GDDR6GDDR6X

GDDR6 and GDDR6X are high-speed graphics memory standards designed for modern GPUs and graphics accelerators. They are siblings in the GDDR family, with GDDR6 serving as the widely adopted baseline and GDDR6X representing a later, higher-speed evolution. The term GDDR6GDDR6X is not an official designation, but is sometimes used informally to refer to the two-generation family in discussions.

GDDR6 was introduced to deliver higher bandwidth than its predecessors while maintaining power efficiency suitable for consumer GPUs. It typically operates around 1.35V and targets data rates up to 16 Gbps per pin. Memory chips come in densities ranging into a few tens of gigabits, and GPUs use multiple chips to form wide memory buses (commonly 256-bit or 320-bit) to achieve large aggregate bandwidth. GDDR6 enabled substantial increases in graphics memory bandwidth, supporting higher resolutions, faster frame rates, and more complex textures.

GDDR6X, announced and implemented a few years after GDDR6, pushes the limits further by using PAM4 signaling, which increases data rate per pin without simply raising clock frequency. This enables speeds around 19 to 21 Gbps per pin, with some implementations reaching up to about 24 Gbps. The resulting memory bandwidth on high-end GPUs—combined with wide memory buses—reaches roughly 800 to 960 GB/s in typical configurations. Micron is a primary supplier of GDDR6X memory, and Nvidia has used GDDR6X extensively in its RTX 30-series graphics cards.

Applications of both standards include gaming GPUs, professional graphics, and compute accelerators that require high memory bandwidth. The two generations reflect a progression toward higher data rates and bandwidth while balancing power and cost considerations.