← Tech
💻Tech

Apex Neural Rendering (ANR) Frame Generation Beta: Install Guide and How It Compares to DLSS FG

ANR FG's public beta, an OptiScaler injector for GTX 16, RX 5000, Arc and Steam Deck. Apex's claimed 1.01 ms vs 1.26 ms and 214 MB vs 562 MB, labeled as manufacturer figures, plus the official install steps.

T Terapep ·

TL;DR: On Nov 11–12, 2026, Apex Compute published a public beta of Apex Neural Rendering Frame Generation (ANR FG) on GitHub. It is an OptiScaler-based injector for games that already offer NVIDIA DLSS Frame Generation, and the company lists support for GTX 16, RX 5000, Intel Arc, the Steam Deck, and newer GPUs. Apex's own figures, also reported by TechPowerUp and Notebookcheck on Nov 12, put 2X generation at 1.01 ms vs 1.26 ms and memory use at 214 MB vs 562 MB on an RTX 5090 at 4K. Those are manufacturer claims, not independent tests. ANR Super Resolution and Ray Reconstruction are not in this beta.

What shipped

Apex Neural Rendering is a suite announced by Apex Compute with three pieces:

  • ANR FG — frame generation. In public beta now.
  • ANR SR — super resolution (upscaling). The project page and GitHub README say "beta soon."
  • ANR RR — ray reconstruction (denoising for ray tracing). Also "beta soon."

The ANR source code is listed as "coming soon." Each release does include OptiScaler's source, under GPL-3.0.

The public GitHub repository was created on Nov 11, 2026. The latest package we checked is FG V.B14, published Nov 12, 2026, 15:04 KST (06:04 UTC). That release notes a Windows fix: with the full package, some games failed to start and reported that DirectX 12 was not supported on systems with an older Microsoft Visual C++ runtime. V.B14's full package no longer depends on msvcp140_atomic_wait.dll, a file that only newer runtimes include.

Manufacturer claims vs DLSS Frame Generation

Apex publishes the following first-party numbers on its ANR page, for an NVIDIA GeForce RTX 5090 at 3840 × 2160. TechPowerUp and Notebookcheck repeated the 2X time and memory figures on Nov 12. Notebookcheck also reported the 4X times. We have not run our own tests, and neither outlet presented these as independent benches. TechPowerUp also noted that it was not sure which DLSS Frame Generation version was used.

Bar charts of Apex Compute's claimed ANR Frame Generation figures versus NVIDIA DLSS Frame Generation on a GeForce RTX 5090 at 4K: 1.01 ms versus 1.26 ms at 2X, 0.61 ms versus 0.82 ms at 4X, and 214 MB versus 562 MB of GPU memory. These are first-party measurements, not independently verified. Sources: Apex Compute, as reported by TechPowerUp and Notebookcheck, Nov 12, 2026.

Measurement (RTX 5090, 4K) ANR FG (claimed) DLSS FG (claimed)
2X, per generated frame 1.01 ms 1.26 ms
4X, mean per generated frame 0.61 ms 0.82 ms
GPU memory at 4K 214 MB 562 MB
Network weights 0.17 MB —

The same page also lists first-party image-quality scores (CGVQM-5, ColorVideoVDP, PSNR) in which Apex claims an advantage for ANR FG from 2X through 6X. Those scores are likewise manufacturer figures. Notebookcheck cautioned that faster frame generation alone does not necessarily mean better overall performance or lower input latency, and that independent testing is needed.

Apex also published Apple silicon times for a native Metal path on an M2 Max, with no DLSS comparison on that table:

Measurement (Apple M2 Max, claimed) 1920 × 1080 3840 × 2160
2X, per generated frame 4.26 ms 15.66 ms
4X, mean per generated frame 2.52 ms 9.25 ms

Supported GPUs and operating systems

From Apex's site and the FG V.B14 release notes:

Vendor Listed support
NVIDIA GeForce RTX 20 series and newer, and GTX 16 series
AMD Radeon RX 5000 series and newer, and Steam Deck
Intel Arc A-series and B-series
Apple Silicon, M1 and later

Operating systems listed on the same pages: Windows 10 and 11, Linux and SteamOS with Proton, and macOS 26 or later. The main GitHub README currently lists the Windows and Linux/SteamOS set under "Supported GPUs" and does not mention Apple there; the official site and the V.B14 notes do.

ANR FG is documented to work in games that already offer NVIDIA DLSS Frame Generation. Compatibility will vary by title. This is a beta that writes into the game's folder.

How to install the beta

Download only from the official apex-compute/ANR Releases page. As of V.B14 the two game packages are:

  • anr-fg-full — for games that do not already use OptiScaler
  • anr-fg-addon — for games that already use OptiScaler

Official steps, from the README and the V.B14 notes:

  1. Close the game. Extract the package and run the installer, passing the folder that contains the game's .exe.
    • Windows: install.cmd. The V.B14 notes say this needs the Microsoft Visual C++ 2015–2022 Redistributable (x64).
    • Linux and Steam Deck (Proton): sh install.sh with that same folder, then set the game's launch options to the line the installer prints. The docs give WINEDLLOVERRIDES="dxgi=n,b" %command%, or winmm=n,b in Vulkan games.
  2. In the game, turn on frame generation — the NVIDIA option, if the game offers more than one.

To uninstall, run uninstall.cmd or sh uninstall.sh with the same folder.

The release also includes Windows, Linux, and macOS test kits (TESTERS.md inside) and an optional recorder, nr-record, for comparing generators. Those are not required for a normal install.

Before you start

  • This is a beta. Expect broken titles, visual issues, and changing packages.
  • It is built around OptiScaler, which injects into the game directory. Do not use it in online or anti-cheat games. See our OptiScaler / DLSS Swapper anti-cheat guide.
  • Download only from the official GitHub releases. Treat random mirrors as untrusted.
  • Keep a copy of the game folder or be ready to use the official uninstaller and, if needed, a storefront verify (Steam, Epic, or similar).

Related

Sources

#anr#apex-compute#frame-generation#optiscaler#dlss#gtx-16#steam-deck#how-to#2026
← Back to all posts