GeForce RTX 4070 DLSS 4.5 vs DLSS 5: what desktop owners should know
The desktop GeForce RTX 4070 fully supports the current DLSS 4.5 stack: Super Resolution, Ray Reconstruction, Frame Generation, and the companion NVIDIA Reflex path. DLSS 5 is different. NVIDIA hasn’t announced it for this SKU, so developers and owners should treat future support as unknown rather than build a plan around rumor.
Published 2560×1440 results provide a clear baseline for what the card does today. As of mid-2026, there is no official DLSS 5 driver, SDK, or game integration targeting this GPU. That leaves the next twelve months as a normal DLSS 4.5 production cycle unless NVIDIA changes the support record.
The RTX 4070’s place in the Ada desktop range
The GeForce RTX 4070 is a mid-range Ada Lovelace desktop card aimed at 1440p. It uses the AD104 die with 12 GB of GDDR6X on a 192-bit bus. Its feature set includes hardware ray tracing, Tensor cores for AI inference, AV1 encoding, DLSS upscaling, and frame generation.
The DLSS implementation is shared with the wider RTX 40 family rather than split into a special branch for this model. The API is the same as on larger Ada cards; the performance budget is what changes. If an RTX 40 DLSS feature runs on an RTX 4090, it runs through the same feature path on the RTX 4070, subject to the card’s lower throughput. That makes the current comparison mostly about NVIDIA’s release timing and support decision.
What DLSS 4.5 runs on this card
Super Resolution reconstructs the output from a lower-resolution internal frame. Ray Reconstruction replaces multiple hand-tuned denoisers for ray-traced lighting, reflections, and ambient occlusion with one neural model. Frame Generation inserts an AI-generated frame between two rendered frames. NVIDIA Reflex then reduces render-queue latency to offset the extra delay introduced by Frame Generation.
A game can expose the first three as independent settings. Each has a measurable cost. Super Resolution uses a small amount of GPU time for neural reconstruction. Ray Reconstruction adds a post-process pass while replacing other denoisers. Frame Generation is the heaviest because it creates an entire intermediate frame and requires Reflex for a well-behaved latency path.
| DLSS 4.5 component | What it does | Where it runs on the RTX 4070 | Main engineering cost |
|---|---|---|---|
| Super Resolution | Upscales a lower internal render to the target output resolution | Tensor cores, parallel with main render | Small per-frame GPU time, image quality tuning |
| Ray Reconstruction | Replaces traditional denoisers with a single neural denoiser | Tensor cores, post-process pass | Replaces multiple denoiser passes, integration work |
| Frame Generation | Inserts an AI-generated intermediate frame between rendered frames | Tensor cores, optical flow accelerator | Extra frame cost, base-frame latency, requires Reflex |
| NVIDIA Reflex | Reduces render queue latency to compensate for Frame Generation | Driver and SDK | Small CPU overhead, requires integration |
Verified 1440p performance under DLSS 4.5
The Notebookcheck desktop GPU page reports three results for this exact card. They are presented at the recorded resolutions and settings without interpolation:
- Cyberpunk 2077 1.6 at 2560×1440, native QHD settings, average 80.3 FPS.
- Cyberpunk 2077 at 2560×1440, QHD with DLSS active, average 83.9 FPS.
- Alan Wake 2 at 2560×1440, native QHD settings, average 68.7 FPS.
Cyberpunk 2077 moves from 80.3 FPS at native QHD to 83.9 FPS with DLSS at the same output resolution. The gain is modest, but it is the exact delta recorded for this build and card. Alan Wake 2 at 68.7 FPS supplies a native reference for one of the heavier RT-focused releases.
Drivers, patches, settings, and scene selection can move all three averages. The GeForce RTX 40 series reference provides hardware context, while the table below keeps the measured result tied to the desktop RTX 4070.
| Game | Resolution | Settings profile | Average FPS | Provenance |
|---|---|---|---|---|
| Cyberpunk 2077 1.6 | 2560×1440 | Native QHD | 80.3 | Notebookcheck direct measurement, desktop RTX 4070 |
| Cyberpunk 2077 | 2560×1440 | QHD with DLSS | 83.9 | Notebookcheck direct measurement, desktop RTX 4070 |
| Alan Wake 2 | 2560×1440 | Native QHD | 68.7 | Notebookcheck direct measurement, desktop RTX 4070 |
Frame Generation needs Reflex and clean pacing
Frame Generation uses the Ada optical-flow accelerator and a small Tensor-core network to interpolate motion between the previous and current rendered frames. Its quality depends heavily on the motion vectors exported by the engine, which is why results vary from one integration to another.
