NBA 2K27 DLSS 5 Verdict: 4 of 6 RTX 50 GPUs Proven [2026]

NVIDIA’s DLSS 5 went live on September 3, 2026, bundled with NBA 2K27 and a new Game Ready driver, and the headline number NVIDIA put on stage was hard to miss: “up to 370 frames per second” on a GeForce RTX 5090 at 4K with ray tracing turned on. Within days, independent testers at outlets including GameGPU, ComputerBase, and PCMag had pulled that number apart, and the picture that emerged is messier than the marketing slide. DLSS 5’s new neural rendering pipeline does something no prior DLSS version attempted, but it also costs 50% to 60% of a card’s base frame rate to run, according to NVIDIA’s own briefing language, before Multi Frame Generation multiplies whatever is left into the eye-catching totals shown in keynote charts.

The bigger story, and the one NVIDIA’s press materials gloss over, is that only four of the six current RTX 50-series desktop GPUs have any independently verified NBA 2K27 numbers at all. The RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070 all appear in NVIDIA’s own performance charts and in third-party benchmarks. The RTX 5060 Ti and RTX 5060, the cards most buyers actually own, do not. Here is what the testing that does exist actually shows, GPU by GPU, resolution by resolution.

Google · Preferred Sources

Don't miss new tech stories on Google

Add Tech Insider once in the Google app and our stories appear in your news suggestions.

Add Now

What NVIDIA Announced on September 3

NVIDIA’s GeForce news post confirmed that “DLSS 5 is available starting now in NBA 2K27, from developer Visual Concepts and publisher 2K, for all GeForce RTX 50 Series GPUs and laptops, and GeForce NOW,” positioning the basketball sim as the flagship launch title for the new rendering technology, according to NVIDIA’s official announcement. The rollout arrived alongside a matching Game Ready driver, with NVIDIA’s driver page describing “DLSS 5 is available now in the newly-released NBA 2K27,” per the Game Ready driver announcement.

Developer Visual Concepts also published a short video walking through how the studio tuned the technology for basketball specifically, with NVIDIA’s GeForce account posting on X: “Hear from @NBA2K developer, Visual Concepts, on how they tuned NVIDIA DLSS 5 to help achieve a more lifelike NBA experience in 2K27. NBA 2K27 is #RTXON 9/3,” a promotional post that framed the collaboration as a joint engineering effort rather than a simple driver-level toggle, per NVIDIA GeForce’s post on X.

What makes this launch different from prior DLSS generations is the underlying technique. DLSS 5 introduces what NVIDIA calls 3D-guided neural rendering. Earlier DLSS versions upscaled a lower-resolution rendered frame (DLSS 1 and 2) or generated interpolated frames between two real frames to boost apparent frame rate (DLSS 3 and 4). DLSS 5 instead has the GPU’s AI hardware actively rewrite lighting, shadows, and surface detail on frames the game engine has already rendered, aiming for a more photorealistic look rather than just a smoother or sharper one. That extra AI workload is also exactly why it costs so much raw performance to run.

The 370 FPS Claim, Decoded

NVIDIA’s own promotional language reads: “Using the full suite of DLSS technologies in NBA 2K27, at 4K with the Ultra Preset and ray tracing, GeForce RTX 5090 owners can play at up to 370 frames per second with DLSS 5, for the definitive experience,” according to the DLSS 5 announcement post. At 1440p Ultra with ray tracing, NVIDIA’s chart lists 590 FPS for the RTX 5090, 410 FPS for the RTX 5080, 350 FPS for the RTX 5070 Ti, and 260 FPS for the RTX 5070.

As previous testing of DLSS 5’s marketing claims has shown, those numbers are not rendered frame rates. They are the display output after Multi Frame Generation inserts several AI-generated frames for every frame the GPU actually renders, at ratios NVIDIA refers to internally as MFG multipliers, running as high as 6x in NBA 2K27’s Ultra preset. Analysis of NVIDIA’s own chart data shows the 370 FPS figure at 4K on the RTX 5090 corresponds to roughly 62 FPS of actual rendered frames, with six generated frames inserted per real one. At 1440p, the RTX 5090’s base render rate sits just under 100 FPS before MFG does its work. NVIDIA has been fairly direct about the tradeoff in briefings, telling press that enabling DLSS 5’s new rendering technology on its own “can knock roughly 50 to 60 percent off frame rates, depending on the game, scene, and resolution,” a figure multiple outlets covering the launch reported as NVIDIA’s own average across the full RTX 50-series lineup.

