NVIDIA RTX 5080 SUPER: Every Leaked Spec, Delayed Launch, and What RTX Spark Means for PC Gaming

NVIDIA's next mid-generation GPU refresh, the RTX 50 SUPER series, has been the subject of intense speculation for over a year. The flagship of the rumored lineup, the GeForce RTX 5080 SUPER, is still unannounced, but a steady stream of leaks from multiple hardware insiders has painted a relatively consistent picture of its specifications — even as its launch date has been pushed back repeatedly.

Here is a fact-checked breakdown of everything that has been reported about the RTX 5080 SUPER, the broader RTX 50 SUPER lineup, and why you should not confuse these unconfirmed cards with the officially announced NVIDIA RTX Spark.

RTX 5080 SUPER: Leaked Specifications

All information about the RTX 5080 SUPER comes from unverified leaks, rumors, and speculation. NVIDIA has not confirmed any specs, pricing, or release dates .

Based on leaks from leakers such as kopite7kimi, Moore's Law is Dead, and MEGAsizeGPU, the reported specifications are:

• GPU Die: GB203-450-A1 (the same full die as the standard RTX 5080)
• CUDA Cores: 10,752 (unchanged from the RTX 5080)
• VRAM: 24GB GDDR7 (a 50% increase over the 16GB RTX 5080)
• Memory Speed: 32–36 Gbps GDDR7
• Memory Bus: 256-bit
• Memory Bandwidth: Approximately 1,024 GB/s to 1,152 GB/s
• TGP / TDP: Approximately 415 watts (up from ~360W on the RTX 5080)
• Boost Clock: Approximately 3.0 GHz (leaked)
• RT Cores: 84
• Tensor Cores: 336

Key takeaway: The core count is expected to be identical to the standard RTX 5080. The main upgrades would be the VRAM increase to 24GB, faster GDDR7 memory, and a substantially higher 415W TDP — roughly 15% more than the vanilla model . This power draw is the clearest indicator that the card will demand a robust power supply.

Power Supply Requirements

NVIDIA has not released any official power supply recommendations for the RTX 5080 SUPER. No specific PSU wattage has been widely leaked. As a general rule, a card with a 415W TGP typically requires an 850W to 1000W power supply. This is an extrapolation, not a confirmed figure, and should be treated with caution.

Launch Timeline: A Moving Target

The launch window for the RTX 50 SUPER series has shifted numerous times, and confidence in any single date is low.

• Mid-2025 leaks: Pointed to a Q4 2025 or holiday 2025 launch .
• Late 2025 leaks: The timeline was revised to March–May 2026 .
• Early 2026 leaks: Suggested a CES 2026 reveal, with possible retail in Q3 2026 due to GDDR7 supply constraints .
• June 2026 leaks (most recent): The most current reports suggest the RTX 50 SUPER series may have slipped to Q4 2026 or even CES 2027 . Leaker MEGAsizeGPU indicated the lineup is "back on track" but with a later window . A report from June 20, 2026, citing BenchLife, places an appearance at CES 2027 as more likely .

Bottom line: The launch has been delayed multiple times. The most recent signals (June 2026) point to a late 2026 or early 2027 release at the earliest .

The RTX 50 SUPER Lineup

The rumored SUPER series reportedly includes three to four models :

Model	Leaked VRAM	Leaked CUDA Cores	Leaked TDP
RTX 5080 SUPER	24GB GDDR7	10,752	~415W
RTX 5070 Ti SUPER	24GB GDDR7	8,960	~350W
RTX 5070 SUPER	18GB GDDR7	6,400	~275W
RTX 5060 SUPER (rumored)	12GB GDDR7	Not yet specified	Not yet specified

A recent development (June 2026) is a leak suggesting NVIDIA may also introduce an RTX 5060 SUPER with 12GB VRAM, expanding the lineup down-market . The VRAM increases across the board are widely attributed to NVIDIA reportedly moving from 2GB to 3GB GDDR7 memory chips .

Important: RTX Spark Is a Different Product

The RTX 50 SUPER series should not be confused with the NVIDIA RTX Spark, which was officially announced at COMPUTEX 2026 by CEO Jensen Huang on June 1, 2026 .

• What it is: An Arm-based system-on-chip (SoC) combining a 20-core NVIDIA Grace CPU with a Blackwell RTX GPU and unified memory .
• Target: Consumer Windows laptops and compact desktop PCs, purpose-built for local AI agents and personal AI workloads .
• AI Performance: Up to 1 petaflop of AI compute .
• Unified Memory: Up to 128GB of unified LPDDR5X memory .
• CUDA Cores: Approximately 6,144 Blackwell CUDA cores .
• NVIDIA's positioning: Described as "the most efficient PC chip ever built," and is NVIDIA's first entry into the consumer PC chip market .
• Availability: Systems shipping fall 2026 .

RTX Spark is a completely separate product from the GeForce RTX 50 SUPER discrete GPUs .

Verdict: What to Believe?

Topic	Status	Source Confidence
RTX 5080 SUPER specs (24GB, 415W, 10,752 CUDA)	Leaks only	Medium — consistent across multiple leakers
RTX 5080 SUPER launch date	Uncertain	Low — timeline has shifted repeatedly
RTX 50 SUPER lineup	Leaks suggest 3-4 models	Medium
RTX Spark	Officially announced	High — confirmed via NVIDIA press releases and media coverage

The leaked specifications for the RTX 5080 SUPER are consistent enough across several reputable leakers to be treated as plausible, but nothing is final until NVIDIA makes an official announcement. The launch timeline is highly uncertain and has been pushed back repeatedly. If you are planning a high-end PC build, treat the SUPER series as a 2027 product until proven otherwise.