Nvidia RTX 50 Super leak points to up to 24GB VRAM and 415W TGP—but launch is reportedly on hold

Verdict: Leaks point to three possible desktop refresh cards: an RTX 5080 Super with 24GB of GDDR7 and a 415W Total Graphics Power (TGP), an RTX 5070 Ti Super with 24GB, and an RTX 5070 Super with 18GB. The configurations are technically plausible because newer 24Gb GDDR7 memory devices have a 3GB-per-chip density. However, Nvidia has announced none of these products. As of August 10, 2026, the latest report says cards have reportedly reached at least one board partner but that Nvidia has put the launch on hold because 3GB GDDR7 is substantially more expensive than existing 2GB devices. There is no confirmed release date, MSRP, final specification sheet, or independent performance testing.

The rumored RTX 50 Super cards could solve the current generation’s most obvious weakness—limited VRAM—but they may also bring higher power targets and higher prices. Buyers should treat the memory configurations as credible leaks, not as finished Nvidia specifications.

Rumored RTX 50 Super specifications

The most consistently repeated leak describes three desktop models. The Super figures below are rumored; the comparison figures for the existing RTX 50 cards are official Nvidia specifications.

#	Preview	Product	Price
1		VIPERA NVIDIA GeForce RTX 4090 Founders Edition Graphic Card	$4,440.00	Buy on Amazon
2		ASUS TUF Gaming GeForce RTX 5070 12GB GDDR7 OC EditionGaming Graphics Card	$937.39	Buy on Amazon

This lineup comes from repeated leaker and media reports, including the original RTX 5070 Super claims and later technical interpretations from Tom’s Hardware and PCWorld. A reported Seasonic calculator listing separately exposed the 415W figure for the 5080 Super. None of this is an Nvidia confirmation.

What would change from the current cards?

• The rumored RTX 5080 Super reportedly uses the same 10,752 CUDA-core configuration as the RTX 5080, but raises memory capacity from 16GB to 24GB, memory speed from 30Gbps to 32Gbps, and TGP from 360W to 415W. Reports also associate it with the same GB203 silicon, although that detail remains unconfirmed.
• The rumored RTX 5070 Ti Super reportedly retains the RTX 5070 Ti’s 8,960 CUDA cores and 28Gbps memory while increasing capacity from 16GB to 24GB and TGP from 300W to 350W.
• The rumored RTX 5070 Super reportedly moves from 6,144 to 6,400 CUDA cores, from 12GB to 18GB, and from 250W to 275W while retaining a 192-bit bus and 28Gbps memory.

One report used the figure 10,753 CUDA cores for the alleged 5080 Super. Nvidia’s official Blackwell documentation lists 10,752 for the RTX 5080, as does the current product information, so 10,753 appears to be an isolated error or typo rather than a specification to rely on. See Nvidia’s Blackwell architecture documentation for the official figure.

#1 Best Overall

VIPERA NVIDIA GeForce RTX 4090 Founders Edition Graphic Card

• 16,384 NVIDIA CUDA Cores
• Supports 4K 120Hz HDR, 8K 60Hz HDR and variable refresh rate as indicated in HDMI 2.1A
• New streaming multiprocessors: up to 2x power and power efficiency
• Fourth generation tensor cores: up to 2x AI power
• Third-generation RT cores: up to 2x ray tracing performance

$4,440.00

Buy on Amazon

Why 24GB GDDR7 makes these capacities plausible

The central technical detail is the difference between 24Gb memory devices and 24GB of graphics memory. A 24Gb device contains 24 gigabits, which equals 3GB—not 24GB.

Samsung has announced 24Gb GDDR7 devices, and Micron’s public GDDR7 catalog lists 24Gb parts at 28GT/s and 32GT/s. Nvidia also already ships a 24GB GDDR7 configuration in the RTX 5090 Laptop GPU and a 96GB GDDR7 configuration in the professional RTX PRO 6000 Blackwell, demonstrating that high-density GDDR7 is not purely theoretical. Those products do not, however, confirm a desktop RTX 50 Super launch.

The rumored capacities fit the existing bus widths without requiring a wider memory interface:

• Eight 3GB devices on a 256-bit bus: 24GB. This is the reported configuration for both the RTX 5080 Super and RTX 5070 Ti Super.
• Six 3GB devices on a 192-bit bus: 18GB. This matches the reported RTX 5070 Super configuration.

