool, and the process is straightforward:
- Download the latest ANR frame generation beta from GitHub.
- Pick the package that matches your game: one for games that already have OptiScaler installed, and a full package for games that do not.
- Extract the files and run the installer against the game’s executable directory.
- Enable multi frame generation in the game’s graphics settings.
The latest release, FG V.B14, arrived on November 12, 2026 and fixes a Windows compatibility issue that stopped some games from launching with older Visual C++ runtimes.
Apex Compute also released a utility called nr-record. It captures frame generation inputs so players can replay them through competing implementations and compare image quality directly. That is a useful tool for anyone who wants to test the claims rather than take them on trust.
ANR is not the only project bringing DLSS class features to more hardware. AMDLSS 5.2.0 recently brought faster DLSS 5 neural rendering to Radeon GPUs, and the open tools scene keeps growing.
What to watch before you trust the claims
The claims are bold, and the honest way to judge them is to wait for independent testing. Here is what to keep an eye on before treating any of the numbers as fact.
- Only frame generation is out. Super resolution and ray reconstruction are still in training, so ANR is not a full DLSS replacement yet. Judge the beta on what it actually does today.
- The code is not open yet. ANR is described as open source, but the frame generation code has not been released. Apex Compute says it will come later. Until then, nobody outside the company can inspect how it works.
- Test it in one game first. The beta is experimental, so expect artifacts and compatibility quirks. Try it in a slow single-player game, where frame generation helps most, and compare it with frame generation off before judging the image quality.
- Use the nr-record tool. It lets you replay the same scene through different frame generation implementations and compare them directly. That is a better test than any benchmark chart.
- Watch for independent benchmarks. The real verdict will come from third-party testers running ANR across many games and GPUs, not from the developer’s own charts.
Why This Matters For PC Gamers
Frame generation has been one of the most vendor-locked features in PC gaming. NVIDIA DLSS multi frame generation needs an RTX 50 series card. AMD FSR frame generation works more widely, but image quality comparisons still favor NVIDIA in many titles. A cross-vendor neural option that runs on a GTX 1660, an RX 5700, or a Steam Deck would change who gets to use the feature.
There are good reasons to stay cautious. The frame generation code is not open source yet, even though the project is described as open source. Apex Compute says it will release the source code later. The beta is experimental, so compatibility issues and visual artifacts should be expected. And every performance number so far comes from the developer’s own benchmarks. Independent testing is still needed, and the real test will be how ANR looks in motion across many games, not in cherry-picked comparisons.
Still, the timing is interesting. Open tools for PC graphics are having a moment: GPU Taskman recently gave players a free way to watch VRAM use in real time, and shader compilation tools are tackling stutter directly. ANR fits the same pattern, a community-facing tool that works across hardware instead of locking players to one brand.
If independent tests confirm even part of the claims, ANR could push NVIDIA, AMD, and Intel all to move faster on open frame generation. For now, the beta is free, the install is simple, and the claims are bold enough to be worth testing yourself.
News Source: Wccftech, with additional reporting from The FPS Review. All performance figures are Apex Compute’s internal benchmarks, as reported by those outlets.
