The video memory landscape is about to change, with rumors suggesting the arrival of GDDR7 chips with unprecedented densities. According to reports from certain media outlets, memory manufacturers are planning to release 4 GB and 6 GB chips by 2027 or 2028. These components could serve as the foundation for future graphics cards, offering total capacities far exceeding those currently available.
It is important to note that this information comes from a leak attributed to a tipster known as Golden Pig, who is active on the Weibo platform. According to this source, the 4 GB chip would represent a 100% increase in density compared to current standards, while the 6 GB version would offer a 200% increase. By way of comparison, 2 GB chips are currently the standard for most of Nvidia’s desktop RTX 50 series and are used in both consumer models and part of the mobile lineup.
A massive increase in potential memory capacity… but at what cost?
Currently, 3 GB GDDR7 chips are already available on the market. They are primarily found in RTX Pro professional graphics cards as well as high-end laptop models, such as the RTX 5090 and RTX 5070. These configurations leverage this increased density to maximize memory capacity within a limited PCB footprint. Although an update to the RTX 50 Super series featuring these 3 GB chips was anticipated, ongoing memory shortages appear to have stalled this project.
If this information proves accurate, Nvidia could potentially offer two or three times more video memory compared to existing models. The major technical advantage lies in the ability to integrate these new chips without requiring a complete PCB redesign. For example, an RTX 5080 Super card equipped with 32 GB of VRAM could become an extremely popular product. However, given the exorbitant costs of current memory, the model most likely to benefit from an upgrade to the 6 GB chip would be the RTX Pro 6000 professional graphics card, whose capacity could reach 192 GB.
Technical Limitations and Release Timeline
The leak does not specify whether these new chips will also offer an increase in transfer speed. Currently, the 2 GB and 3 GB GDDR7 chips on GeForce cards deliver transfer rates ranging from 28 GB/s to 30 GB/s, resulting in a memory bandwidth of 1.79 TB/s. Even a slight improvement in speed per chip could prove significant on cards equipped with a 512-bit bus.
Nevertheless, expectations should be tempered. Nvidia appears to be planning to reserve these innovative chips for its next-generation GeForce RTX 60 series cards, which are themselves expected to launch in 2028. This strategy would allow the company to highlight massive increases in VRAM compared to the RTX 50 series, without having to deal with complex PCB wiring. AMD could also take advantage of these components for its future graphics cards based on the RDNA 5 architecture.
Finally, the feasibility of these projects remains contingent on the resolution of memory shortages. If the situation does not improve by 2028, the cost could make these configurations prohibitively expensive for consumers. The hope remains that the market will stabilize enough to allow for the launch of this new hardware.