RTX 5070 Super Leak: 18GB VRAM and What It Means for 1440p Gamers

The leak in one paragraph

The claim, sourced from kopite7kimi and amplified across outlets like TweakTown, Wccftech, VideoCardz, and Tom's Hardware, is that Nvidia is preparing a "Super" refresh of the Blackwell RTX 50 lineup. The headline change is memory. The RTX 5070 Super would carry 18GB of GDDR7 (up from 12GB), the RTX 5070 Ti Super and RTX 5080 Super would both move to 24GB, and there may even be an RTX 5060 Super with 12GB. None of this is confirmed by Nvidia. Every number below should be read as a well-sourced rumor, not a spec sheet.

RTX 5070 Super specs: 18GB and not much else

Here is the part that will disappoint anyone expecting a big performance leap. The RTX 5070 Super specs 18GB rumor is almost entirely a memory story. The leaked configuration keeps the same GB205 silicon and only nudges the shader count from 6,144 CUDA cores to 6,400 — a bump of roughly 4%. That is a rounding error in most games.

What changes is the VRAM. The current RTX 5070 uses 2GB GDDR7 modules across a 192-bit bus to reach 12GB. The Super would swap in the newer 3GB GDDR7 modules on the same 192-bit bus, and six 3GB chips get you to 18GB without widening the memory interface or redesigning the board. Memory bandwidth reportedly stays flat at 672 GB/s because the chips still run at 28Gbps. Power creeps up about 10%, from a 250W TGP to a rumored 275W, which hints at slightly higher boost clocks.

Leaked RTX 50 Super lineup vs current cards

Notice the pattern across the whole rumored stack: the Ti Super and 5080 Super keep identical CUDA core counts to their non-Super versions. The uplift, if it materializes, comes from more memory and slightly higher clocks — not more shaders.

Why 12GB is the RTX 5070's real problem

To understand why an 18GB refresh matters, you have to understand why 12GB became a sore point so fast. When the RTX 5070 launched at $549, reviewers broadly agreed the raw horsepower was fine for 1440p but the 12GB buffer was tight for a card sold as a high-refresh 1440p and entry-4K option.

VRAM is not about frame rate on paper. It is about what happens when you run out. When a game needs more memory than the card physically has, the driver starts shuffling assets between VRAM and system RAM over the PCIe bus, which is dramatically slower. The symptom is not a lower average FPS — it is stutter, texture pop-in, sudden frame-time spikes, and 1% lows that crater even while the average looks healthy in a benchmark bar chart.

Modern releases have made this worse. Games leaning on high-resolution texture packs, ray tracing, and frame generation all consume extra memory. Ray tracing needs to store BVH acceleration structures. Frame generation buffers additional frames. Load a demanding title at 1440p with ray tracing and a texture pack, and 12GB can fill up in a hurry.

What 18GB actually unlocks at 1440p

Jumping from 12GB to 18GB is a 50% increase in capacity, and at 1440p that is the difference between "usually fine" and "comfortable headroom." It matters in a few concrete ways.

Texture quality without compromise. The single cheapest way to make a game look better is maxing texture resolution, and textures cost VRAM, not compute. With 18GB you can leave textures on Ultra in memory-hungry titles without triggering the pop-in that plagues 12GB cards.

Ray tracing plus frame generation together. These two features are where 12GB cards get squeezed hardest. Extra buffer lets you stack path-traced lighting and DLSS frame generation at 1440p without the frame-time spikes that ruin the smoothness these features are supposed to deliver.

Longevity. A GPU bought in 2026 should still be viable in 2029. VRAM requirements only trend upward. An 18GB card has a materially longer runway than a 12GB one, which is the whole reason this leak generated so much attention.

Light 4K and high-res mods. The RTX 5070 Super will not become a 4K powerhouse — the GPU is still a mid-tier chip and bandwidth is unchanged. But for 4K with DLSS upscaling, or for heavily modded games at 1440p, having the memory to hold everything in the buffer removes a real bottleneck. If you want to see where this class of card lands against the rest of the stack, our GPU tier list is a quick sanity check.

The catch: bandwidth and cores stay the same

It is worth being blunt about the ceiling here. More VRAM does not make a GPU faster in scenarios where it was never memory-limited. If you play esports titles at 1440p that already run at hundreds of frames per second on a 12GB RTX 5070, the Super's 18GB will do essentially nothing for you. Bandwidth is identical, and the roughly 4% core bump is invisible in practice.

The 18GB matters specifically in memory-constrained situations — high textures, ray tracing, frame generation, 4K, mods. That is a meaningful slice of demanding modern games, but it is not "every game." Treat the Super as a fix for a specific weakness, not a generational leap. If you want to check whether your current setup already clears the bar for the games you play, Can I Run It will tell you fast.

Pricing rumors: same money, more memory

The most encouraging part of the leak is pricing. Multiple reports suggest Nvidia intends to hold the line on MSRP, with the Super cards landing at or near the same prices as the originals they replace.

