Key takeaways
- Check the source’s track record. Some leakers have consistently correct specification details; trade sites that aggregate rumours often repeat earlier posts.
- Look for corroboration from supply-chain data. PSU calculators, shipping manifests and driver strings are stronger evidence than anonymous posts.
- Ignore precise dates without sources. A named month with no cited partner or document is usually a guess.
- Remember that memory economics change. If chip prices fall, a refresh becomes more feasible; if they rise, delays continue.
- Wait for an official specification sheet before treating any number as settled.
As of October 2026, NVIDIA has not announced, priced or dated an RTX 5080 Super or RTX 5070 Super. What exists is a body of leaks and trade reports: a 24GB RTX 5080 Super built on the same 10,752-core GB203 chip as the RTX 5080, an 18GB RTX 5070 Super, and a launch window that has slipped repeatedly, with the most recent reports pointing to early 2027. If you need a GPU now, the current RTX 50 cards are the real choice; the Super refresh is a rumour worth tracking, not a product you can plan around.
This guide separates what is credible from what is speculation, explains why 3GB memory modules are the centre of the story, shows what each rumoured card would change, and gives a decision framework for buy-now versus wait.
Our top picks
As an Amazon Associate we earn from qualifying purchases at no extra cost to you.
What Is Actually Confirmed
Nothing about the Super refresh has been confirmed by NVIDIA. The company did not present the cards at CES 2026, has not published a specification sheet, and has not issued a statement on timing. Every number in this article that refers to a Super card comes from leakers, retail-adjacent listings or trade press citing board-partner sources. Treat each figure as provisional.
| Claim | Source type | Reliability |
|---|---|---|
| RTX 5080 Super uses the same 10,752 CUDA cores as the RTX 5080 | Leaker report, repeated by several outlets | Moderate to high |
| RTX 5080 Super has 24GB of GDDR7 on a 256-bit bus | Leaker reports and PSU-calculator listing | Moderate to high |
| Memory rated around 32 Gbps | Leaker report | Moderate |
| Board power around 415W or higher | Leaks and a PSU-calculator entry | Moderate |
| RTX 5070 Super has 18GB | Leaker report | Moderate |
| Performance gain of 7 to 15 percent over non-Super cards | Single leaker estimate | Low |
| Launch in early 2027 | Industry report citing partners | Low to moderate |
Why 3GB Memory Modules Drive Everything
Current RTX 50 cards use 2GB (16Gb) GDDR7 chips. The Super refresh is widely reported to use 3GB (24Gb) chips, which raise capacity without widening the memory bus. The arithmetic is simple: a 256-bit bus connects to eight memory chips, so eight 2GB chips give 16GB and eight 3GB chips give 24GB. A 192-bit bus connects to six chips, so six 2GB chips give 12GB and six 3GB chips give 18GB. That is exactly why leaks describe 24GB for the 256-bit RTX 5080 and RTX 5070 Ti tier, and 18GB for the 192-bit RTX 5070.
| Card (rumoured Super) | Bus width | Chips | Chip size | Total VRAM |
|---|---|---|---|---|
| RTX 5070 Super | 192-bit | 6 | 3GB | 18GB |
| RTX 5070 Ti Super | 256-bit | 8 | 3GB | 24GB |
| RTX 5080 Super | 256-bit | 8 | 3GB | 24GB |
The same math explains why leaks about a 24GB RTX 5070 or 16GB RTX 5060 never made sense without a wider bus: capacity follows bus width and chip density.
Why the Launch Keeps Slipping
Several threads appear in the reporting, and they reinforce each other.
- Memory cost. Reports claim the 3GB GDDR7 chips cost roughly three times as much as 2GB modules at the time, which would add substantial cost to every card.
- Supply competition. Memory makers have been prioritising chips for AI accelerators, which tightens supply for consumer parts.