In other words, Nvidia could reportedly replace the existing 2GB memory devices with 3GB devices while keeping the same number of chips and the same memory-controller arrangement. That explains why capacities of 18GB and 24GB are plausible even though they are unusual for consumer graphics cards. It is evidence of feasibility, not proof of product intent.

The commercial problem is cost. On July 17, 2026, VideoCardz reported, citing an anonymous board-partner source, that 3GB GDDR7 modules could cost approximately $60–$70 each compared with roughly $20 for 2GB devices. Those prices cannot be independently verified against public supplier price lists, but they provide a possible explanation for why a technically straightforward memory upgrade could become difficult to launch at an attractive price.

Capacity, bandwidth, compute, and software are different things

More VRAM is useful, but it is not a universal performance upgrade. Four separate factors affect a GPU workload:

1. Capacity: how much data can fit in the framebuffer.
2. Bandwidth: how quickly data can move between the GPU and memory.
3. Compute: shader, ray-tracing, and Tensor processing capability.
4. Software features: technologies such as DLSS, frame generation, ray reconstruction, and driver optimizations.

The rumored Super cards directly improve capacity. The 5080 Super may also gain a modest theoretical bandwidth increase. But more capacity alone does not make the CUDA cores execute shaders faster, and it does not guarantee a particular frame-rate increase.

Rumored bandwidth calculations

Memory bandwidth is calculated as memory speed multiplied by bus width, divided by eight:

Bandwidth = memory speed × bus width ÷ 8

The current RTX 5080 specification comes from Nvidia’s Blackwell architecture PDF, while Nvidia lists the current RTX 5070 Ti and RTX 5070 figures on its RTX 5070 family page. Under the leak, the 5080 Super would gain about 6.7% theoretical bandwidth—not 50%. The 50% number applies to memory capacity: 16GB to 24GB, or 12GB to 18GB.

That distinction matters. If a game or application already fits within 16GB, moving to 24GB may have little or no effect on average frame rates. If it exceeds the available framebuffer, additional capacity can prevent texture reductions, stuttering, asset streaming problems, or severe performance drops. The benefit depends on the workload and settings.

Where the extra VRAM could help

The strongest case for these rumored cards is not a guaranteed increase in ordinary 1440p gaming performance. It is additional headroom for workloads that are already approaching the limits of current cards.

• Native 4K gaming: very high-resolution textures, large asset packs, and extreme settings can consume substantially more memory than typical 1080p or 1440p play.
• Ray tracing and path tracing: complex lighting data and high-resolution assets can increase framebuffer pressure.
• Heavily modded games: high-resolution texture packs and expansive worlds may exceed the capacity of a 12GB or 16GB card.
• Local AI: model weights, context windows, caches, and auxiliary data must fit within available VRAM or be moved to system memory, often at a substantial performance cost.
• Video, 3D, and content creation: large timelines, scenes, textures, simulations, and renders can have capacity requirements that differ from games.

Independent RTX 5080 testing has found situations where 16GB can be exhausted at 4K with extreme settings. PCGamesN’s RTX 5080 review provides one example, while TechSpot’s VRAM testing shows why memory use varies greatly by game, resolution, texture settings, and measurement method. These results demonstrate a possible limitation; they do not mean every game needs 24GB or that a Super card will be faster in every title.

There are no independent benchmarks for an unreleased RTX 5080 Super, RTX 5070 Ti Super, or RTX 5070 Super. Reported performance gains of approximately 9–16% for the 5080 Super, 7–11% for the 5070 Ti Super, and 8–12% for the 5070 Super originated with a Moore’s Law Is Dead-sourced report and remain unverified. They should not be treated as expected review results.

A 415W TGP is significant, but it is not a PSU recommendation

The rumored 415W figure is a board-level Total Graphics Power target for the RTX 5080 Super. It is not the power consumption of the complete computer, and it is not a claim that a 415W power supply would be sufficient.

Nvidia officially rates the current RTX 5080 at 360W TGP and lists 850W required system power in its technical documentation. The rumored 5080 Super would therefore represent an additional 55W, or approximately 15.3%. The other reported increases are approximately 16.7% for the RTX 5070 Ti Super—350W compared with 300W—and 10% for the RTX 5070 Super—275W compared with 250W.

The extra power could reflect faster memory on the reported 5080 Super, higher clock or voltage targets, a larger board power limit, board-level changes, or a combination of factors. The leak does not establish the cause.

