According to President and Chief Executive Officer Sanjay Mehrotra, Micron plans to release its GDDR7 memory in the first part half of 2024. This announcement comes as Samsung has already said that it too is planning the production of its GDDR7 memory. Samsung, however, has not updated on when its next-gen VRAM will become available but an announcement could be forthcoming. In the meantime, Micron is planning on using its 1ß node for the new memory which will be available for consumer and datacenter GPU solutions.
From Micron (via Tom’s Hardware):
“We plan to introduce our next-generation G7 product on our industry-leading 1ß node in the first half of calendar year 2024.”
Micron did not share any specifications for its GDDR7 memory in its earning call but if Samsung’s December announcement is anything to go by it could support bandwidth up to 36 Gbps. By comparison, GDDR6X which is used in various NVIDIA GeForce RTX 30 and 40 series graphics cards, offers up to ~23 Gbps when overclocked. When used on a 256-bit bus Samsung also said its G7 memory could offer up to 1,152 GB/s and jumps up to 1,728 GB/s on a 384-bit bus.
It has been noted that consumers should not expect any new graphics cards from AMD, NVIDIA, or Intel, to use GDDR7 to its max potential, at least not in the first round, due to it using PAM3 signaling and different encoding than GDDR6/GDDR6X, and it will require new memory controllers. As Micron plans to introduce its G7 memory Cadence has already completed a G7 verification process so GPU manufacturers could begin implementing the next-gen memory for new products, which are expected by either 2024 or 2025. On a side note, NVIDIA was recently reported to be looking into HBM3E memory as a solution for its data center GPUs so between Micron, Samsung, and SK Hynix, the next round of high-speed GPU memory is on its way.