RTX 5090: The Only Card With Data at Every Resolution

The RTX 5090 is the only GPU in the lineup that independent reviewers tested at 1080p, 1440p, and 4K, which makes it the clearest picture of how DLSS 5 actually behaves once the marketing multiplier is stripped away. GameGPU’s NBA 2K27 testing found that at 4K with max settings and ray tracing, turning on DLSS 5 dropped the RTX 5090 from 192 FPS to 76 FPS, close to a 60% loss. A separate 4K test reported by a Spanish-language outlet found a near-identical pattern: 204 FPS natively versus 78 FPS with DLSS 5 enabled, a drop the reviewer calculated at 61.76%. The same outlet also measured DLSS Quality mode with 6x Multi Frame Generation at 4K, where the displayed output hit 371 FPS against a base of 730 FPS with DLSS 5’s neural rendering switched off, translating to a 49.17% cut in underlying rendered performance once the neural rendering pass is active.

Independent hands-on testing from hardware journalist Ryan Shrout matched that range closely. Running the RTX 5090 at 4K with settings maxed, DLAA enabled, and no frame generation of any kind, Shrout recorded gameplay going from roughly 150 FPS with neural rendering off to roughly 75 FPS with it on, essentially a straight halving of frame rate. German outlet ComputerBase ran its own pass and found the RTX 5090 dropping from 205 FPS to 132 FPS at 1440p (a 36% cut) and to 77 FPS at 4K, a loss the outlet pegged at 57%.

ResolutionNative FPS (RTX 5090)DLSS 5 FPS (RTX 5090)Performance ChangeSource
1080p, Max + RT180 FPS134 FPS-26%Independent 2026 hands-on test
1440p, Max + RT176 FPS108 FPS-39%Independent 2026 hands-on test
1440p, Ultra + RT205 FPS132 FPS-36%ComputerBase
4K, Max + RT192 FPS76 FPS-60%GameGPU
4K, Native (alt. run)204 FPS78 FPS-61.76%Spanish-language outlet review
4K, Ultra + RT (DLAA, no MFG)~150 FPS~75 FPS~-50%Ryan Shrout hands-on test
4K, Ultra + RT77 FPS-57%ComputerBase

Read across all seven data points and a pattern holds: the higher the resolution, the bigger DLSS 5’s neural rendering tax. At 1080p, the RTX 5090 loses about a quarter of its frame rate. At 4K, it loses well over half. That single fact, more than the 370 FPS headline, is the number that should shape how buyers think about which RTX 50-series card is actually worth pairing with DLSS 5.

RTX 5080: NVIDIA’s Numbers vs. Independent Testing

NVIDIA’s own chart puts the RTX 5080 at 410 FPS at 1440p Ultra with ray tracing and DLSS 5 enabled, second only to the RTX 5090’s 590 FPS on the same chart. Independent testing paints a more modest picture. ComputerBase’s 4K pass found the RTX 5080 falling to roughly 48 FPS with DLSS 5’s neural rendering pass active, a card that is nominally the second-tier option in the lineup landing well below a comfortable 60 FPS floor once ray tracing and neural rendering both compete for the same GPU die at native 4K.

No outlet in the current wave of testing has published a dedicated RTX 5080 breakdown at 1080p the way they have for the RTX 5090. Given the RTX 5090’s own resolution curve, where the neural rendering penalty shrinks from roughly 60% at 4K to roughly 26% at 1080p, it is reasonable to expect the RTX 5080 to clear 1080p comfortably. But that is an inference from the RTX 5090’s scaling pattern, not a number anyone has actually measured yet on a 5080 specifically, and readers should treat it as such until a reviewer runs the test.

RTX 5070 Ti and RTX 5070: Where NVIDIA’s Chart Stops

The RTX 5070 Ti and RTX 5070 both appear in NVIDIA’s own 1440p Ultra chart, at 350 FPS and 260 FPS respectively with DLSS 5 and ray tracing enabled. Those are the lowest tier of GPUs NVIDIA chose to feature by name in its NBA 2K27 marketing materials, and as with the RTX 5080, no independently published GameGPU, ComputerBase, or PCMag test in the current round of coverage breaks out per-resolution native-versus-DLSS-5 numbers for either card. Buyers shopping at this price tier are, for now, working from NVIDIA’s own marketing figures rather than a third-party benchmark, which is a meaningfully different level of confidence than the RTX 5090 numbers above.