Rank #2

ASUS TUF Gaming GeForce RTX 5070 12GB GDDR7 OC EditionGaming Graphics Card

• Powered by the NVIDIA Blackwell architecture and DLSS 4 OC mode: 2640MHz/Default mode: 2610MHz (Boost Clock)
• Military-grade components deliver rock-solid power and longer lifespan for ultimate durability
• Protective PCB coating helps protect against short circuits caused by moisture, dust, or debris
• 3.125-slot design with massive fin array optimized for airflow from three Axial-tech fans
• Phase-change GPU thermal pad helps ensure optimal thermal performance and longevity, outlasting traditional thermal paste for graphics cards under heavy loads

$937.39

Buy on Amazon

For a future buyer, the relevant system-design questions will include:

• Is the PSU powerful enough for the complete system, including the CPU and storage?
• Does it support an appropriate ATX 3.x design and transient-load behavior?
• Does it provide a native 12V-2×6 connector, or will an adapter be required?
• Are the cables correctly installed and free from sharp bends near the connector?
• Can the case provide enough airflow for a potentially higher-power card?
• Will the card’s cooler fit without blocking adjacent slots or interfering with a radiator?

Do not choose a specific 850W or 1,000W PSU for the rumored cards solely from the 415W number. Nvidia and the board partners have not published the final system-power or connector requirements. Once a product is official, buyers should use Nvidia’s specifications, the specific AIB model’s requirements, and a reputable power-supply calculator such as Seasonic’s calculator. Seasonic describes its calculator output as a minimum system recommendation that accounts for other components and headroom; its listing of a product name does not prove Nvidia’s specifications.

How strong is the evidence?

The evidence is not a single confirmed leak. It is a chain of claims with different levels of reliability:

1. Initial leaker claim: On June 29, 2025, Kopite7kimi reportedly posted details for an RTX 5070 Super with 18GB, a 192-bit interface, 6,400 CUDA cores, and a 275W power target. A Lowyat summary recorded the claim. This is the origin of the 5070 Super specification set, not an Nvidia announcement.
2. Secondary technical reporting: Tom’s Hardware and PCWorld subsequently reported a broader three-card lineup, adding the alleged 24GB RTX 5080 Super and RTX 5070 Ti Super. Repetition across outlets makes the story more notable, but repeating a source does not independently verify every number.
3. Seasonic database entry: The reported appearance of the RTX 5080 Super in Seasonic’s calculator supports the existence of at least an internal or preliminary product entry and gives the 415W number additional context. Databases can contain preliminary, stale, or incorrect data, so this is not official confirmation of the GPU’s memory, cores, release date, or final power requirement.
4. Timing reports: BenchLife reported a possible early-2027 window, reportedly around CES 2027. That is a rumor-based estimate, not a release commitment.
5. Board-partner report: VideoCardz reported in July 2026 that completed or nearly completed cards had reached at least one AIB, while Nvidia allegedly told partners to hold the launch because 3GB GDDR7 was too expensive. This would be stronger evidence that engineering work progressed, but it remains an anonymous-source report and does not prove that Nvidia will ship the cards.

The most important distinction is between hardware reportedly existing and a retail product being guaranteed. A board partner can have engineering or production samples even if a launch is delayed, specifications change, pricing becomes unattractive, or the product is canceled.

RTX 50 Super launch status: the timeline has changed repeatedly

The earlier October 2025 and Q4 2025 predictions are no longer relevant launch guidance. Nvidia also did not announce the cards at CES 2026, and the alleged hold means CES 2027 itself should not be treated as certain. As of August 10, 2026, no Nvidia launch event, MSRP, final specification sheet, or firm release date is public.

Should you buy an RTX 5080, 5070 Ti, or 5070 now?

The answer depends more on your current workload than on the existence of the leak.

Buy a current card if:

• Your existing GPU is failing or cannot meet an immediate gaming or work requirement.
• You can find an RTX 5080, RTX 5070 Ti, or RTX 5070 at a price you consider acceptable.
• You play primarily at 1440p or use settings that do not approach the current card’s VRAM limit.
• You need a known quantity rather than an uncertain product with no confirmed date.
• Waiting would force you to miss a project deadline, gaming release, or other time-sensitive need.

