Nvidia has confirmed that DLSS 5 is coming to GeForce RTX 40-series graphics cards through an official update rather than the unauthorized patches that have circulated online since late August 2026. But the company’s own language on the rollout, delivered to The Verge on September 4, 2026, makes clear that RTX 40 owners will not get the same access, timeline, or feature set as buyers of the newer RTX 50-series cards. The announcement resolves one question raised at DLSS 5’s September 3 launch in NBA 2K27, and immediately raises several more about how much say gamers actually get over the technology running on their own hardware. More than a week later, as of September 14, 2026, Nvidia has still not named an exact release date or driver version for the RTX 40 rollout.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Nvidia actually confirmed on September 4
Nvidia spokesperson Rajal Maharaj told The Verge that RTX 40-series support for DLSS 5 is real and coming, just not yet. “Our current focus is on optimizing performance for the GeForce RTX 50 Series with model updates expected later this fall,” Maharaj said in a statement to the outlet. “Once RTX 50 Series performance is more fully tuned, we plan to work on expanding official support to the GeForce RTX 40 Series.” A separate line attributed to Maharaj put it more bluntly: “Right now the focus is of course on 50-series. One step at a time.”
That statement is a reversal, or at least a softening, of Nvidia’s original position. DLSS 5 launched on September 3, 2026 tied to NBA 2K27, and at that point the company described the feature as available only on RTX 50-series desktop and laptop GPUs plus GeForce Now cloud streaming instances. RTX 40 and RTX 30 owners were told, in effect, to wait or upgrade. Less than 48 hours later, Nvidia walked that framing back just enough to confirm a future RTX 40 rollout, while giving no firm date beyond “later this fall” and no commitment at all for RTX 30-series or older Turing and Ampere cards.
The sequencing matters. Nvidia is not saying RTX 40 cards will get DLSS 5 at the same time, with the same feature completeness, or on the same schedule as RTX 50 cards. It is saying RTX 50 comes first, gets tuned first, and only after that work is “more fully tuned” does RTX 40 support begin. For an install base that vastly outnumbers RTX 50 owners, that is a meaningful gap between what Nvidia calls universal DLSS 5 support and what actually ships.
The control gamers still don’t have
The most important detail buried in Nvidia’s September 4 statement is what it does not promise: user-facing control over which parts of DLSS 5 run on which hardware. DLSS 5’s headline addition is neural rendering, a departure from the frame-by-frame upscaling and generation model of DLSS 3 and DLSS 4. Coverage of Nvidia’s own materials notes that neural rendering is currently available only to RTX 50 owners, with Nvidia declining to say whether or when RTX 30 or RTX 20 cards will ever get it at all.
That is a familiar shape for Nvidia’s upscaling roadmap. DLSS 3’s Frame Generation feature launched tied exclusively to RTX 40-series hardware, using the Optical Flow Accelerator built into Ada Lovelace silicon. DLSS 4’s Multi Frame Generation then launched tied exclusively to RTX 50-series Blackwell hardware. Each generational leap in DLSS has arrived with a headline feature gamers on the previous generation could not access no matter how they configured their settings, and DLSS 5’s neural rendering pipeline looks positioned to repeat that pattern a third time. Nvidia has not published a features matrix that lets an RTX 40 owner toggle individual DLSS 5 components on or off; the company controls that gate entirely, on its own timeline, feature by feature.
What makes the September 4 statement notable is not that Nvidia drew a line, but that the line moved within 48 hours of launch, without Nvidia fully explaining why RTX 40 cards, which shipped with dedicated Tensor Cores and the same general class of AI acceleration hardware used for DLSS 3, were excluded from day one in the first place. The most plausible explanation, based on early testing of unofficial builds, is performance, not raw compatibility.
What the unofficial patches already showed
Before Nvidia’s official statement, modders had already forced DLSS 5 to run on RTX 40 and even RTX 30-series cards, and the results explain a lot about Nvidia’s caution. Reporting on those unofficial builds found render latency exploding on RTX 30-series hardware when the neural rendering path was engaged, with numbers reportedly jumping from around 29 milliseconds in a standard render path to roughly 3,326 milliseconds once neural rendering activated. In the worst recorded cases, framerate on that older Ampere hardware collapsed to around 1 frame per second. Separate figures describe a standalone neural-rendering evaluation step taking roughly 1.2 seconds per pass on unsupported cards.
Those numbers are not from Nvidia and were never presented as representative of a finished, optimized build. They come from community experimentation on hardware Nvidia has not tuned the technology for, and reports note the modder responsible did not always have every card generation available to test directly. But the scale of the slowdown, roughly two orders of magnitude worse render latency in the RTX 30 case, gives Nvidia’s “one step at a time” framing more weight than a simple product-segmentation explanation would. If neural rendering genuinely runs at 1 fps on older Tensor Core hardware without heavy optimization, gating the feature by GPU generation looks less like an artificial lock and more like triage.
