Increased GDDR7 VRAM capacity is poised to bring about a significant change in the world of graphics cards in the coming years. Recent information suggests that memory manufacturers may release new chips with 4GB and 6GB capacities in 2027 or 2028.

The vast majority of GDDR7 chips used in current graphics cards offer a standard structure with 2GB of capacity per chip. These new generation solutions will allow for a significant increase in the total VRAM size without needing to make any changes to the memory data bus width.

Next Generation Memory Technology

Leaks shared by Golden Pig Upgrade reveal that memory manufacturers are preparing GDDR7 chips with 4GB and 6GB capacities. Samsung, SK Hynix, or Micron have not yet made any official product announcements regarding these capacities.

Increasing the capacity to 4 GB makes it possible to double the total VRAM using the same number of memory chips. 6 GB chips, on the other hand, promise a significant leap by offering three times more capacity per chip compared to existing 2 GB solutions.

For example, a design in the GeForce RTX 5060 class, which has a 128-bit memory bus and offers 8 GB of VRAM with four 2 GB chips, can reach 16 GB of capacity with 4 GB chips. Similarly, theoretically, 24 GB of VRAM capacity becomes possible when 6 GB chips are used on a 128-bit data bus.

Of course, for these capacity increases to occur, the memory controller on the GPU and the card design must support the relevant chips. This development on the GDDR7 side has actually already started with 3 GB capacity chips.

SK Hynix has shown that this technology is no longer just a leak, by showcasing both 2GB and 3GB GDDR7 products in 2026. The company’s 3GB solution finds its place in its product portfolio, especially for graphics workloads.

Cost and Performance Balance

It has long been rumored that NVIDIA will also switch to 3GB GDDR7 chips in its GeForce RTX 50 SUPER series. This structure allows for the use of higher VRAM capacities without changing the memory data bus in existing models.

The release schedule of the RTX 50 SUPER family is not yet clear due to the impact of memory prices. It is claimed that the 9GB GeForce RTX 5050 model, which previously came up with a 96-bit data bus and three 3GB chips, has been canceled or postponed.

High memory costs and GDDR7 supply constraints are among the main obstacles to such projects. The availability of 4GB and 6GB chips in 2027 or 2028 may coincide with the GeForce RTX 60 series, expected to use NVIDIA’s Rubin architecture.

The new capacities could affect not only high-end cards but also mid-range designs. With 4GB chips, it becomes possible to use 16GB of VRAM in a 128-bit design, 24GB in a 192-bit design, and 32GB in a 256-bit design.

6GB chips, on the other hand, allow for capacities of 24GB, 36GB, and 48GB respectively on the same data buses. Which combinations manufacturers will prefer will depend entirely on GPU design, cost, and product positioning.

The biggest obstacle to increasing memory capacity remains cost. The fact that 3 GB GDDR7 chips are more valuable than 2 GB chips directly affects the plans for new graphics cards. Do you think higher VRAM capacity or more affordable models are more valuable in graphics cards?