NVIDIA Clarifies RTX 5000 Non-DLSS Gains After Benchmark Backlash

NVIDIA’s RTX 50-series cards do deliver genuine performance gains without DLSS, but the company’s biggest CES 2025 numbers came from DLSS 4—especially Multi Frame Generation—not conventional rendering. Additional Editor’s Day slides reported by Windows Central put the RTX 5070 roughly 20% ahead of the RTX 4070, the RTX 5080 about 15% ahead of the RTX 4080, and the RTX 5090 about 30% ahead of the RTX 4090 in selected non-DLSS comparisons.

Those figures were NVIDIA’s own claims, not independent review averages. They also do not erase the reason for the backlash: CES charts prominently highlighted feature-assisted frame rates that could be several times higher than the underlying rendered-frame rate.

#	Preview	Product	Price
1		Nvidia RTX 5000 Ada Quadro RTX 5000 32 GB GDDR6	$4,135.00	Buy on Amazon

As an Amazon Associate I earn from qualifying purchases.

Why NVIDIA’s RTX 50-series claims caused controversy

At CES 2025, NVIDIA presented the GeForce RTX 50 series around its Blackwell architecture, improved ray-tracing and Tensor Core hardware, neural rendering, and DLSS 4. The most dramatic demonstrations combined DLSS Super Resolution, Frame Generation, and the new Multi Frame Generation feature with demanding ray-traced or path-traced games.

In selected DLSS-enabled scenarios, NVIDIA presented results such as the RTX 5070 performing comparably to an RTX 4090. That was a feature-assisted comparison—not a claim that the RTX 5070 rendered games as quickly as an RTX 4090 at native resolution. NVIDIA’s launch material continued to emphasize very high DLSS 4 frame rates, increasing the pressure for a clearer raw-performance baseline.

#1 Best Overall

Nvidia RTX 5000 Ada Quadro RTX 5000 32 GB GDDR6

• NVIDIA GPUDirect remote direct memory access (RDMA) support
• NVIDIA Quadro Sync II compatibility
• 3D stereo support with stereo connector
• NVIDIA GPUDirect for Video support
• NVIDIA Mosaic technology

$4,135.00

Buy on Amazon

The later information was not a clearly documented standalone apology or formal correction. Rather, NVIDIA shared additional Editor’s Day performance slides, which were reported by Windows Central on January 16, 2025. A later ComputerCity article appears to be a derivative account of the same story, so it should not be treated as an independent confirmation.

What NVIDIA claimed without DLSS

The reported comparisons were as follows:

These numbers come from NVIDIA-provided comparisons as reported by Windows Central. They show what NVIDIA claimed, not a universal performance average across every game.

RTX 5070: roughly 20% over RTX 4070

NVIDIA reportedly claimed about a 20% improvement for the RTX 5070 over the RTX 4070 at 1440p with maximum settings and DLSS disabled. The RTX 5070 launched at a $549 MSRP, compared with $599 for the RTX 4070 at launch.

That is a potentially useful generational improvement, particularly for owners of older RTX 20- or 30-series cards. It is less compelling as an upgrade for an RTX 4070 owner expecting a transformative native-performance jump. The figure also should not be read as a guaranteed 20% gain in every game.

RTX 5070 Ti: avoid false precision

Coverage described the RTX 5070 Ti as offering a similar uplift over the RTX 4070 Ti, but the exact non-DLSS percentage is not documented as clearly in the available reporting. It is therefore safer to describe the figure as NVIDIA’s approximate estimate rather than label it a confirmed 20% average.

The 5070 Ti was also shown reaching 2.9 times the performance of its predecessor in a selected Indiana Jones and the Great Circle DLSS 4 scenario. That result demonstrates the potential of the feature stack; it is not a general 5070 Ti performance claim.

RTX 5080: roughly 15% over RTX 4080

For the RTX 5080, NVIDIA reportedly claimed about 15% more performance than the RTX 4080 at 4K with DLSS disabled. The RTX 5080 launched at the same nominal $999 MSRP as the RTX 4080.

