GeForce RTX 2070 SUPER DLSS 4.5 vs DLSS 5: outlook

GeForce RTX 2070 SUPER DLSS 4.5 vs DLSS 5

No official DLSS 5 support has been announced for the GeForce RTX 2070 SUPER. Its documented ceiling is DLSS 4.5 Super Resolution and Ray Reconstruction, so that is the feature set a shipping build should target.

In 2026, the card occupies an awkward middle ground. It was the last desktop GeForce tier in its generation without the later dedicated AI acceleration path for upscaling, yet it remains the strongest older-generation model that many developers still target. Published results for this exact desktop GPU show where the practical limits sit.

The 2019 desktop card still has clear limits

The GeForce RTX 2070 SUPER launched in July 2019 with a TU104 GPU, 2,560 CUDA cores, 8 GB of GDDR6 on a 256-bit bus, and 215 W board power. It is a desktop-only model sold as a Founders Edition and through board partners. Laptop and Max-Q variants are separate SKUs with lower power limits and slower memory.

Results from an RTX 2070 SUPER laptop, a Max-Q card, or a non-SUPER RTX 2070 should not be transferred to this desktop GPU. Its 8 GB VRAM buffer also limits the modern texture sets and output resolutions a studio can promise.

NVIDIA has delivered several DLSS generations on the same Turing tensor cores since launch. DLSS 4.5 Super Resolution and Ray Reconstruction are the newest modes documented for the RTX 20 series. Turing tensor cores support INT8 and FP16 operations, which helps explain why later feature tiers have skipped this generation.

DLSS 4.5 covers upscaling and ray reconstruction

DLSS 4.5 is the newest NVIDIA stack documented for the RTX 20 series. Its two supported techniques address different parts of the frame:

DLSS Super Resolution renders below the output resolution and reconstructs the final image using Quality, Balanced, Performance, or Ultra Performance ratios. With 8 GB of VRAM, Quality and Balanced are the realistic choices for 1440p.
DLSS Ray Reconstruction replaces hand-written ray-tracing denoisers with a trained network. That can reduce the rays per pixel needed for a given visual target and determine whether ray-traced reflections or global illumination fit the budget.

DLSS 4.5 is the upper bound a studio can assume for this card. Newer frame generation, multi-frame generation, and the Transformer-based architecture introduced for the RTX 40 and RTX 50 series are outside its documented feature set.

DLSS 5 support remains unannounced

No official DLSS 5 launch support has been announced for the GeForce RTX 2070 SUPER, and future support remains unknown. NVIDIA has added DLSS capabilities to older hardware through later driver work before, but it has made no promise here.

Three hardware limits shape that uncertainty:

Turing predates the Optical Flow Accelerator 2.0 used by newer frame generation paths. Any DLSS 5 feature relying on hardware optical flow would need a fallback, and none has been documented for the RTX 20 series.
Turing tensor cores use INT8 and FP16, while newer transformer-based upscaling and reconstruction networks tend to assume FP8 or INT4 inference. Support may be possible, but the same model would carry a different runtime cost.
The 8 GB VRAM limit must hold intermediate DLSS buffers as well as the final image. Quality and Balanced remain the only realistic internal-resolution choices for 1440p output.

Do not include DLSS 5 in the shipping plan for this card unless NVIDIA confirms support. The mode may arrive later, but current code and UI should not assume it will.

Published 1440p results set the native baseline

These are the only published average FPS results for this exact GPU in shipped games, taken from Notebookcheck. The GeForce 20 series reference supplies the wider architecture context. Each figure is an average at 2560×1440 using the benchmark source’s settings, not a maximum from a tuned scene or a worst-case stress result.

Game	Resolution	Average FPS	Source
Cyberpunk 2077 1.6	2560×1440	46.6	Notebookcheck, NVIDIA GeForce RTX 2070 Super GPU benchmarks page
Cyberpunk 2077	2560×1440	50.5	Notebookcheck, NVIDIA GeForce RTX 2070 Super GPU benchmarks page
Alan Wake 2	2560×1440	31.4	Notebookcheck, NVIDIA GeForce RTX 2070 Super GPU benchmarks page

The two Cyberpunk 2077 entries come from separate test passes on the same GPU. Their 46.6 and 50.5 averages are consistent with different scene mixes rather than different presets. Alan Wake 2 is the harsher reference: its heavy ray-traced lighting brings the card to a 31.4 average at native 1440p before upscaling.

No DLSS-on result appears in the source for these passes. Measure the on-versus-off difference in the scene and build you plan to ship.

Ada and Blackwell have hardware Turing lacks

The RTX 40 series based on Ada Lovelace added Optical Flow Accelerator 2.0 and tensor cores suited to newer transformer-based reconstruction networks. The Blackwell RTX 50 series extends that path and provides more headroom for multi-frame generation. The GeForce 20 series predates the current optical flow design and uses an earlier tensor pipeline.