RTX 5060 Ti and RTX 5060: The Two GPUs Nobody Has Tested Yet

This is the gap that matters most for the widest slice of buyers. PCMag’s hands-on coverage of the launch states plainly: “To use DLSS 5, however, you need an Nvidia GeForce RTX 50-series graphics card,” describing NBA 2K27 as the first title built around the technology, per PCMag’s review. That requirement covers the entire RTX 50-series family in principle. In practice, every specific FPS figure published so far, whether from NVIDIA’s own chart or from GameGPU, ComputerBase, or independent hands-on testers, covers only the RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070.

NVIDIA’s generalized guidance that DLSS 5 costs “50 to 60 percent” of frame rate across all RTX 50-series GPUs implies the RTX 5060 Ti and RTX 5060 meet the architectural and VRAM requirements to run the feature. But implication is not a benchmark. Given that the RTX 5080 already drops close to, or below, a smooth 60 FPS at native 4K with DLSS 5 active, the realistic expectation for the RTX 5060 Ti and RTX 5060, both considerably less powerful cards, is that 4K with DLSS 5 will be a rough experience and 1440p will need careful settings management. Until a reviewer actually runs the numbers, that remains an educated projection rather than a confirmed fact, and it is worth being explicit about that distinction given how much of the coverage so far blurs “supported” with “tested.”

1080p Is Where DLSS 5 Stops Collapsing Frame Rates

The single most useful number to come out of this launch week is the RTX 5090’s 1080p result: 180 FPS native down to 134 FPS with DLSS 5 enabled, a roughly 26% hit. That is dramatically smaller than the 57-61% losses recorded at 4K on the same card. Part of the explanation is that NBA 2K27, like most sports titles, becomes CPU-bound at 1080p well before the GPU is the bottleneck, so DLSS 5’s added rendering cost eats into a smaller share of the total frame budget. The practical upshot is that 1080p is currently the one resolution where DLSS 5’s neural rendering pass does not threaten to push a high-end card below a smooth, responsive frame rate, even before Multi Frame Generation is layered on top.

That finding cuts against the way NVIDIA has marketed DLSS 5, which leans almost entirely on 4K and 1440p Ultra charts designed to showcase the biggest possible FPS multipliers. For anyone actually deciding whether to turn DLSS 5 on in NBA 2K27 today, on hardware that exists today, 1080p is the resolution where the technology’s cost is least likely to be noticeable, while 4K is where it is most likely to require compromises elsewhere in the settings menu.

Power Draw: The Bill Nobody Mentions in the Marketing Slide

DLSS 5’s neural rendering pass does not just cost frame rate, it costs power, a pattern earlier power-draw benchmarking on DLSS 5 had already flagged before NBA 2K27 shipped. ComputerBase’s testing found RTX 5090 board power climbing from 299W to 500W at 1440p when DLSS 5 was switched on, an increase of roughly 200 watts, and from 417W to 561W at 4K, an increase of about 140 watts. That is a meaningful jump for anyone running a 1000W or smaller power supply alongside other high-draw components, and it is a cost that scales with how aggressively the neural rendering pass is engaged, not just with resolution.

For system builders, the combination of a heavier power draw and a lower base frame rate means DLSS 5 is not simply a free image-quality toggle. It is closer to a new, GPU-intensive rendering mode that trades both performance headroom and thermal and power headroom for a different look, and it should be budgeted for accordingly when tuning clocks and power limits on a new RTX 50-series build.

Full RTX 50-Series DLSS 5 Comparison Table

Putting NVIDIA’s own chart alongside the independent testing that exists so far produces a clearer readiness picture than either source alone.

GPUNVIDIA 1440p Ultra + RT ClaimIndependent 1080p/4K Data Published?DLSS 5 Real-World Status
RTX 5090590 FPSYes, at 1080p, 1440p, and 4KConfirmed across every resolution; 1080p loses ~26%, 4K loses 57-62%
RTX 5080410 FPSPartial (4K only, ~48 FPS with DLSS 5)Confirmed at 4K; no independent 1080p figures yet
RTX 5070 Ti350 FPSNo independent breakdown publishedNVIDIA marketing figure only
RTX 5070260 FPSNo independent breakdown publishedNVIDIA marketing figure only
RTX 5060 TiNot listed by NVIDIA for this titleNo data publishedAssumed supported by architecture; unverified in NBA 2K27
RTX 5060Not listed by NVIDIA for this titleNo data publishedAssumed supported by architecture; unverified in NBA 2K27

Four of six current RTX 50-series desktop GPUs have some form of published, attributable NBA 2K27 DLSS 5 performance data. Two do not. That gap is worth watching, because the RTX 5060 Ti and RTX 5060 are, by unit volume, typically the most common cards in NVIDIA’s mid-range stack, based on how NVIDIA’s own GPU lineup has historically skewed toward mid-range sales volume.