On a conventional native-FPS basis, that makes the upgrade story considerably less dramatic than the CES DLSS 4 charts suggested. The RTX 5080 can still make sense for new 4K buyers or users who value ray tracing and DLSS 4, but RTX 4080 owners should not expect the kind of uplift implied by multi-frame-generated display rates.

RTX 5090: roughly 30% over RTX 4090

NVIDIA reportedly claimed approximately 30% more performance for the RTX 5090 than the RTX 4090 with DLSS disabled at 4K. The RTX 5090 launched at $1,999, compared with the RTX 4090’s $1,599 launch MSRP.

The 30% figure should remain attributed to NVIDIA unless a clearly defined independent average supports it. The RTX 5090’s larger 32GB GDDR7 memory configuration and stronger AI and ray-tracing hardware also matter for demanding games, path tracing, rendering, and creator workloads—but those advantages do not make every conventional game 30% faster.

What “DLSS disabled” actually tells you

In general, a DLSS-disabled test removes NVIDIA’s DLSS upscaling and frame-generation features. It does not automatically prove that every other testing variable was identical or that the image was rendered in the same way across all comparisons.

A meaningful reading of any “non-DLSS” result requires answers to several questions:

• Was the game rendered natively at the stated output resolution?
• Was ray tracing or path tracing enabled, or was this a rasterization-only test?
• Were Frame Generation and Multi Frame Generation both disabled?
• Were identical quality settings, drivers, CPUs, memory configurations, and BIOS settings used?
• Was the benchmark GPU-limited rather than CPU-limited?
• Was the result from one selected title or an average across a broad test suite?
• Were both cards reference-clocked, or were factory-overclocked models involved?

The available reporting identifies the RTX 5070 comparison as a 1440p maximum-settings test and the RTX 5080 and RTX 5090 comparisons as 4K tests. It does not provide a complete, independently reproducible methodology for every card. “Non-DLSS” is therefore useful shorthand, but not the same thing as a fully documented independent benchmark average.

It is also important not to equate non-DLSS with rasterization-only performance. DLSS can be disabled while ray tracing remains enabled, producing a native ray-tracing result rather than a traditional raster benchmark.

Why DLSS 4 creates much larger numbers

There are three different kinds of frames to distinguish:

1. Native rendered frames: conventionally produced by the GPU at the game’s rendering resolution.
2. Upscaled frames: reconstructed from a lower-resolution internal render using DLSS Super Resolution.
3. Generated frames: synthesized between rendered frames using motion data and AI models.

Multi Frame Generation increases the number of frames delivered to the display by creating additional frames between conventionally rendered ones. It does not mean the GPU is rendering every displayed frame in the traditional sense.

As Tom’s Hardware’s testing explained, a game producing 50 base frames per second can display substantially more frames with frame generation while still sampling player input at the lower base rate. A counter showing 200 FPS after Multi Frame Generation should not be described as equivalent to 200 FPS of native rendering.

Generated frames can improve perceived smoothness, particularly in supported single-player games, but they do not eliminate the need for a sufficiently high base frame rate. Responsiveness, latency, artifacts, fast-motion behavior, and implementation quality still matter. Competitive players and anyone sensitive to input delay may value native rendered frames more highly than a larger generated-FPS number.

What independent testing added

Later testing helped separate DLSS 4’s feature gains from the cards’ underlying rendering performance. Tom’s Hardware reported substantial scaling from Multi Frame Generation on both the RTX 5080 and RTX 5090:

Those results are useful evidence that Multi Frame Generation can raise displayed FPS dramatically in selected conditions. They are not raw rasterization gains, and they should not be summarized as the RTX 5080 or RTX 5090 being three or four times faster than the previous generation.

