DLSS Preset C vs Preset B: Should You Switch Frame Generation? Tests, HUD Issues and How to Try It (2026)
NVIDIA's new DLSS Frame Generation Preset C: what it changes, independent FPS tests on RTX 40/50, UI blur caveats, who should switch, official enable steps and how to roll back.
Short answer: if your game already runs at 60 FPS or more before frame generation, try Preset C. In the independent tests we found it gave a small to moderate FPS gain (roughly 3–28%, depending on the rig and game). If your base frame rate is under about 40 FPS, or you care about a clean HUD, minimap and quest text, stay on Preset B. Preset C is still a beta in the NVIDIA app, and you can switch it off in seconds.
This guide covers one question: should you switch DLSS Frame Generation from Preset B to the new Preset C, and how do you test it fairly? For the bigger October news (the faster DLSS 5 model, Hogwarts Legacy and RTX Remix dates), see our DLSS 5 October update explainer.
What Preset C is (and what NVIDIA actually said)
On 13 October 2026 NVIDIA released a beta of a new Performance Mode for Frame Generation, labelled Preset C in the NVIDIA app. According to NVIDIA's announcement:
- Frame Generation and Multi Frame Generation use two AI models. One uses motion vectors and depth to work out object motion between two rendered frames. The second calculates optical flow for things like UI and ray-traced effects.
- Preset C replaces the second model with a "faster and lighter algorithm", with "minor trade-offs in image quality".
- It covers Frame Generation, Multi Frame Generation and Dynamic Multi Frame Generation on GeForce RTX 40 and RTX 50 GPUs and RTX Spark laptops.
- NVIDIA's headline figure is "up to 35% faster" on a GeForce RTX 5060.
Club386 reported NVIDIA's breakdown of that figure: the cost of frame generation drops by 10% in 2x mode, 23.8% in 4x and 36.4% in 6x, measured on an RTX 5060 8GB in Cyberpunk 2077 at 1440p. Note the wording. That is a cut in the cost of frame generation, not a 35% jump in your total FPS. NVIDIA did not say whether RTX 40 cards see the same gain.
What independent testers found so far
These are early, mostly single-rig results from YouTube-based testers and small outlets. Treat them as indications, not verdicts.
| Source | Hardware / game | Preset B | Preset C |
|---|---|---|---|
| PC Central (citing the Gaming Optimized channel; RTX 4060, driver 617.14) | Resident Evil Requiem | 86 FPS | 99 FPS |
| same | Gears of War (opening scene) | 101 FPS | 129 FPS |
| same | Cyberpunk 2077 (streets) | 85 FPS | 93 FPS |
| GameGPU (MxBenchmarkPC; RTX 5070, Cyberpunk, 1440p, FG 2x) | Scene 1 | 95 FPS, 49 ms | 98 FPS, 47 ms |
| same | Scene 2 | 100 FPS, 49 ms | 106 FPS, 47 ms |
| YouTube tester (1440p, ray tracing, DLSS Balanced, FG 2x) | Cyberpunk / Doom | 67 / 107 FPS | 75 / 115 FPS |
Patterns across these tests:
- FPS gains are real but uneven: from about 3% in a light scene to about 28% in the heaviest cases (PC Central's headline "up to 28%" on an RTX 4060). GameGPU's tester saw gains of up to about 20% at 4x/6x under maximum load.
- Latency dropped slightly (about 2 ms in two separate tests), which is what you would expect when the extra FPS comes from a cheaper frame-gen step. One tester said NVIDIA's files show no change to the latency code itself, so any latency gain comes from the higher frame rate.
- Image-quality downsides are concentrated on the UI. PC Central reports blurring and ghosting on UI elements when the base frame rate falls below about 40 FPS, and a possible black-screen bug in Cyberpunk 2077. One YouTube tester described faded doubles of world icons and smearing behind quest text at 4x. This tester also said, based on his own analysis of NVIDIA's files, that Preset C cannot run alongside the part that draws the HUD separately. We could not verify that file-level claim, and NVIDIA has not confirmed it, so read it as a hypothesis for why the smearing happens.
Who should try Preset C
| Your situation | Suggestion |
|---|---|
| RTX 4060/5060-class GPU, base FPS 60+ with frame gen on | Try it. This is where gains look most useful |
| You use 4x or 6x Multi Frame Generation | Try it. The cost reduction is largest here per NVIDIA's own numbers |
| Base FPS below ~40 before frame generation | Skip it. UI ghosting reportedly gets worse, and frame generation is already a poor fit at low base FPS |
| Heavy HUD or text games (strategy, RPG quest logs, minimaps) | Test carefully. Keep Preset B if you see smearing |
| Competitive shooters where latency matters | Optional. Latency gains were small (about 2 ms in tests), so don't expect miracles |
| You want zero risk | Wait for the final release. It is a beta |
How to switch to Preset C (official steps)
From NVIDIA's announcement:
- Install the newest NVIDIA app update.
- Open Settings → About and tick "Opt in to access Beta or experimental features".
- Wait for the beta to download and click the on-screen message to install it.
- Open the Graphics section, choose a supported Frame Generation, Multi Frame Generation or Dynamic Frame Generation game. Or choose Global Settings at the top to apply it to every supported game.
- Scroll to Driver Settings and click DLSS Override – Model Presets.
- Click Custom.
- Open the Frame Generation drop-down and choose Preset C. If you don't see it, reboot.
- Click Apply.
Tip: apply it per game first, not globally. A global override forces the preset on every supported game, including ones where it may behave badly.
A fair 5-minute A/B test
- Pick one repeatable scene: a camera pan through a busy street, or a built-in benchmark.
- Keep everything else fixed: same resolution, DLSS Super Resolution mode, frame-gen multiplier and driver.
- Record average FPS and 1% lows with the NVIDIA overlay or your usual tool, once on Preset B and once on Preset C.
- Look at the UI on purpose. Turn the camera quickly and watch the minimap, quest markers and subtitles for smearing or doubled icons.
- Check stability: play 10–15 minutes, because the one reported black-screen issue was in Cyberpunk 2077.
- If the gain is under about 5% or the HUD looks worse, go back to Preset B.
How to roll back
Return to the same menu (Driver Settings → DLSS Override – Model Presets → Custom → Frame Generation), pick the previous preset (B) or reset the game to default, and click Apply. If a game misbehaves, restart it after changing the preset.
What this does not do
- It does not add DLSS 5 to RTX 40 cards. RTX 40 owners get Preset C for frame generation, but DLSS 5 neural rendering is listed for RTX 50, GeForce NOW and RTX Spark.
- It does not change image sharpness in the world itself. The change is aimed at the frame-generation step.
- It is not proven to help every game. NVIDIA's own claim is "up to".
- Do not rely on DLL swapping to force this. If you are weighing third-party tools, see DLSS Swapper vs DLSS Manager vs OptiScaler, and use the official NVIDIA app route for this feature.
Caveats
- NVIDIA's 35% figure is "up to", measured on one GPU (RTX 5060) in one game at 1440p.
- Independent numbers here come from a handful of testers on single rigs, and some are YouTube-based, so results will vary with your CPU, driver and game version.
- Preset C is a beta, and details can change before a final release.
Sources: NVIDIA GeForce news (13 Oct 2026), TechPowerUp (13 Oct), Club386 (14 Oct), GameGPU (13 Oct), PC Central (17 Oct) and a YouTube tester's video, all checked 18 Oct 2026.
← Back to all posts