Seasonic's calculator leaks again: three RTX 50 SUPER models show up in unison, with the 5080 SUPER heading toward 415W. Some manufacturers keep their new products tightly under wraps, while others... even their power supply selection tool spills all the beans. Seasonic, the globally renowned PSU brand, has had its online wattage calculator thoroughly combed through by eagle-eyed netizens recently—and there, plain as day, sit three unreleased NVIDI

💡 What You Will Learn

Seasonic's calculator leaks again: three RTX 50 SUPER models show up in unison, with the 5080 SUPER heading toward 415W. Some manufacturers keep their new products tightly under wraps, while others...

📜 Table of Contents

• Power Draw on All Three Cards: Each One Heavier Than the Last
• VRAM Jumps 50% — NVIDIA Finally Listened
• What Can 24GB VRAM Do? Way More Than Gaming
• Wait or Buy Now? The Advice Is Straightforward

Seasonic's Calculator Leaks Again: Three RTX 50 SUPER Cards Appear Together, 5080 SUPER Heading for 415W

Some manufacturers keep their new products tightly under wraps, while others... leak everything through their own power supply selector tool.

Seasonic, the globally renowned PSU brand, had its online PSU wattage calculator recently dug through by eagle-eyed netizens—and there, plain as day, sat three unreleased NVIDIA graphics cards: the RTX 5080 SUPER, RTX 5070 Ti SUPER, and RTX 5070 SUPER.

The irony here is that a power supply calculator—designed to tell you "what size PSU you need"—has become the primary leak channel for NVIDIA's new products. Seasonic's engineers have to pre-configure power recommendations for upcoming GPUs, and once those TDP figures are written into the product database, the public-facing tool updates automatically—more reliable than any insider tip.

This isn't the first time Seasonic has played the "mole." A few months ago, the same tool leaked the 5070 Ti SUPER and 5070 SUPER, with only the 5080 SUPER missing at the time. Now all three are together, completing the SUPER family lineup. Statistically speaking, when a PSU maker's selection tool leaks unreleased product TDPs twice in a row, it's safe to assume these are the final specs.

Power Draw on All Three Cards: Each One Heavier Than the Last

All three SUPER cards see a bump over their standard counterparts, but the margins vary significantly:

✅ RTX 5070 SUPER — 275W 25W more than the standard 5070 (250W). A modest increase, typical of the conservative SUPER approach—the previous-gen 4070 SUPER only added 20W over the 4070. A 25W bump puts virtually no strain on cooling or PSUs; existing 650W units will handle it with ease.

✅ RTX 5070 Ti SUPER — 350W A full 50W higher than the 5070 Ti (300W). That's a sizable jump—50W is genuinely rare in a mid-cycle refresh. Most of the extra power goes to the memory subsystem upgrade and TDP headroom. If you're currently on a 750W PSU, you'll likely want to step up to 850W for peace of mind.

⏸️ RTX 5080 SUPER — 415W 55W above the RTX 5080 (360W), edging close to the previous-gen flagship RTX 4090's 450W. An 80-class card touching 90-class power levels hasn't happened in several generations. For PSU pairing, we'd recommend starting at 1000W.

Looking at all three together, NVIDIA's SUPER strategy this generation is straightforward: skip process-node efficiency gains, just unlock the TDP ceiling, and trade power for more VRAM capacity and stable boost clocks. This is a complete departure from the RTX 30-series SUPER approach—back then it was about doubling CUDA core counts; this time, the bet is on doubling VRAM capacity.

VRAM Jumps 50% — NVIDIA Finally Listened

The biggest highlight across the entire SUPER lineup is memory.

NVIDIA has switched from 2GB GDDR7 modules to 3GB GDDR7 modules, a 50% capacity increase per chip. The impact on specific models is immediate:

Both the 5080 SUPER and 5070 Ti SUPER break into the 24GB club, while the 5070 SUPER lands at 18GB.

This comes down to a natural advantage in memory bus width: the 5080 SUPER and 5070 Ti SUPER both use a 256-bit bus, and 8× 3GB modules fill it perfectly with zero waste. The 5070 SUPER uses a 192-bit bus, and 6 modules at full capacity gives exactly 18GB. On memory speed, the 5080 SUPER also gets a bump from 30 Gbps to 32 Gbps, pushing bandwidth just past the 1TB/s threshold.

As for core counts—sorry, nothing to see here. The 5080 SUPER keeps the same 10,752 CUDA cores, and the 5070 Ti SUPER stays at 8,960. The only change is on the 5070 SUPER, which goes from 48 SMs to 50, adding 256 cores—basically within margin of error.

NVIDIA's message couldn't be clearer: benchmarks can wait, but your VRAM anxiety is getting cured right now. Performance gains might only be single-digit percentages, but VRAM is up a solid 50%—and that's far more useful than a 5% score bump.

What Can 24GB VRAM Do? Way More Than Gaming

More VRAM benefits far more than just gamers.

Gamers are the first to benefit, of course. The 18GB 5070 SUPER is already a ceiling-tier card at 1440p—maxed-out path tracing in Cyberpunk 2077 with full texture packs won't worry about VRAM overflow. The 24GB 5080 SUPER and 5070 Ti SUPER give you ample headroom for 4K ray tracing, and with a wave of domestic AAA titles hitting over the next two to three years, that capacity buffer is just right.

But the real surprise lies beyond gaming.

24GB of VRAM means you can run mid-sized local LLM inference on a consumer-grade card. Current mainstream open-source models in the 13B–34B parameter range (Llama 3, Qwen 2.5, DeepSeek, etc.) fit right within 24GB after INT4 quantization—a single card can run them end-to-end, no dual-GPU setups, no NVLink required.

Six months ago, this configuration required a workstation card like the A5000, at triple the price starting out. Now an 80-series consumer card can do the same job. The gap? You know it if you know it.

For AI developers, 3D rendering designers, and content creators cutting high-bitrate video, the SUPER lineup's VRAM upgrade might be the most tangible selling point of the Blackwell generation. A 10% benchmark gain might go unnoticed, but jumping from 16GB to 24GB is something you feel immediately when working—no more closing one project to open another, no more dropping from FP16 to INT4 precision just to make things fit.

Wait or Buy Now? The Advice Is Straightforward

On the timeline, the SUPER series is expected to be announced at CES in early 2027, with retail availability rolling out through Q1.

That's a full year later than originally planned. GDDR7 global shortages, AI demand squeezing consumer-grade capacity, DRAM price hikes—the objective factors stack up one after another, and NVIDIA's SUPER schedule kept slipping. Rumors of "cancellation" and "indefinite delay" circulated, but thankfully none came true.

But late also has its upside: after a year of production ramp-up, 3GB GDDR7 yields have improved and costs are coming down. Beyond product cadence, NVIDIA's choice of this window also signals an intent to use VRAM advantage to box in AMD—if the competition pushes large-VRAM models in the mid-range, the standard 12GB and 16GB options would be caught flat-footed.

So back to the question everyone's asking: Is 16GB on the 5080 really enough?

Honestly, buying a 16GB 5080 now means you'll likely regret it a year from now. Wait six more months, and you get 24GB VRAM + a higher TDP ceiling + a more mature driver ecosystem—the math works out in your favor no matter how you slice it. For those who've been sitting on the fence for the better part of a year, early 2027 might be the real moment to pull the trigger in the Blackwell generation.

Will you wait for SUPER, or grab the standard version now? Drop your take in the comments.