For buyers, the useful comparison is not one single FPS number but several separate measurements:

• Native raster performance: important for games without DLSS and for buyers who prefer conventional rendering.
• Native ray-tracing performance: important for RT-heavy games when upscaling is disabled.
• DLSS Super Resolution performance: relevant when image reconstruction is acceptable.
• Frame Generation performance: relevant when the base FPS and latency are already satisfactory.
• Multi Frame Generation performance: relevant to supported games and users prioritizing smoother display motion.
• Latency and frame-time behavior: essential because a high displayed FPS does not guarantee equally responsive controls.

Which RTX 50-series cards make sense?

RTX 5070

The RTX 5070 is most defensible for a new 1440p build, owners of substantially older GPUs, and buyers who specifically want DLSS 4 and Blackwell’s newer feature set. The case is weaker for RTX 4070 owners seeking a major native upgrade or for players whose libraries lack DLSS support.

RTX 5070 Ti

The 5070 Ti offers more headroom for high-refresh 1440p and entry-level 4K gaming. Its exact non-DLSS gain over the RTX 4070 Ti should be treated as NVIDIA’s approximate claim, not a precisely verified average. Do not use the 2.9x Indiana Jones result as a universal performance expectation.

RTX 5080

The RTX 5080 is aimed at 4K players, ray-tracing enthusiasts, and buyers who want DLSS 4 support. Its reported native uplift over the RTX 4080 is only about 15%, however, so existing RTX 4080 owners should judge the purchase on their need for newer features and additional headroom—not on the largest AI-assisted CES numbers.

Its 16GB of VRAM may also be a consideration for buyers planning to keep the card for demanding 4K games over many years.

RTX 5090

The RTX 5090 is the choice for maximum 4K performance, heavy ray tracing or path tracing, and demanding AI, rendering, or creator workloads. It is a poor value proposition for buyers who simply want the most native FPS per dollar, and RTX 4090 owners satisfied with current performance may find the reported native uplift insufficient to justify the higher $1,999 launch MSRP.

Buying based on the right benchmark

Buyers who disable DLSS should prioritize independent reviews showing average FPS, 1% lows, native 4K results, ray tracing with upscaling disabled, VRAM capacity, and performance in their own games. At 1080p—and sometimes at 1440p—a CPU bottleneck can prevent a faster GPU from showing its full advantage.

Buyers who use DLSS-supported games should additionally examine Super Resolution image quality, base FPS before generation, latency, artifact behavior, and whether the games they play support Frame Generation or Multi Frame Generation. AMD Radeon cards and FidelityFX Super Resolution remain relevant alternatives for buyers prioritizing conventional raster performance, cross-vendor upscaling, or price-sensitive gaming. NVIDIA generally retains stronger appeal for DLSS, ray tracing, CUDA, and AI-related software support.

The cited prices—$549 for the RTX 5070, $749 for the RTX 5070 Ti, $999 for the RTX 5080, and $1,999 for the RTX 5090—are January 2025 launch MSRPs. They are not verified August 2026 street prices, so current buying decisions should use current retailer pricing rather than these historical figures.

Separate issue: missing-ROP RTX 50-series cards

Some RTX 5090, RTX 5090D, RTX 5080, and RTX 5070 Ti cards were later found to have eight missing Raster Operations Pipelines. This manufacturing issue is separate from the CES benchmark controversy.

GamersNexus testing of a defective RTX 5080 with 104 ROPs instead of the expected 112 found losses ranging from negligible in some tests to roughly 10–11% in others. Owners can use GPU-Z to inspect reported hardware characteristics, but the software is not a warranty-resolution tool. Anyone who finds a potentially affected card should contact the retailer or board partner rather than attempt a hardware repair.

Quick Recap

Bestseller No. 1

Nvidia RTX 5000 Ada Quadro RTX 5000 32 GB GDDR6

NVIDIA GPUDirect remote direct memory access (RDMA) support; NVIDIA Quadro Sync II compatibility

$4,135.00

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.