Waiting makes more sense if:

• You already have a suitable GPU and can comfortably tolerate an indefinite delay.
• Your main reason for upgrading is VRAM capacity rather than an immediate need for more shader or ray-tracing performance.
• You regularly use native or heavily modded 4K, path tracing, large creative projects, or local AI workflows that are constrained by 12GB or 16GB.
• You are willing to reassess after Nvidia publishes real prices, benchmarks, power requirements, and availability.

Do not pay a large premium for an existing RTX 50 card on the assumption that the Super cards are definitely arriving soon. The reported memory costs could produce higher launch prices than old rumors suggested, and the cards could still be delayed, changed, or canceled. Conversely, do not postpone an urgent purchase solely because an unannounced refresh might eventually offer more VRAM.

What remains unknown

Even if the current leak is broadly accurate, the following details remain unverified:

• MSRP and real retail pricing.
• Launch date, regions, quantities, and availability.
• Clock speeds and boost behavior.
• Final memory suppliers, memory bins, and sustained speeds.
• Founders Edition designs.
• Board dimensions, cooler size, slot thickness, and weight.
• Connector requirements and adapter details.
• Official system-power recommendations.
• Independent gaming, creator, and AI performance.
• Whether Nvidia will use the Super branding at all.
• Whether additional rumored products, including an RTX 5060 with 12GB or an RTX 5050 with 9GB, will launch.

The possible RTX 5060 Super and RTX 5050 configurations belong in the low-confidence category. They have appeared in rumor coverage, but the most consistently repeated and technically coherent report remains the three-card 5080 Super, 5070 Ti Super, and 5070 Super lineup.

Bottom line for buyers

The RTX 50 Super rumor is technically credible because 24Gb GDDR7 devices can produce 18GB and 24GB graphics cards on the same 192-bit and 256-bit bus widths used by the current models. The alleged 415W RTX 5080 Super power target is also supported by a reported Seasonic calculator entry, although that entry is not an Nvidia specification.

What is not established is more important for a purchase decision: performance, price, connectors, dimensions, and release timing. The latest report says cards have reached a board partner but that Nvidia is holding the launch over the cost of 3GB GDDR7 modules. Treat the Super lineup as a possible future option for buyers who need more framebuffer capacity—not as a guaranteed near-term replacement for the current RTX 50 cards.

Frequently Asked Questions

Has Nvidia confirmed the RTX 50 Super lineup?

No. As of August 10, 2026, Nvidia has not announced an RTX 5080 Super, RTX 5070 Ti Super, or RTX 5070 Super. The specifications, pricing, branding, and release timing remain based on leaks and third-party reports.

Will 24GB of VRAM make the RTX 5080 Super 50% faster?

No. The reported 50% increase is in memory capacity, from 16GB to 24GB. More capacity can prevent slowdowns when a workload exceeds the existing framebuffer, but it does not directly increase shader, ray-tracing, or Tensor performance. Independent benchmarks for the rumored card do not exist.

Does a rumored 415W TGP mean the RTX 5080 Super needs a 415W power supply?

No. TGP describes the graphics card’s board-level power target, not the complete PC or the minimum PSU. A final PSU recommendation will depend on the GPU, CPU, connectors, transient behavior, and the specific board design. Nvidia has not published that requirement for the rumored card.

Is CES 2027 the confirmed RTX 50 Super launch date?

No. Early 2027, potentially around CES 2027, is a reported timing estimate. The latest report says the launch is on hold, and no replacement date has been announced.

The Bottom Line

More VRAM is the most credible part of the RTX 50 Super rumor; the launch is not. The reported 24GB/24GB/18GB capacities are technically plausible with 3GB GDDR7 devices, but there is no confirmed price, date, performance, or final design. Buy a current card when you have a real need and an acceptable deal; wait only if additional VRAM is essential and you can tolerate the possibility of another delay.

Quick Recap

Bestseller No. 1

VIPERA NVIDIA GeForce RTX 4090 Founders Edition Graphic Card

16,384 NVIDIA CUDA Cores; Supports 4K 120Hz HDR, 8K 60Hz HDR and variable refresh rate as indicated in HDMI 2.1A

$4,440.00

Bestseller No. 2

ASUS TUF Gaming GeForce RTX 5070 12GB GDDR7 OC EditionGaming Graphics Card

3.125-slot design with massive fin array optimized for airflow from three Axial-tech fans; Auto-Extreme precision automated manufacturing helps ensure higher reliability

$937.39

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.