GDDR7 memory chips with denser capacities than 2GB or 3GB are reportedly planned for 2027 or 2028, which could power next-gen graphics cards. According to known tech leaker Golden Pig on Chinese social media platform Weibo, memory manufacturers are expected to release 4GB and 6GB GDDR7 chips, which represent a 100% and 200% density increase compared to the usual 2GB chips used today.
Leaving AMD and Intel graphics cards aside, as they’re still using GDDR6 memory for now, 2GB GDDR7 memory chips are currently the standard capacity by Nvidia. These chips are used by the entire desktop RTX 50 Series lineup, plus a portion of the mobile RTX 50 range.
There are 3GB GDDR7 chips available now, but the only cards using them are Nvidia’s RTX Pro cards and high-end RTX 5090 and RTX 5070 laptop cards, both of which use that higher density to pack more capacity into a limited PCB area. The rumoured RTX 50 Super refresh was also expected to feature 3GB chips, but the ongoing memory shortages seem to have halted this plan.
If this leak is accurate, then Nvidia could potentially offer double or triple the capacity compared to existing models, potentially as a drop-in replacement without needing to redesign the PCB. A 32GB RTX 5080 Super would probably fly off the shelves, for example. That said, given the extreme cost of memory right now, the card that has the best chance of getting a 6GB chip upgrade is the RTX Pro 6000, which would boost its capacity up to 192GB.
However, the leaker didn’t specify if these new chips will also offer a speed boost. Current 2GB and 3GB GDDR7 chips’ speeds range from 28Gb/s to 30Gb/s on GeForce cards, which results in 1.79TB/s of memory bandwidth. Even a small increase in per-chip speed could snowball into meaningful gains on cards rocking a wide 512-bit bus.
That said, we must temper our expectations, since Nvidia is likely to keep these fancy new chips for its next-gen GeForce RTX 60 Series, which turns out is also expected to release in 2028. Doing so would allow it to advertise massive VRAM increases compared to the RTX 50 Series without requiring complicated PCB wiring. AMD may also take advantage of these chips for its RDNA 5 GPUs.
But none of this will be feasible, at least at a reasonable cost, if current memory shortages remain. Hopefully by 2028, the situation will have calmed down a bit, so consumers can finally enjoy some new hardware.