Early pricing leaks were wider — the RTX 5070 Super was floated anywhere from $549 to $599, and the RTX 5080 Super from $999 to $1,199 — before more recent reports narrowed things back to the original MSRPs. If those numbers hold, the value proposition is straightforward: the same asking price for 50% more memory. That would be a genuinely good deal by recent GPU standards. The obvious caveat is that MSRP and street price are two different animals. The current RTX 5070 launched at $549 but has frequently sold closer to $599-$630 at US retailers like Newegg, Best Buy, and Amazon, only occasionally dipping toward MSRP during sales. Expect the same gap between sticker and reality if and when the Super arrives.

The messy timeline: 2026, 2027, or never

This is where honesty is required. The release window for the RTX 50 Super series has been a moving target for the better part of a year.

The series was widely expected at CES in January 2026 with a big VRAM bump across the board. It never appeared. In January 2026, multiple reports went further: leakers claimed Nvidia had told its board partners the Super refresh was "delayed indefinitely," language some outlets read as corporate-speak for quietly cancelled. Then the story flipped again. Around Computex in mid-2026, one leaker claimed the lineup was "back on track" for a late-2026 launch, while other reports pointed to a slip into early 2027 and a possible CES 2027 reveal. Earlier rumors had even floated a Q4 2025 window that obviously came and went.

As of July 2026, all of this remains unconfirmed. Nvidia has not officially announced any Super card, none have shipped, and the reporting is openly contradictory — some sources say 2026, some say 2027, and some suggest it may never arrive in this exact form.

The reported reason for the delays is the same thing that makes the 18GB and 24GB configurations possible in the first place: high-density 3GB GDDR7 modules. That memory is expensive and reportedly in short supply, with manufacturers prioritizing far more profitable AI datacenter demand over consumer graphics cards. There is also a strategic angle: with no new competing AMD Radeon flagship pressuring Nvidia in 2026, the company has little urgency to spend scarce memory on a mid-cycle refresh — especially if a next-generation RTX 60 series is only a year or so out. In other words, the exact component that makes the Super refresh compelling is also the bottleneck holding it up. Until GDDR7 supply loosens, timing stays genuinely uncertain.

Should you buy an RTX 5070 now or wait?

The unsatisfying but correct answer: it depends on your timeline and how you play.

Buy now if you need a GPU today, you play mostly at 1440p in titles that do not hammer VRAM, and you can find an RTX 5070 near its $549 MSRP. It is a capable 1440p card right now, and waiting for an unannounced product with a slipping — possibly cancelled — schedule is a gamble.

Wait if you specifically want to max ray tracing plus frame generation, dabble in 4K, or keep a card for four-plus years — and you can tolerate an open-ended wait that might stretch into 2027 or never materialize. The 18GB buffer directly addresses the RTX 5070's weakest point, and if the same-MSRP pricing holds, it would be the better long-term buy.

One more consideration: if the Super series does launch, it will likely push current RTX 5070 and 5070 Ti prices down as Nvidia clears inventory. Even if you do not want the Super itself, its arrival could be a good moment to grab a discounted current-gen card.

FAQ

Is the RTX 5070 Super officially confirmed?

No. As of July 2026, Nvidia has not announced the RTX 5070 Super or any card in the rumored RTX 50 Super series. All specs, including the 18GB VRAM figure, come from leakers — primarily kopite7kimi — and should be treated as rumor until Nvidia says otherwise.

How much VRAM will the RTX 5070 Super have?

Leaks point to 18GB of GDDR7 on a 192-bit bus, up from 12GB on the current RTX 5070. The increase is made possible by newer 3GB GDDR7 memory modules, which let Nvidia raise capacity 50% without widening the memory interface.

Is 18GB of VRAM enough for 1440p and 4K gaming?

For 1440p, 18GB is comfortable headroom and should handle max textures, ray tracing, and frame generation without the stutter that can affect 12GB cards. For 4K, 18GB helps, but the RTX 5070 Super's unchanged 672 GB/s bandwidth and mid-tier core count mean it is best suited to 4K with DLSS upscaling rather than native 4K in the most demanding titles.

When will the RTX 5070 Super be released?

The timeline is unsettled and the reporting is contradictory. Early rumors pointed to CES 2026; that came and went. In January 2026, some leakers said Nvidia had told partners the refresh was "delayed indefinitely." More recent reports are split between a late-2026 launch and a slip into early 2027 (a possible CES 2027 reveal), reportedly gated by tight, expensive 3GB GDDR7 supply being prioritized for AI. No official date exists.

Will the RTX 5070 Super be worth the upgrade over the RTX 5070?

If you own a 12GB RTX 5070 already, probably not — the roughly 4% core bump is negligible and you would be paying again for memory you may not need. The Super makes more sense for new buyers who want the larger frame buffer for longevity, ray tracing, and high-resolution textures at 1440p.