Nvidia RTX 50 Super: Leaked Specs, Delays, and the New 12GB Entry-Level Card

The path to market for Nvidia’s RTX 50 Super refresh has been anything but smooth. After a year of shifting timelines, leaked cancellations, and a no-show at Computex 2026, the mid-generation Blackwell refresh has been declared alive again by one prominent leaker. As of June 5, 2026, the story has turned on a dime: the cards are said to be "back on track" and have grown to include an additional entry-level model that would offer a meaningful VRAM boost for 1080p users. The catch is that Nvidia still hasn’t said a word.

The Planned Lineup and Memory Overhaul

The Super refresh is fundamentally a memory play. The proposed models—based on leaked specifications from kopite7kimi and MEGAsizeGPU—would swap the current 2 GB GDDR7 modules for 3 GB dies, allowing for a 50 percent increase in graphics memory without redesigning the memory bus. According to a report from Guru3D, the leaked specifications are as follows :

• RTX 5080 Super: Built on GB203 silicon with a full 84 SM count, this card is expected to move from 16 GB to 24 GB of GDDR7, with memory speeds reaching 32 Gbps and a bandwidth of approximately 1,024 GB/s. The total graphics power was listed at up to 415W .
• RTX 5070 Ti Super: Also featuring 24 GB of GDDR7 (up from 16 GB) on a 256-bit bus, with 28 Gbps memory .
• RTX 5070 Super: A bump from 12 GB to 18 GB of GDDR7 on a 192-bit bus, with 6,400 CUDA cores—a slight 4% core-count increase over the base 5070 .

The performance gain over the non-Super variants is estimated at a modest 5 to 10 percent . The primary value proposition for buyers—and the reason the memory upgrade is so critical—is the larger VRAM buffer for high-resolution gaming and entry-level AI tasks.

A Surprise Entry: The 12 GB RTX 5060

The most recent twist came on June 5, 2026, when leaker @Zed__Wang (MEGAsizeGPU) posted on X that the revived Super plans would now include a 60-class product with 12 GB of VRAM . This model would reportedly achieve its capacity by using four 3 GB GDDR7 modules on a standard 128-bit bus, sidestepping the performance compromises that a 96-bit cutdown (for a 9 GB variant) would have introduced .

• Naming uncertainty: The card may be called the RTX 5060 12GB or RTX 5060 Super 12GB .
• Core configuration: Leaked at roughly 4,096 CUDA cores .
• Estimated entry price: Around $349, a $50 premium over the 8 GB RTX 5060 .
• Target market: Designed for 1080p gaming and budget-friendly AI workloads .

The leaker stated there is no definitive launch window but expressed a personal belief that the cards will ship this year .

A Chaotic Timeline of Delays and Cancellations

The journey to this latest "back on track" rumor has been defined by production realities that sidelined gaming hardware. The Super refresh, internally codenamed "Kicker," has had its launch window slide rightward repeatedly :

• Mid to Late 2025: Initial rumors targeted a Q4 2025 or early 2026 launch .
• September 2025: Reports pushed the timeline to a March–May 2026 window, citing soaring VRAM costs and tight GDDR7 supply .
• November 2025: Nvidia reportedly told add-in board partners the launch had not been canceled but postponed to Q3 2026 .
• January 2026: Sources described an "indefinite delay" as Nvidia prioritized AI GPU production capacity .
• February 2026: The Information reported the dramatic news that Nvidia had shelved the Super refresh entirely and would release no new gaming GPUs in 2026. The RTX 60 series was said to be delayed until 2028 .
• April 2026: Igor's Lab reported that the project was still "up in the air" with no reliable confirmation of its status either way .

Throughout this period, the root causes were consistent: GDDR7 cost surges, a global memory supply shortage, and an overriding economic incentive for Nvidia to allocate high-margin AI chips over consumer GPU wafers and memory capacity .

The Computex 2026 Context

Jensen Huang's Computex 2026 keynote on June 1 was a landmark event for Nvidia's AI hardware aspirations but a silent one for PC gaming. The show was dominated by RTX Spark, an Arm-based superchip for Windows laptops, alongside the Vera CPU and the Vera Rubin rack-scale system . Nvidia confirmed no new GeForce consumer GPUs were announced at the keynote . This silence was expected; TechRadar’s live coverage noted there was "no appetite from Nvidia to launch a new GPU any time soon" given the ongoing memory shortages .

The June 5 rumor arrived just days after the keynote, creating a dramatic pivot in the narrative from cancellation to cautious optimism—a pivot driven not by Nvidia, but by a single leaker's claim . Until an official announcement surfaces, the Super lineup remains on paper, its story a testament to how AI demand is reshaping the gaming GPU market.