ShiftDelete.Net reports that Apex announced the suite on November 11, 2026. Its broad hardware ambitions make this worth watching, but the performance headline needs careful handling: Apex’s own published timing table shows smaller advantages in its disclosed comparisons, not a blanket 2.4× improvement.
What the ANR beta actually delivers
Apex describes three components: ANR FG for frame generation, ANR SR for upscaling, and ANR RR for ray-tracing denoising. Its repository lists the latter two as forthcoming and says ANR source code is still coming. A public GitHub repository, in other words, does not establish that the rendering implementation is already open source.
The documented PC beta supports:
- NVIDIA: GeForce RTX 20, 30, 40 and 50 series, plus GTX 16.
- AMD: Radeon RX 5000, 6000, 7000 and 9000 series, plus Steam Deck.
- Intel: Arc A-series and B-series.
The README lists Windows 10 and 11, Linux, and SteamOS through Proton. These are vendor-stated compatibility boundaries—not proof that every game works.
Apple support needs a separate qualification. Apex’s product page lists M1 and newer Apple silicon running macOS 26 or later. Its V.B13 release notes, dated November 12, describe an Apple-silicon test kit comparing ANR’s Metal implementation with Apple MetalFX frame interpolation on identical frames. That should not be mistaken for the Windows game-installation package arriving unchanged on macOS.
The installation path is through OptiScaler
This is a third-party beta, not an announcement of native publisher integration. Apex provides two packages: anr-fg-full for games without OptiScaler and anr-fg-addon for games already using it. The documented activation path uses the game’s existing NVIDIA frame-generation option.
For Windows testing, Apex’s release instructions specify:
- Close the game and extract the appropriate package.
- Ensure the Microsoft Visual C++ 2015–2022 Redistributable, x64 prerequisite is installed.
- Run
install.cmd, supplying the folder containing the game executable. - Launch the game and enable its NVIDIA frame-generation option where offered.
- To remove the installation, run
uninstall.cmdagainst the same folder.
The practical takeaway is simple: a supported GPU is only one requirement. The documented method also depends on the game exposing a suitable frame-generation path.
Why “2.4× faster” is not a gaming-FPS promise
Apex’s product page reports these measurements on a GeForce RTX 5090 at 3840 × 2160:
| Measurement | ANR FG | DLSS FG |
|---|---|---|
| 2X, time per generated frame | 1.01 ms | 1.26 ms |
| 4X, mean time per generated frame | 0.61 ms | 0.82 ms |
| GPU memory | 214 MB | 562 MB |
The company also lists network weights of 0.17 MB. All are Apex-published figures.
Dividing the disclosed timings gives approximately 1.25× at 2X and 1.34× at 4X. Those ratios describe this measured operation, not whole-game frame rates. Treating either as an equivalent FPS increase would go beyond the evidence.
ShiftDelete.Net also describes a separate company social-media chart whose 2X values—0.34 against 0.81—produce roughly the advertised 2.4× ratio. However, the chart lacks a unit label. Without that and a clear methodology, it cannot reconcile the headline with the website’s different measurements.
Apex’s quality table likewise favors ANR: its 2X results include 96.59 versus 93.66 for CGVQM-5, 9.24 versus 8.77 for ColorVideoVDP, and 33.84 versus 31.22 dB for PSNR. These remain vendor results, even with published statistical intervals.
The most useful evidence is hardware-specific
Beyond the headline, Apex’s V.B12 notes report an Intel Arc B-series optimization using XMX matrix units. On its B580 test, 4K 2X generation fell from 6.88 ms to 5.99 ms; Arc A-series was unchanged. That is a more actionable distinction than assuming identical benefits across supported GPU families.
For troubleshooting, Apex asks testers to provide the game, GPU, driver, operating system, settings and newest nr_fg-*.log from the game’s OptiScaler/custom/Apex Neural Rendering FG folder.
ANR therefore merits attention as an experimental cross-vendor option—not yet a demonstrated replacement for the complete DLSS suite. The sensible test is whether it improves a particular game on a particular PC without introducing unacceptable visual or stability problems. The benchmark headline can wait its turn.
References
- Apex Neural Rendering Arrives as a Cross Platform DLSS Rival - ShiftDelete.Net ShiftDelete.Net · 2026-11-12T04:43:14+00:00
- GitHub - apex-compute/ANR: Apex Neural Rendering (ANR) by Apex Compute · GitHub github.com
- Releases · apex-compute/ANR · GitHub github.com