In scenes with limited motion, the visible frame rate can roughly double because an extra frame appears between each rendered pair. That generated frame still trails the newest input, and the rendered frame spends longer in the queue. Reflex reduces the queue so input-to-photon latency stays closer to the no-Frame-Generation baseline. Shipping Frame Generation without Reflex is therefore a correctness problem on this card, not a cosmetic preference.
- Enable Reflex whenever Frame Generation is enabled, so the user does not see a regression in input responsiveness.
- Export high-quality motion vectors from the engine, because Frame Generation quality scales with motion vector accuracy.
- Avoid stacking Frame Generation on top of variable refresh rate displays that operate below the generated frame rate, because the visual cadence becomes uneven.
- Profile the post-process cost of Ray Reconstruction, because it runs after the main render and before Frame Generation.
No official DLSS 5 path exists in mid-2026
NVIDIA has not announced DLSS 5 support for the desktop RTX 4070. There is no official DLSS 5 driver, SDK, game integration, release date, feature list, or measured uplift for this card. The same uncertainty applies to the wider RTX 40 desktop family.
What is known is enough for a current plan: DLSS 4.5 works under the present Game Ready driver and SDK, including Frame Generation paired with Reflex. Whether NVIDIA extends DLSS 5 to RTX 40 desktop, when it might happen, and which features it might include are all open questions.
Prepare the engine without hard-coding a version
Drive the graphics menu from SDK capability queries. If a future SDK exposes a feature on the installed card, the menu can show it. If not, the option should remain absent. A hard-coded version string is how unsupported controls reach players.
Keep the motion-vector, depth, and alpha exports clean, even if the current path doesn’t consume every channel. Later DLSS models have tended to ask more of per-pixel motion and transparency data, and those buffers still consume the RTX 4070’s existing bandwidth. Reserve a modest post-process allowance for another neural pass, but don’t assign a specific cost before the SDK exists.
- Drive the in-game DLSS menu from the SDK, not from a hard-coded version string, so a new SDK release updates the menu automatically.
- Keep motion vectors, depth, and alpha exports clean, so a future DLSS version can consume them without engine rework.
- Reserve a small amount of post-process budget for an additional neural pass, because each DLSS generation tends to add a new pass on top of the previous one.
- Test Frame Generation with Reflex on every supported title, because the latency interaction is a correctness concern, not a polish concern.
The Notebookcheck RTX 4070 desktop results show that the card already operates above 60 FPS at QHD in the listed demanding games. That gives a 60 Hz or 120 Hz project some room, but it isn’t unlimited. Profile this card directly rather than infer its budget from an RTX 4090.
What the present results mean for owners and studios
For most current single-player releases at 1440p, the card remains a reasonable target with DLSS 4.5. Native averages of 80.3 FPS in Cyberpunk 2077 1.6 and 68.7 FPS in Alan Wake 2 both clear 60 FPS in the recorded tests. Competitive games are a different case because players often disable Frame Generation to avoid its base latency and rely on the native frame rate instead. In the most demanding competitive titles, that native 1440p budget is not generous.
A studio can also keep the desktop RTX 4070 on its supported hardware list. It offers DLSS 4.5, Reflex, ray tracing, and AV1 encoding, which cover four major features expected by current PC releases. The one unsupported item is DLSS 5, and a store page shouldn’t claim it before NVIDIA’s SDK does.
Minimum feature set for a shipping PC build
- DLSS Super Resolution with at least Quality, Balanced, and Performance presets, exposed in the in-game graphics menu.
- DLSS Ray Reconstruction enabled by default on presets that use hardware ray tracing.
- DLSS Frame Generation as a separate toggle, off by default in competitive modes.
- NVIDIA Reflex enabled whenever Frame Generation is enabled, with an on-screen indicator.
- DLSS Override option in the NVIDIA control panel, so users with global driver settings can still follow the in-game toggle.
All five belong to the current DLSS 4.5 surface on this card and don’t depend on a future driver or SDK.
What would change the support answer
An official DLSS 5 SDK naming the RTX 40 desktop family would confirm the subset of features it exposes. A Game Ready driver note could do the same for its listed functions. If neither appears, the RTX 4070 remains a DLSS 4.5 card.