Turing documentation and the wider GeForce 20 series history explain why some modes appear on newer cards but not older ones. This is an architectural gap, not merely a menu or marketing decision.

A game using transformer or optical-flow features cannot send the RTX 2070 SUPER down the same path. It must fall back to DLSS 4.5 Super Resolution and Ray Reconstruction, or use a separate lower-quality option for Turing.

1440p is constrained by memory, compute, and ray tracing

Three limits matter when profiling this card:

VRAM ceiling. The 8 GB of GDDR6 holds the framebuffer, DLSS intermediate buffers, ray-tracing acceleration structures, and high-resolution textures. A 4K texture pack can make that budget tight at 1440p, so studios commonly cap the optional texture setting at High rather than Ultra.
Compute ceiling. The 2,560 CUDA cores and 320 Turing tensor cores can run DLSS 4.5 at Quality or Balanced without the upscaler becoming the bottleneck. Performance and Ultra Performance are available but seldom worthwhile here because their image-quality loss is visible.
Ray-tracing ceiling. First-generation Turing RT cores accelerate BVH traversal at lower throughput than the second- and third-generation hardware in newer cards. Ray Reconstruction helps, but it cannot replace a faster RT pipeline in scenes with many bounces or dense reflections.

Alan Wake 2’s 31.4 average at 1440p shows what happens when a demanding workload meets that first-generation RT pipeline. Extra effects still have to fit inside the same frame-time and memory limits.

Choose the DLSS 4.5 mode by workload

This is a configuration guide, not a measured benchmark. It maps common workloads to sensible DLSS 4.5 modes. DLSS 5 is absent because NVIDIA has not announced launch support for this GPU.

Workload	Target output	Recommended DLSS 4.5 mode	Why
Heavy ray-traced global illumination, 1440p output	2560×1440	DLSS 4.5 Quality + Ray Reconstruction	Keeps internal resolution above 720p, lets the network denoise the RT result
Standard forward rendering, 1440p output	2560×1440	DLSS 4.5 Quality	Closest to native image, fits inside the 8 GB VRAM budget
1080p output with high refresh target	1920×1080	DLSS 4.5 Balanced or Performance	Lowers the internal cost to keep frame time under a high refresh budget
4K output on this card	3840×2160	Not recommended	VRAM budget and RT throughput cannot sustain 4K reliably on this generation
Multi-frame generation path	Any	Not supported on this card	Requires an Optical Flow Accelerator generation that Turing does not have

Balanced or Performance makes the most sense for a high-refresh 1080p target. At 1440p, the more aggressive ratios usually cost too much image quality. The 4K warning comes from the combined VRAM and RT-throughput limits, not from the DLSS version alone.

Profile the exact shipping scenes

A studio targeting the GeForce RTX 2070 SUPER should run three checks before locking its DLSS 4.5 profile.

Measure DLSS 4.5 Quality at the target resolution. Capture average FPS, 1 percent low, and VRAM use with DLSS off and then at Quality in the same heavy shipping scene. A synthetic scene is not a substitute.
Compare Ray Reconstruction with the existing denoiser. Run the trained network beside any custom reflections or global-illumination denoiser. The visual result is not always better, and network cost at low internal resolutions can consume the expected gain.
Set the older-card fallback before release. Expose DLSS 4.5 Quality and Balanced, hide transformer- or optical-flow-dependent modes, and document the cutoff. This avoids promising an unannounced DLSS 5 path.

Public release notes should name the DLSS modes available to each NVIDIA generation. Clear cutoffs reduce support tickets from players who think an option is missing.

Possible DLSS 5 outcomes

NVIDIA has not documented an outcome for this card. Based on its treatment of older DLSS hardware, the possibilities are:

No RTX 20 release. Turing’s tensor and optical-flow design is a generation behind, leaving DLSS 4.5 Super Resolution and Ray Reconstruction as the ceiling.
A reduced RTX 20 feature set. NVIDIA could deliver upscaling and reconstruction without the frame generation path, as it has shipped tensor-only DLSS updates before.
A version close to the headline release. This looks less likely on the architecture, but driver work could bring the network cost within Turing’s limits.

None of these scenarios justifies advertising support now. Keep DLSS 5 out of the RTX 2070 SUPER plan until NVIDIA publishes a list that includes the card.

Match the menu to the supported card list

Players can still use DLSS 4.5 Quality for current games at 1440p. The published averages provide a native 1440p reference for demanding titles, but DLSS 5 and multi-frame generation should not appear in this card’s menu.