Why NVIDIA Leans on Multi Frame Generation Multipliers

NVIDIA’s decision to headline DLSS 5 with Multi Frame Generation output numbers rather than base render rates is not new; the company took a similar approach with DLSS 3 and DLSS 4. What is different this time is the size of the gap between the two numbers. Earlier frame generation implementations that doubled frame rate (2x MFG) created a roughly 2:1 gap between real and displayed frames. DLSS 5’s 6x MFG ratio in NBA 2K27’s Ultra preset creates a much wider spread, which is exactly why a roughly 62 FPS base can be marketed as 370 FPS.

That marketing choice matters because Multi Frame Generation’s inserted frames are synthetic. They do not reduce input latency the way a genuinely higher render rate would, and in a competitive sports title where reaction time to on-court action matters, the gap between a smooth-looking display and a genuinely responsive one is not cosmetic. Reviewers testing DLSS 5 in NBA 2K27 have generally focused their published numbers on the base rendered frame rate for exactly this reason, treating the MFG-multiplied output figure as a display statistic rather than a performance one.

Historical Context: From DLSS 2 Upscaling to DLSS 5 Neural Rendering

DLSS has moved through distinct technical eras since its 2019 debut. The first two versions were pure upscaling technologies: render a lower-resolution frame, then use AI to reconstruct a sharper image at the target resolution, a technique documented in detail on the Deep Learning Super Sampling overview. DLSS 3, introduced with the RTX 40-series, added Frame Generation, inserting one AI-generated frame between two rendered frames to boost apparent smoothness without touching the underlying render resolution. DLSS 4 extended that idea to a 4x multiplier on RTX 50-series hardware.

DLSS 5 represents a structurally different step. Rather than only reconstructing resolution or interpolating motion, its neural rendering pass actively rewrites lighting, shadow, and surface detail on top of frames the engine has already produced, a shift NVIDIA frames as a move toward AI-assisted photorealism rather than pure performance scaling, according to details published on NVIDIA’s DLSS technology page. That is also precisely why it costs so much more raw compute than earlier DLSS generations, and why NVIDIA restricted it to RTX 50-series hardware at launch rather than extending it back to RTX 40-series cards the way some earlier DLSS features eventually were.

Market Impact: What This Means for RTX 50-Series Demand

DLSS 5 being tied exclusively to RTX 50-series hardware gives NVIDIA a fresh, title-specific reason for RTX 40-series owners to consider upgrading, at a moment when GPU pricing across the RTX 50 lineup has already been under pressure from tariff talk and component costs through 2026. NBA 2K27’s enormous annual player base, larger than most single-player showcase titles NVIDIA has used to launch prior DLSS generations, gives the feature broader visibility than a niche ray-tracing tech demo would.

At the same time, the performance data now public works against NVIDIA’s own marketing in one specific way: buyers researching whether to pay a premium for an RTX 5080 or RTX 5090 specifically for DLSS 5 in NBA 2K27 are finding independently verified numbers showing a 50-60% frame rate cost at higher resolutions, not the 370 FPS headline. That is likely to steer performance-conscious buyers who care about DLSS 5 in this title toward running it at 1080p, where the RTX 5090’s 26% hit is far more tolerable, rather than at native 4K. It is one more data point in the broader AI chips and GPU pricing story playing out across the RTX 50-series lineup this year.

What Comes Next: 5 Predictions

  • Expect GameGPU, ComputerBase, TechPowerUp, and Tom’s Hardware to publish full RTX 5060 Ti and RTX 5060 NBA 2K27 benchmarks within the next few weeks, closing the current data gap on the lineup’s most popular tier.
  • NVIDIA will likely add more titles to the DLSS 5 roster before the end of 2026, following the same RTX 50-series exclusivity pattern it used at DLSS 5’s debut.
  • Driver updates in the coming months will probably narrow the performance gap somewhat as NVIDIA optimizes the neural rendering pass, similar to how DLSS 3’s overhead shrank across its first several driver revisions after RTX 40-series launch.
  • Expect competitive and esports-focused NBA 2K27 players to largely leave DLSS 5 off in favor of native rendering or lighter DLSS modes, prioritizing input responsiveness over the neural rendering look.
  • AMD and Intel will likely use DLSS 5’s performance cost as a talking point in their own upscaling and frame generation marketing, positioning FSR and XeSS as lower-overhead alternatives even without matching DLSS 5’s visual feature set.