A development watchlist can cover the NVIDIA developer blog, DLSS SDK release notes, and Game Ready driver notes. Once one names DLSS 5 on RTX 40 desktop, the team can evaluate a patch. Until then, it isn’t a production dependency.
Misconceptions worth clearing up
The RTX 4070 isn’t automatically barred from every later DLSS feature by its hardware. It shares the RTX 40 Tensor-core generation, optical-flow accelerator family, and driver surface. Nor is DLSS 5 an automatic free upgrade for every RTX card; NVIDIA hasn’t announced that. Finally, Frame Generation performance doesn’t match an RTX 4090 simply because the API is shared. It starts from the RTX 4070’s own native 1440p budget.
| Misconception | What the evidence says | What to plan around |
|---|---|---|
| The desktop RTX 4070 cannot run any new DLSS feature. | The card has the same Tensor cores, optical flow accelerator, and driver surface as the rest of the RTX 40 desktop family. | Hardware is not the gating factor for future DLSS versions on this card. |
| DLSS 5 is a free automatic upgrade on every RTX card. | DLSS 5 support for this card has not been announced, and the release date is unknown. | Treat any DLSS 5 claim on this card as unverified until NVIDIA confirms it. |
| Frame Generation on this card matches the high-end cards. | Native 1440p budget on this card is above 60 FPS in the listed measurements, which is the Frame Generation input budget. | Profile Frame Generation on the actual desktop RTX 4070, not on a higher-tier card. |
Recommendations for the next twelve months
Keep the RTX 4070 in the support matrix and ship the complete DLSS 4.5 path. Put Frame Generation behind its own control, enable Reflex with it, and let the SDK decide which options appear. Owners can continue treating the card as a 1440p model with measured DLSS 4.5 gains. Studio budgets should regard DLSS 5 as an unanswered support question.
When the answer changes, it will change in NVIDIA’s SDK and Game Ready driver notes. Until then, the current hardware facts and the 80.3, 83.9, and 68.7 FPS measurements are the useful record.
Frequently asked questions
Does the GeForce RTX 4070 support DLSS 5?
No official support has been announced, and future compatibility remains unknown. The confirmed feature set is DLSS 4.5 Super Resolution, Ray Reconstruction, Frame Generation, and Reflex.
Which DLSS 4.5 features run on the card?
Super Resolution, Ray Reconstruction, and Frame Generation run on the RTX 4070. NVIDIA Reflex is the latency-reduction feature paired with Frame Generation.
How much performance did DLSS add in the cited QHD test?
Cyberpunk 2077 increased from 80.3 FPS at native 2560×1440 to 83.9 FPS with DLSS at the same output resolution in the Notebookcheck results.
Is Frame Generation worth using?
It suits single-player content where a higher visible frame rate matters more than minimum latency. Pair it with Reflex. Competitive players often leave it off and target the card’s native frame rate.
Does the hardware rule out every future DLSS version?
No. The GPU shares the Ada Tensor-core and optical-flow families with other RTX 40 desktop cards. Official NVIDIA enablement, not a lack of those blocks, is the current gate.
What are the native 1440p reference results?
Notebookcheck reports 80.3 FPS in Cyberpunk 2077 1.6 and 68.7 FPS in Alan Wake 2 at native 2560×1440.
Should a new release advertise DLSS 5 on RTX 40 desktop?
No. The current SDK doesn’t back that claim for the RTX 4070 or wider RTX 40 desktop family. Ship DLSS 4.5 and let future SDK capability checks expose anything NVIDIA later confirms.
What should be profiled before enabling Frame Generation?
Measure Ray Reconstruction post-process time, Frame Generation optical-flow time, and render-queue depth with Reflex both off and on. Confirm that the higher perceived rate doesn’t push input-to-photon latency beyond the no-Frame-Generation baseline.
Will it remain a 1440p card for another year?
The listed native results remain above 60 FPS, so it still belongs in the 1440p tier. Future games may demand more, but the hardware itself won’t change. DLSS 4.5 remains the supported software path until NVIDIA publishes another SDK entry.
Where can developers verify official status?
Check the NVIDIA DLSS SDK notes, Game Ready driver notes, and developer blog. The RTX 40 series page provides hardware context, and the Notebookcheck desktop page provides the cited frame-rate measurements. Treat unsupported claims from elsewhere as unverified.
Leave a Reply