Apex Compute's free ANR frame generation beta claims to beat DLSS Frame Generation on speed and image quality, and it works on GTX 16, RTX 20 and 30, Radeon, Intel Arc and even Steam Deck. The headline numbers are the company's own, but you can download it and test it tonight.
If you are still gaming on an RTX 3080, a GTX 1660 or a Radeon RX 6700 XT, Nvidia has spent the last few years telling you its AI frame generation is not for you. A small company called Apex Compute just said otherwise. Its new Apex Neural Rendering frame generation, ANR FG for short, is out now as a free public beta, and it runs on GPUs from all three PC graphics makers, including plenty of cards Nvidia’s own DLSS Frame Generation will never touch.
Apex announced the full ANR suite on November 11 with a bold pitch: “cross-platform Super Resolution, Ray Reconstruction, and Frame Generation technology, delivering up to 2.4× faster performance and higher quality rendering than NVIDIA DLSS.” Only the frame generation piece is available today. The company says its upscaling and ray reconstruction models “are currently in training,” with both listed as “Beta soon” on its GitHub page.
Introducing Apex Neural Rendering (ANR), our cross-platform Super Resolution, Ray Reconstruction, and Frame Generation technology, delivering up to 2.4× faster performance and higher quality rendering than NVIDIA DLSS.
, Apex Compute (@apexcompute) November 11, 2026
ANR rebuilds neural rendering around a new architecture that… pic.twitter.com/DcSkwrouKS
Who can run it
This is the part that has PC players lit up. Apex’s supported list covers:
- Nvidia: GeForce RTX 20, 30, 40 and 50 series, plus GTX 16 series
- AMD: Radeon RX 5000, 6000, 7000 and 9000 series, plus Steam Deck
- Intel: Arc A-series and B-series
- Apple: M1 and later, on macOS 26 or later
It works on Windows 10 and 11, and on Linux and SteamOS through Proton. For context, Nvidia’s DLSS Frame Generation has always been locked to RTX 40 series cards and newer, which leaves every RTX 20 and RTX 30 owner on the outside looking in, even though those cards have the AI hardware for upscaling. AMD’s FSR frame generation already runs across brands, so cross-vendor frame gen on its own isn’t new. What Apex is selling is a neural model that it says looks better than Nvidia’s and runs faster, on cards that Nvidia’s version skips. If that holds up, it’s a big deal for anyone squeezing more life out of an older GPU.
The numbers, and whose numbers they are
Every benchmark here comes from Apex itself, and that matters. On its ANR page, Apex says an RTX 5090 at 4K takes 1.01 ms to generate a frame in 2X mode with ANR, against 1.26 ms for DLSS Frame Generation. In 4X mode the averages are 0.61 ms versus 0.82 ms. ANR also uses 214 MB of GPU memory at 4K compared with 562 MB for DLSS Frame Generation, and Apex lists the model’s network weights at just 0.17 MB.
Notice something: those millisecond figures put ANR about 20% faster per generated frame, not 2.4 times faster. The “2.4×” headline comes from a separate Apex benchmark chart, which VideoCardz points out lists 0.34 for ANR against 0.81 for DLSS in 2X mode on the RTX 5090 without saying what unit it measures. That same chart shows ANR beating FSR 3.1 on the RX 7800 XT, Arc B580 and Arc A770, but on the RX 9060 XT the picture is mixed, with FSR 3.1 slightly faster at 2X and AMD’s FSR 4 machine learning frame gen ahead at 3X and 4X. So the honest read is this: Apex’s own clearly labeled numbers show a real but smaller speed edge over DLSS, plus a big memory saving, and the 2.4× claim needs independent testing before anyone runs with it.
On image quality, Apex publishes three objective metrics, CGVQM-5, ColorVideoVDP and PSNR, and ANR comes out ahead of DLSS Frame Generation in every mode from 2X through 6X. In CGVQM-5, for example, Apex scores ANR 96.59 against 93.66 for DLSS at 2X, and the gap widens to 93.50 versus 87.82 at 6X. The company also shows drag-to-compare shots from The Last of Us Part II Remastered and test scenes with particles, smoke, grass and a spinning propeller, which are exactly the spots where frame generation usually smears.
Our take: the memory number is the one we’d watch. Older cards like the RTX 3060 Ti and RTX 3070 shipped with 8 GB of VRAM, and saving roughly a third of a gigabyte at 4K is the difference between a stable game and a stuttering one on those cards. Speed bragging rights against an RTX 5090 are nice. Making a five-year-old GPU feel new again is the real story.
Apex's ANR brings AI frame gen to older GPUs. Are you trying it?
Vote and see what other Xbox fans think.
Like having a say? Get OnPoint Weekly or argue it out on Discord.
How to try it, and the fine print
Don’t expect a toggle in your driver panel. ANR FG installs through OptiScaler, the popular modding tool that swaps upscalers and frame generation inside games. The GitHub release offers two packages: a full one for games without OptiScaler and an add-on for games that already have it. You close the game, extract the package and run the installer pointed at the folder holding the game’s .exe (install.cmd on Windows, install.sh on Linux and Steam Deck). On Proton you also add a launch option the installer prints for you. Then, in the game, you turn on frame generation, choosing the Nvidia option when the game offers more than one. Uninstalling is one script.
That means it works game by game, and how well it works will vary from game to game. Apex has been shipping fixes fast. The beta is already on build V.B14. Recent builds fixed a “DirectX 12 is not supported” error on PCs with older Visual C++ runtimes, added support for Vulkan games, shipped follow-up fixes for titles like No Man’s Sky and DOOM: The Dark Ages, and sped up Intel Arc B-series cards by using their XMX units. Apex also says ANR’s source code is “coming soon.” It isn’t public yet, though OptiScaler itself is open source under GPL-3.0.
The timing is wild, too. Apex dropped this in the same week Nvidia’s board is weighing the future of CEO Jensen Huang and Nvidia stock is sliding. Nobody needed a new DLSS rival on Nvidia’s to-do list this week, but PC gamers won’t complain. Competition is how older hardware gets support, and if independent testers back up even half of what Apex is claiming, Nvidia will have to answer the obvious question: why couldn’t RTX 30 owners get this from the company that sold them the card?
If you try it on your rig, tell us your GPU and what you saw. We want to hear it.
Image: Apex Compute
Sources
- Apex Compute on X (November 11, 2026): Introducing Apex Neural Rendering (ANR)
- Apex Compute: ANR: Apex Neural Rendering
- GitHub: apex-compute/ANR
- GitHub: ANR FG releases
- VideoCardz (November 12, 2026): Gamers now have an open-source DLSS Frame Generation alternative, Apex claims faster performance and better quality
- TechPowerUp (November 12, 2026): Apex Neural Rendering Beats NVIDIA DLSS in Both Image Quality and Performance
- Notebookcheck: New Nvidia DLSS rival promises faster frame generation on older AMD, Intel and GeForce GPUs