How This Compares to Prior DLSS Launches

DLSS 3’s 2022 debut with the RTX 40-series also drew scrutiny over marketing framerates that leaned on Frame Generation multipliers, but the base-to-output gap was smaller because DLSS 3 only doubled frame rate at most. DLSS 4’s 4x Multi Frame Generation on RTX 50-series hardware widened that gap further when it launched, and DLSS 5’s 6x ratio in NBA 2K27 widens it again. Each generation has followed a similar arc: a headline number built on the maximum MFG multiplier, followed within days by independent testing that isolates the real render rate underneath it. What is new with DLSS 5 is the scale of the frame rate cost from the rendering technique itself, at 50-60%, compared with earlier DLSS versions where the underlying render pass was largely unaffected and only the interpolation stage added overhead.

Frequently Asked Questions

When did DLSS 5 launch in NBA 2K27?

DLSS 5 went live in NBA 2K27 on September 3, 2026, alongside a matching NVIDIA Game Ready driver, according to NVIDIA’s official GeForce announcement.

Which GPUs support DLSS 5 in NBA 2K27?

DLSS 5 requires an RTX 50-series GPU. NVIDIA’s own performance charts and independent testing currently cover the RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070 by name. The RTX 5060 Ti and RTX 5060 are expected to meet the hardware requirements based on NVIDIA’s general RTX 50-series guidance, but no independently published NBA 2K27 benchmark for either card exists yet.

Does DLSS 5 work on RTX 40-series cards?

No. NBA 2K27’s DLSS 5 implementation is limited to RTX 50-series GPUs, laptops, and GeForce NOW at launch, per NVIDIA and PCMag’s coverage of the rollout. NVIDIA has separately confirmed a patched, more limited version of DLSS 5 for RTX 4000-series cards in other titles, but that support does not extend to NBA 2K27.

Is the 370 FPS number on the RTX 5090 real?

It is a real displayed output figure, but it reflects Multi Frame Generation inserting several AI-generated frames for every frame the GPU actually renders, at ratios as high as 6x. Analysis of NVIDIA’s own chart suggests the true base render rate behind that number is closer to 62 FPS at 4K Ultra with ray tracing.

Does DLSS 5 hurt performance more at 4K or 1080p?

The performance cost is significantly larger at 4K. Independent RTX 5090 testing found roughly a 57-62% frame rate loss at 4K versus roughly 26% at 1080p, largely because NBA 2K27 becomes CPU-bound at lower resolutions, shrinking the share of the frame budget DLSS 5’s neural rendering pass can eat into.

How much more power does DLSS 5 use?

ComputerBase’s testing measured the RTX 5090 drawing roughly 200 more watts at 1440p (299W to 500W) and roughly 140 more watts at 4K (417W to 561W) when DLSS 5 was enabled versus disabled.

Should I turn DLSS 5 on in NBA 2K27?

That depends on your priorities and your GPU. On an RTX 5090 at 1080p, the frame rate cost is modest enough that the visual upgrade is likely worth it for most players. At native 4K on any current RTX 50-series card, the frame rate and power cost are steep enough that many players will prefer running DLSS 5 off, or pairing it with a lower resolution, especially for competitive play where input responsiveness matters more than lighting fidelity.

Will DLSS 5 come to more games soon?

NVIDIA has not published a full game roadmap for DLSS 5 beyond NBA 2K27 at launch. Given the pattern set by prior DLSS generations, additional titles are likely to be announced in the coming months, though specific games and dates have not been confirmed.

Related Coverage

Elias Virtanen

Elias Virtanen

Cybersecurity Analyst

Elias Virtanen is the Cybersecurity Analyst at Tech Insider, bringing hands-on expertise from his background in penetration testing and security consulting. He previously worked as a security researcher at F-Secure in Helsinki, where he focused on threat intelligence and vulnerability disclosure. Elias covers ransomware trends, zero-trust architecture, and the evolving regulatory landscape including NIS2 and the EU Cyber Resilience Act. He holds a CISSP certification and an MSc in Information Security from Aalto University.

View all articles