If the production target includes the RTX 2070 SUPER, lock the DLSS 4.5 path, hide features the card lacks, and publish the cutoff. A studio that excludes this GPU can leave DLSS 5 enabled for its supported cards. The choice depends on how much of the game’s PC audience still uses Turing.

Validate the profile before release

For each question, record a number, screenshot, or named test scene from the same release candidate:

What is the average FPS in the heaviest shipped scene at 1440p with DLSS 4.5 Quality on?
What is the 1 percent low in the same scene under the same settings?
What is the VRAM use in the same scene under the same settings?
Which DLSS modes are visible in the menu on this card, and which are hidden?
What does the release note say about DLSS 5 support on the RTX 20 series?

Those five answers provide enough evidence for a public patch note without inventing unsupported performance claims.

The published data has important gaps

There is no published result for this exact GPU with DLSS 4.5 enabled in Cyberpunk 2077 or Alan Wake 2, no 4K figure, and no frame-generation result because that feature is outside Turing’s documented matrix. A claimed DLSS 4.5-versus-native delta must name its source and test scene; otherwise it may be an interpolation from another GPU.

Use the published figures as a native reference, not a replacement for testing. The actual DLSS 4.5 delta depends on your build, target hardware, scene, and workload.

If support arrives later

An official DLSS 5 release for the RTX 2070 SUPER would require three changes:

Update the support matrix. Revise menus, profiles, and release notes to include the card under the same conditions as other Turing SKUs.
Replace the fallback for new modes. If DLSS 5 Quality becomes available, DLSS 4.5 Quality is no longer the only documented ceiling and the new mode can appear in the menu.
Measure a new comparison. The 46.6, 50.5, and 31.4 averages remain the native 1440p baseline, but the DLSS-on comparison gains another mode.

Until NVIDIA makes that announcement, keep the card on DLSS 4.5, retain the fallback, and measure against the published native 1440p averages.

Frequently asked questions

Does the GeForce RTX 2070 SUPER support DLSS 5?

No official support has been announced, and future support remains unknown. DLSS 4.5 Super Resolution and Ray Reconstruction are the newest modes documented for the card.

What is the latest DLSS version the RTX 2070 SUPER actually runs?

DLSS 4.5 is the newest documented stack. It combines Super Resolution with Ray Reconstruction, runs on Turing tensor cores, and does not require the newer optical-flow hardware used for RTX 40 and RTX 50 series frame generation.

Can the RTX 2070 SUPER run Cyberpunk 2077 at 1440p?

Yes. Published native results at 2560×1440 show 46.6 FPS in Cyberpunk 2077 1.6 and 50.5 FPS in a different Cyberpunk 2077 test pass. Neither result uses DLSS.

What about Alan Wake 2 on the RTX 2070 SUPER?

The same Notebookcheck page reports 31.4 FPS at 2560×1440. It is a useful native baseline for a demanding current game before DLSS is enabled.

Is DLSS 4.5 Quality enough for ray tracing on the RTX 2070 SUPER?

DLSS 4.5 Quality with Ray Reconstruction is the documented 1440p option for ray-traced reflections and global illumination. Its trained network can replace a hand-written denoiser and reduce the rays per pixel needed for the same target. Quality stays closer to native than Performance or Ultra Performance and fits the 8 GB VRAM budget.

Should a GameDev studio ship DLSS 5 modes on a game that targets the RTX 2070 SUPER?

No. Expose the documented DLSS 4.5 modes, hide anything dependent on newer transformer or optical-flow features, and state the cutoff in the release notes.

Does frame generation work on the RTX 2070 SUPER?

Not through the documented RTX 40 and RTX 50 series path. That implementation depends on a newer Optical Flow Accelerator than Turing provides, and frame generation is absent from the RTX 2070 SUPER support matrix.

What is the realistic DLSS 4.5 mode for a 1080p high-refresh target on this card?

DLSS 4.5 Balanced or Performance is the realistic choice. It lowers the internal resolution while retaining acceptable 1080p output and can free enough frame time for a high-refresh target.

What is the realistic DLSS 4.5 mode for a 4K target on this card?

4K is not a realistic target. The 8 GB VRAM limit and first-generation RT pipeline cannot sustain it reliably, while aggressive DLSS ratios give up too much image quality.

What should a studio do if NVIDIA does announce DLSS 5 for the RTX 2070 SUPER later?

Update the support matrix, expose the new mode, and measure its DLSS-on result in the heaviest scene. The native 1440p baseline will not change. Until then, DLSS 4.5 remains the ceiling and the existing fallback stays in place.

For the current release, lock DLSS 4.5 in the PC profile, hide transformer- and optical-flow-dependent modes, and document the cutoff.

GeForce RTX 2070 SUPER DLSS 4.5 vs DLSS 5: what ships today