That does not fully explain why RTX 40 cards, which sit one generation closer to RTX 50 than RTX 30 does, were also excluded at launch. Nvidia’s own statement frames the RTX 40 delay as a matter of engineering sequencing, optimizing RTX 50 first, then moving down the stack, rather than a hardware limitation specific to Ada Lovelace. That distinction, sequencing versus hard capability limit, is exactly the kind of detail Nvidia has left vague, and it is the detail that determines whether RTX 40 owners are looking at a few weeks of delay or a permanently reduced feature set once support does arrive.
DLSS 5 feature access by GPU generation
| GPU Generation | DLSS 5 Status as of Sept. 14, 2026 | Neural Rendering Access | Path to Support |
|---|---|---|---|
| RTX 50-series (Blackwell) | Officially supported at launch (Sept. 3) | Yes, full access | Already shipping via NBA 2K27 and GeForce Now |
| RTX 40-series (Ada Lovelace) | Officially confirmed, not yet live | No, pending Nvidia rollout | Nvidia says “later this fall,” after RTX 50 tuning |
| RTX 30-series (Ampere) | No official commitment | No; unofficial builds show severe slowdowns | Undisclosed; Nvidia has not stated plans |
| RTX 20-series (Turing) | No official commitment | No | Undisclosed; Nvidia has not stated plans |
| GeForce Now (cloud) | Officially supported at launch | Yes, via cloud-side RTX 50 hardware | Already live regardless of the subscriber’s own GPU |
That table captures the core tension in Nvidia’s rollout strategy. A GeForce Now subscriber streaming from Nvidia’s own RTX 50-series cloud hardware gets full DLSS 5 access today, regardless of what GPU sits in their own PC. An RTX 40 owner who paid full price for a card marketed as a premium gaming GPU as recently as this year does not, and has no firm date for when that changes. That asymmetry, cloud parity now, local parity later or never, is likely to be the more durable story here, well past the news cycle around Nvidia’s September 4 statement.
Historical pattern: how DLSS access has narrowed with every generation
Nvidia’s DLSS roadmap shows a consistent trend: each major DLSS version keeps broad upscaling support backward-compatible while gating its single headline feature to the newest hardware generation. DLSS 3’s Frame Generation required RTX 40-series or newer. DLSS 4’s Multi Frame Generation required RTX 50-series. DLSS 5’s neural rendering pipeline is following the identical shape, restricted to RTX 50 at launch with RTX 40 promised “later” and everything older left unaddressed.
The pattern is commercially rational for Nvidia. Each new headline feature gives buyers of the current-generation card a reason to feel their purchase was worth the premium, and gives owners of older cards a reason to consider upgrading rather than waiting. What is different about the DLSS 5 rollout, compared to the DLSS 3 and DLSS 4 launches, is the public reversal. Nvidia did not simply announce RTX 50-exclusive neural rendering and hold that line; it announced RTX 50-exclusive support, faced two days of community testing and coverage showing partial functionality already existed on older cards through unofficial means, and then confirmed an RTX 40 expansion without giving ground on RTX 30 or offering a specific date.
| DLSS Generation | Headline Feature | Minimum GPU Requirement | Backward Compatibility for Base Upscaling |
|---|---|---|---|
| DLSS 3 | Frame Generation | RTX 40-series (Ada Lovelace) | Yes, DLSS Super Resolution works on RTX 20/30/40 |
| DLSS 4 | Multi Frame Generation | RTX 50-series (Blackwell) | Yes, transformer-based Super Resolution rolled out to RTX 20-40 |
| DLSS 5 | Neural Rendering | RTX 50-series at launch; RTX 40 confirmed, undated | Unclear; RTX 30/20 status undisclosed as of Sept. 14, 2026 |
Why Nvidia is moving cautiously on older Tensor Cores
Neural rendering is a heavier computational workload than the convolutional and transformer-based models that powered earlier DLSS versions. Instead of processing a finished or near-finished frame, reports describe the neural rendering approach as evaluating rendering decisions earlier in the pipeline, which requires running larger AI models more frequently and with tighter latency budgets. RTX 50-series Blackwell cards ship with Nvidia’s newest generation of Tensor Cores, built specifically with this class of workload in mind. RTX 40-series Ada Lovelace cards have capable but earlier-generation Tensor Cores, and RTX 30-series Ampere cards are a full two architectural generations removed from Blackwell.
That gap in raw AI throughput is the most likely technical reason behind the staged rollout, and it lines up with the render-latency figures from unofficial RTX 30 testing. Whether the same bottleneck applies as severely to RTX 40 cards is the open question Nvidia’s statement does not resolve. Ada Lovelace’s Tensor Cores are meaningfully faster than Ampere’s, so it is plausible RTX 40 support becomes viable with far less optimization work than RTX 30 would require, which would explain why Nvidia named a fall timeline for RTX 40 while leaving RTX 30 entirely open-ended.