- Lack of competitive pressure. Reports indicate no new AMD Radeon desktop cards are expected in 2026, so NVIDIA has less reason to rush a mid-cycle refresh.
- Product readiness. Some reports say the cards are fully developed and have reached partners, which suggests timing, not engineering, is the blocker.
Timeline in reports so far: a late-2025 target gave way to a CES 2026 reveal that did not happen, then a January 2026 report that the refresh was delayed indefinitely, then partner guidance pointing to the third quarter of 2026, and most recently a Q1 2027 target. Sources disagree on whether this is a delay or a quiet shelving, and one hardware site of uncertain reliability is behind some of the most specific dates. Weigh claims accordingly.
What Each Rumoured Card Would Change
RTX 5080 Super
The rumoured configuration keeps the 10,752 CUDA cores and adds 8GB of memory at a higher speed. If the 32 Gbps figure is right, bandwidth rises from the current RTX 5080’s 960 GB/s to about 1,024 GB/s. Board power reportedly climbs above 400W, versus 360W on the standard card. The performance uplift is therefore modest, with the leaker estimates of 7 to 15 percent most likely coming from clocks and bandwidth in memory-bound scenes. The real benefit is capacity: 24GB removes VRAM ceilings at 4K with path tracing and eases local AI workloads.
RTX 5070 Ti Super
Leaks suggest 24GB on the existing 256-bit design. If it materialises, this would be the most interesting card for creators, because it would give 24GB of memory at a mid-high tier.
RTX 5070 Super
The rumoured 18GB variant would lift the mainstream card from 12GB to 18GB, addressing the main criticism of the RTX 5070. Core count details remain unconfirmed in the available reporting, so be cautious about any specific number that circulates.
Current Cards vs Rumoured Refresh
| Card | CUDA cores | VRAM | Bus | Power | Status |
|---|---|---|---|---|---|
| RTX 5070 | 6,144 | 12GB | 192-bit | 250W | Shipping |
| RTX 5070 Super | Not confirmed | 18GB (rumour) | 192-bit | Not confirmed | Unannounced |
| RTX 5070 Ti | 8,960 | 16GB | 256-bit | 300W | Shipping |
| RTX 5080 | 10,752 | 16GB | 256-bit | 360W | Shipping |
| RTX 5080 Super | 10,752 (rumour) | 24GB (rumour) | 256-bit | about 415W (rumour) | Unannounced |
| RTX 5090 | 21,760 | 32GB | 512-bit | 575W | Shipping |
Buy Now or Wait? A Decision Matrix
| Your situation | Recommendation | Reason |
|---|---|---|
| Your GPU is broken or unusable | Buy now | A rumour with no date is not a plan |
| Gaming at 1440p, current card is 4 or more years old | Buy an RTX 5070 or 5070 Ti now | Delay costs months of use |
| Want 24GB for local AI or 4K path tracing | Wait if you can, otherwise consider RTX 5090 | Refresh targets this gap |
| Already own an RTX 5080 | Do not wait | Expected gain is a single-digit to low double-digit percent |
| Price-sensitive buyer | Buy current stock | Refresh cards would likely launch at higher tiers |
| Can tolerate 6 to 12 months of waiting | Watch CES 2027 announcements | Reported target window |
A worked wait-or-buy calculation
Suppose you play about 15 hours a week. Waiting nine months for a card that may not arrive costs roughly 600 hours of gaming on an inferior or broken setup. Even if the refresh arrives on schedule, early supply at launch is typically tight, and the first weeks often favour the higher tiers. Unless you specifically need the extra memory, the expected value of waiting is low. If you do need it, wait only until a credible announcement appears; reports without NVIDIA confirmation have already been wrong several times.
How to Evaluate Future Leaks
- Check the source’s track record. Some leakers have consistently correct specification details; trade sites that aggregate rumours often repeat earlier posts.
- Look for corroboration from supply-chain data. PSU calculators, shipping manifests and driver strings are stronger evidence than anonymous posts.
