Nvidia RTX 60 delayed to 2028: flagship VRAM stacked to 48GB, gamers wait an extra year and pay more

Veteran leaker kopite7kimi gave a clear statement on September 21: Nvidia's next-generation gaming graphics card GR20x (the gaming version of the Rubin architecture, corresponding to the RTX 60 series) is a "2028 product," and the launch plan has been delayed once again. Following the pace of one generation every two years over the past, the new cards should have been in 202

📜 Table of Contents

• How 48GB of VRAM Gets Stacked
• Halving the Modules, PCB Benefits Too
• Why Gamers Have to Wait an Extra Year
• How to Plan for the Wait

NVIDIA RTX 60 Delayed to 2028: Flagship VRAM Stacked to 48GB, Gamers Wait an Extra Year and Pay More

Veteran leaker kopite7kimi stated clearly on September 21: NVIDIA's next-gen gaming GPU GR20x (the gaming version of the Rubin architecture, corresponding to the RTX 60 series) is a "2028 product," with the launch plan delayed once again. Following the past two-year-per-generation cadence, the new cards should have debuted at CES in early 2027 — now it's at least another year of waiting.

This leaker's track record is worth noting: his spec leaks for the RTX 30 and RTX 40 series were almost entirely accurate, and he's also correctly called key details about AMD's current-gen lineup multiple times ahead of launch. In this post, he added a line saying he'll "try to post as little as possible until there's confirmed news" — the wording carries a sense of resignation.

How 48GB of VRAM Gets Stacked

The key is die density. The current flagship RTX 5090's 32GB of VRAM consists of 16× 2GB GDDR7 chips on a 512-bit bus. 3GB modules from Samsung, SK Hynix, and Micron have already entered mass production this year, with 4GB and 6GB modules expected to land in 2027–2028 — right in the RTX 60 launch window. Next-gen cards could ship with high-density VRAM from day one.

The RTX 6090 keeps the 512-bit bus, using 8× 4GB chips for 32GB, or going straight to 8× 6GB chips for 48GB. The sub-flagship RTX 6080 gets 6 modules, starting at 24GB, or up to 36GB with 6GB modules. The RTX 6070 starts at 16GB, or 24GB with 6GB modules.

Halving the Modules, PCB Benefits Too

Cutting the chip count from 16 to 8 has direct benefits: signal routing is greatly simplified, PCB operating temperatures drop, and double-sided soldering is no longer needed. Manufacturers can also use high-capacity modules to build larger local VRAM pools, making it easier to run complex neural network models. Whether these designs roll out broadly mainly hinges on the initial cost of the new modules — final specs won't be locked in until the engineering testing phase.

Why Gamers Have to Wait an Extra Year

The delay math is easy to work out. TSMC is prioritizing capacity for far more profitable AI server accelerators, pushing consumer GPUs down the queue. The VRAM side is equally tight — the RTX 50 SUPER mid-cycle refresh was already cancelled due to DRAM price hikes, and 2GB GDDR7 modules have risen to about $20 per unit. The RTX 5090 launched at $1,999, and street prices have now doubled, generally sitting above $4,000.

What's more telling is what NVIDIA did with its financial reports: the gaming business has disappeared as a reporting category, merged into "edge computing" alongside robotics, and the latest quarterly report doesn't mention gaming once. For a company making its money on AI data centers, you can imagine where gaming cards rank in priority.

How to Plan for the Wait

Look on the bright side: an extra year of waiting gets you a card with more VRAM, fewer modules, and a cooler PCB. Look on the dark side: by the time the new cards launch, nobody can confidently predict where VRAM prices will be. That card you've got now — are you planning to make it last until 2028?