Market and competitive context
Nvidia’s staged rollout lands at a moment when AMD’s FSR and Intel’s XeSS are both positioning broader hardware compatibility as a selling point against Nvidia’s tiered approach. Where Nvidia gates its most advanced upscaling and frame-generation features to the newest silicon generation by generation, AMD has historically emphasized that FSR works across a wider spread of GPU vendors and generations, including non-RTX and non-AMD hardware in many implementations. That contrast does not automatically translate into image-quality parity, DLSS has generally been rated ahead of FSR in independent testing throughout 2025 and 2026, but it does give AMD and Intel a straightforward marketing angle: buy our card, or use our upscaler, and you are not waiting on a staged rollout schedule set by someone else.
For Nvidia, the calculus is different. The company holds a commanding share of the discrete GPU market, and its data center and AI infrastructure business has increasingly overshadowed gaming GPU revenue in its own earnings reporting this year. That gives Nvidia less immediate financial pressure to rush DLSS 5 support to older gaming cards, and more incentive to use headline features as an upgrade lever for the RTX 50 lineup specifically. The risk is reputational rather than financial in the short term: a large, vocal base of RTX 40 owners who feel they were sold an unfinished promise at launch, only to see the feature list expand once the initial narrower framing drew criticism.
How this compares with the leaked and patched builds
The gap between what modders demonstrated in late August 2026 and what Nvidia is officially shipping is instructive. Unofficial patches proved DLSS 5’s code path could technically execute on RTX 40 and RTX 30 hardware weeks before Nvidia’s own confirmation. That is a meaningfully different claim than proving the feature runs well on that hardware. The render-latency spikes recorded on RTX 30 in particular suggest the unofficial builds were running an unoptimized, effectively unsupported code path, not a preview of what a tuned official release will look like.
What the leaks did accomplish was competitive and reputational pressure. Once it became public that DLSS 5 was technically reachable on older cards, even in a broken state, Nvidia’s original “RTX 50 only” framing looked more like a business decision than a hardware limitation. The September 4 statement to The Verge reads, in that light, less like a scheduled roadmap update and more like a response to that pressure, an acknowledgment that RTX 40 support was always coming but had not yet been publicly committed to a timeline until scrutiny forced the issue.
What RTX 40 and RTX 30 owners can expect this fall
Based on Nvidia’s own statement, RTX 40-series owners should expect an official DLSS 5 update sometime in the fall 2026 window, following further performance tuning on RTX 50-series hardware. Nvidia has not said whether that update will include the full neural rendering feature set or a scaled-down version optimized for Ada Lovelace’s Tensor Core throughput. Games that support DLSS 5, currently limited to NBA 2K27 at launch, will need per-title updates to enable RTX 40 support once Nvidia ships the driver and SDK-level changes, meaning the rollout will likely feel gradual rather than a single flag flipping on for every supported title simultaneously. Nvidia has not named a driver version or build number for the RTX 40 update, and as of September 14, 2026 the company has not narrowed “later this fall” to a specific date.
RTX 30 and RTX 20 owners are in a less certain position. Nvidia has made no statement, official or unofficial, about a support timeline for those generations, and the performance data from unofficial testing suggests a straightforward driver update alone would not be enough to make neural rendering usable on that hardware without significant additional optimization work. Owners of those cards should not expect DLSS 5 access in 2026, and there is no confirmed indication it is coming in any form.
Analysis: what “official support” actually buys gamers
The distinction Nvidia is drawing, between unofficial community patches and official, supported DLSS 5 access, is really a distinction about who controls the feature’s rollout pace, not about whether the underlying hardware is capable. Modders showed weeks ago that DLSS 5’s code could be coaxed onto RTX 40 and RTX 30 silicon. What official support adds is driver-level integration, per-game SDK support, and, critically, Nvidia’s own decision about which sub-features are enabled on which cards and when. That is the “full control” gamers still lack: an RTX 40 owner cannot simply flip on neural rendering today the way Nvidia’s own marketing around DLSS has long implied the technology works, tunable per user, per game, per preference. Instead, access is gated centrally, on Nvidia’s schedule, for reasons the company has only partially explained.
That model is not unusual in the GPU industry, but it does clash with how Nvidia has marketed DLSS technology broadly, as an ecosystem benefit that rewards staying within the GeForce lineup regardless of which specific card a gamer owns. The September 4 statement narrows that promise considerably: staying within the GeForce lineup gets you eventual access, on a schedule Nvidia sets unilaterally, with no public commitment on which features arrive in what order for RTX 40, and no commitment at all below that tier.