- Ignore precise dates without sources. A named month with no cited partner or document is usually a guess.
- Remember that memory economics change. If chip prices fall, a refresh becomes more feasible; if they rise, delays continue.
- Wait for an official specification sheet before treating any number as settled.
What to Buy Today If You Cannot Wait
The current lineup already covers each tier. For 1440p at a sensible tier, the RTX 5070 with 12GB is the entry point and the RTX 5070 Ti with 16GB removes memory worries. For 4K, the RTX 5080 handles most titles with upscaling, and the RTX 5090 is the only card with 32GB. Pair any of them with a power supply matched to the board power: 650W for the RTX 5070, 750W for the RTX 5070 Ti, 850W for the RTX 5080 and 1000W for the RTX 5090, per NVIDIA’s recommendations.
If you buy now and a refresh appears later, the resale value of a 16GB card will hold up better than that of a 12GB model, because memory capacity remains the first thing used buyers check.
What Past Super Refreshes Teach Us
NVIDIA has used the Super label before, and the pattern is instructive. The RTX 20 Super cards arrived in mid-2019, roughly ten months after the original RTX 20 series, and moved several models to a larger die for a real performance gain. The RTX 40 Super cards arrived in January 2024, about fifteen months after the RTX 4090 launched, and shifted the RTX 4070, RTX 4070 Ti and RTX 4080 up with more cores at the same or lower launch tiers. In both cases the refresh came with an official announcement at a major event, and the original cards were repositioned or replaced in the stack.
Two lessons follow. First, a refresh usually lands between one year and eighteen months after the original generation, which means the RTX 50 Super window has already passed; the delay is unusual. Second, the best Super cards of past generations improved the value of the mid-range, not the flagship. If the RTX 50 refresh follows that pattern, the 18GB RTX 5070 Super and 24GB RTX 5070 Ti Super would matter more to most buyers than the RTX 5080 Super.
What 24GB Gives You in Practice
A larger memory pool is useful only if your workload can use it. The table below maps common tasks to their typical memory needs.
| Workload | Typical VRAM need | 16GB enough? | 24GB enough? |
|---|---|---|---|
| 1440p gaming, high settings | 8 to 12GB | Yes | Yes |
| 4K gaming with ray tracing | 12 to 16GB | Usually | Yes |
| 4K path tracing with frame generation | 14 to 20GB | Sometimes not | Yes |
| Texture-modded games | 16 to 22GB | No | Usually |
| Local image generation (large models) | 12 to 24GB | Partly | Yes |
| Local language models (quantized, mid-size) | 12 to 24GB | Smaller models | Larger models |
| 8K video editing with effects | 16 to 24GB | Struggles | Comfortable |
For ordinary gaming at 1440p, 24GB is more than any current title needs, so the extra memory is insurance rather than an immediate gain. For AI and content work, the capacity can decide whether a model loads at all.
Power, PSU and Case Implications
If the reported board power of about 415W holds for the RTX 5080 Super, the card would draw roughly 15 percent more than the standard RTX 5080’s 360W. Add a 125W processor and 100W for the rest of the system and you reach about 640W at peak, which suggests an 850W to 1000W power supply for transient headroom. Owners of 750W units would be near the limit.
A 415W card also generates more heat. Case airflow matters more, and the cooler is likely to be a large triple-fan design. Before any future purchase, check three things: the length and thickness of the cooler against your case, the cable type your supply provides, and whether your motherboard slot placement leaves room for a thick card.
Energy cost scales with it. At 350W average during three hours of daily play, the card uses about 383 kWh a year. At 0.15 per kWh, that is about 57 units of currency annually, versus around 49 for a 300W average, a small but real difference.
