Three colorful graphics cards connect to a monitor displaying a futuristic spaceship racing through a city.
Apex Neural Rendering has entered the PC graphics arena with a cross-vendor frame-generation beta and an ambitious claim: performance up to 2.4 times faster than NVIDIA DLSS. But the useful distinction is between what gamers can install and what Apex Compute still promises. Frame generation is in beta; super resolution and ray reconstruction are not yet available as beta features, according to the company’s GitHub documentation.

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.

Three colorful graphics cards connect to a monitor displaying a futuristic spaceship racing through a city. 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:

  1. Close the game and extract the appropriate package.
  2. Ensure the Microsoft Visual C++ 2015–2022 Redistributable, x64 prerequisite is installed.
  3. Run install.cmd, supplying the folder containing the game executable.
  4. Launch the game and enable its NVIDIA frame-generation option where offered.
  5. To remove the installation, run uninstall.cmd against 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:

MeasurementANR FGDLSS FG
2X, time per generated frame1.01 ms1.26 ms
4X, mean time per generated frame0.61 ms0.82 ms
GPU memory214 MB562 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

  1. Apex Neural Rendering Arrives as a Cross Platform DLSS Rival - ShiftDelete.Net ShiftDelete.Net 2026-11-12T04:43:14+00:00
  2. GitHub - apex-compute/ANR: Apex Neural Rendering (ANR) by Apex Compute · GitHub github.com
  3. Releases · apex-compute/ANR · GitHub github.com