Five predictions for the DLSS 5 rollout through early 2027
- Nvidia will ship a scoped-down version of DLSS 5 for RTX 40-series cards in Q4 2026, likely trading some neural rendering fidelity or frame-rate ceiling for compatibility, rather than a feature-complete port of the RTX 50 implementation.
- RTX 30-series support, if it comes at all, will not arrive before mid-2027, and Nvidia may never officially extend full neural rendering to Ampere hardware given the render-latency figures already observed in unofficial testing.
- AMD and Intel will use the staged DLSS 5 rollout in marketing materials this fall, emphasizing broader day-one hardware compatibility for FSR and XeSS updates as a contrast to Nvidia’s generation-gated approach.
- Additional games beyond NBA 2K27 will add DLSS 5 support before the RTX 40 rollout completes, meaning early adopters on RTX 50 hardware will have a widening library advantage over RTX 40 owners for at least one more full quarter.
- Community pressure and continued unofficial patch testing will keep surfacing performance data on older cards, which will likely keep shaping Nvidia’s public statements the same way it appears to have shaped the September 4 announcement.
The bigger picture for Nvidia’s GPU upgrade cycle
DLSS 5’s staged rollout arrives during a year in which Nvidia’s gaming GPU business has increasingly played second fiddle to its data center and AI infrastructure revenue. That context matters for how to read this decision. A company under pressure to sell more gaming GPUs might rush broad DLSS 5 compatibility to reassure a wide installed base. A company whose growth is dominated elsewhere has more room to treat gaming feature rollouts as a slower, tiered process, prioritizing the newest and highest-margin hardware first. Nvidia’s September 4 statement fits the second pattern far more closely than the first.
For gamers, the practical takeaway is that “coming to older GPUs” and “coming with full control” are two different promises, and Nvidia has only made the first one. RTX 40 owners now have official confirmation their cards will eventually run DLSS 5’s neural rendering pipeline. They do not have a date, a confirmed feature list, or any indication that Nvidia intends to hand over the kind of granular, user-level control over the technology that would make “official support” functionally different from a better-optimized version of the same patches modders already built.
Frequently asked questions
Is DLSS 5 officially available on RTX 40-series GPUs right now?
No. As of September 14, 2026, DLSS 5 is still officially available only on RTX 50-series GPUs and through GeForce Now. Nvidia has confirmed RTX 40-series support is coming but has not given an exact date or driver version beyond “later this fall.”
Will RTX 30-series cards ever get DLSS 5?
Nvidia has not made any official commitment for RTX 30-series or older GPUs. Unofficial testing on RTX 30 hardware showed severe performance problems, including render latency spiking into the thousands of milliseconds, which suggests full support may not be technically practical without significant optimization work.
What is neural rendering in DLSS 5?
Neural rendering is DLSS 5’s headline new feature, described in coverage of the launch as a departure from the frame-based upscaling and generation approach of DLSS 3 and DLSS 4. It is currently the sub-feature most tightly restricted to RTX 50-series hardware.
Which game supports DLSS 5 today?
NBA 2K27 was the launch title for DLSS 5, going live on September 3, 2026, with RTX 50-series and GeForce Now support at release.
Can I run DLSS 5 on an RTX 40 card using unofficial patches?
Modders demonstrated that DLSS 5’s code path could be forced to run on RTX 40 and RTX 30 hardware before Nvidia’s official confirmation, but these were unsupported, unoptimized builds, not a preview of the finished official release.
Does DLSS 5 replace DLSS 4’s Multi Frame Generation?
DLSS 5 adds neural rendering as a new capability on top of the existing DLSS upscaling and frame-generation lineage rather than simply replacing it outright, though Nvidia has not published a full technical breakdown of how the two interact on RTX 50 hardware.
Why did Nvidia originally restrict DLSS 5 to RTX 50-series only?
Nvidia’s own statement points to performance tuning: the company said it is focused on optimizing RTX 50-series performance first before expanding official support, rather than citing a hard hardware limitation on older cards.
Does GeForce Now give access to DLSS 5 regardless of my own GPU?
Yes. GeForce Now was included in DLSS 5’s official launch on September 3, 2026, meaning subscribers can access the feature through Nvidia’s cloud-side RTX 50-series hardware even if their local GPU is an older generation.
Related Coverage
- NVIDIA Confirms DLSS 5 for RTX 40 GPUs, RTX 50 First [2026]
- NBA 2K27 DLSS 5 Verdict: 4 of 6 RTX 50 GPUs Proven [2026]
- Nvidia RTX Spark N1X Launches October: 128GB RAM [2026]
- Nvidia DLSS 5: 797 FPS Claim Is Really 133 FPS [2026]
- Nvidia DLSS 5 Tanks FPS 48%, Guzzles 802W [2026]