The Used-Market Angle
A delayed Super series affects second-hand prices. Buyers who want 24GB of memory today have limited choices: the previous-generation flagships and the current RTX 5090. Used high-end cards from earlier generations often have the memory but lack Multi Frame Generation and newer encoders, and their power efficiency is worse. When buying used, ask for proof of working condition, check fan operation and temperatures, and avoid cards with signs of cryptocurrency mining or heavy sustained use. A new-card warranty is worth real value at this tier.
If you plan to sell your current GPU when the refresh arrives, sell before the announcement, not after. Prices for older cards typically soften quickly once a replacement is official.
Signals That Would Confirm a Launch
- An official event invitation or keynote slot from NVIDIA for a trade show.
- Driver releases that add new device IDs, which usually precede launches by weeks.
- Retailer listings with model numbers from major board partners, together with box art leaks.
- Price cuts or stock clearance on current RTX 5070 Ti and RTX 5080 cards, a classic sign of an imminent refresh.
- Review embargo information from partners, which appears days before release.
Until at least two of these appear, assume the cards are not imminent.
How the Delay Affects Existing RTX 50 Cards
Because the refresh is not close, the current RTX 50 line is not being cleared out to make room. Market tiers have stayed steady, and the presence of rumoured upgrades has not driven demand away from existing models. That means you are not likely to benefit from waiting for a discount either. If a surprise announcement does appear, expect a short window in which retailers mark down older stock before supply of the new cards stabilises.
Comparing Rumoured Value Against the RTX 5090
The RTX 5090 is the only shipping GeForce card with more than 16GB, offering 32GB on a 512-bit bus. A 24GB RTX 5080 Super would sit below it in raw throughput, because the flagship has roughly twice the CUDA cores, but would fill a gap for buyers who want more than 16GB without the flagship’s 575W appetite. Per NVIDIA’s published numbers, the RTX 5090 offers 1,792 GB/s of bandwidth, compared with the rumoured 1,024 GB/s of the Super. For workloads that scale with bandwidth, such as large-model inference, the gap is meaningful; for games, the shader count gap dominates.
In short, the refresh would not replace the flagship. It would give the high-end tier a more balanced memory configuration, and for many buyers that is the missing piece of the current lineup.
What Would Change in Your Build Plan
Treat the Super rumours as a reason to future-proof the parts around the GPU, not as a reason to pause. Choose a power supply with at least 850W and a native 12V-2×6 cable, pick a case with generous clearance for a 330 mm or longer card and good front-to-back airflow, and buy a motherboard with a PCIe 5.0 x16 slot. These choices cost little today and make a later swap to any higher-power card painless. If your current plan is a 1440p build around an RTX 5070 or RTX 5070 Ti, none of the rumoured changes would alter your experience in a way that justifies waiting.
Frequently Asked Questions
Has NVIDIA announced the RTX 5080 Super?
No. As of October 2026 there is no official announcement, specification sheet or price. All details come from leaks and industry reports.
Will the RTX 5080 Super really have 24GB of VRAM?
Multiple independent leaks agree on 24GB of GDDR7 on a 256-bit bus built from 3GB chips, which is consistent with how memory capacity is calculated. It remains unconfirmed until NVIDIA publishes the specification.
When is the RTX 50 Super series expected?
The most recent reports suggest early 2027, possibly around CES. Earlier predictions of late 2025 and mid-2026 did not come true, so treat any date as uncertain.
Why would NVIDIA delay the Super cards?
Reports point to the high cost and limited supply of 3GB GDDR7 modules, strong demand for memory from AI hardware, and limited competition from AMD in the desktop segment this year.
Should I wait for the RTX 5070 Super with 18GB?
Only if you specifically need more than 12GB and can accept an open-ended wait. If you need a card now, the 16GB RTX 5070 Ti is available today and covers the same need.
Would a Super card be much faster than the regular card?
Leaks suggest modest gains, mostly from faster memory and higher clocks. The main benefit is the larger memory pool, not a major jump in shader performance.
Write Your Review
No reviews yet. Be the first to share your experience!