GDDR6 vs GDDR7

GDDR6 and GDDR7 are successive generations of graphics DRAM, used as GPU memory. They differ mainly in signaling. GDDR6 uses NRZ, two signal levels, per Micron and NVIDIA. GDDR7 uses PAM3, three levels that NVIDIA says carry 1.5 bits per cycle. GDDR6X, the in-between part, used PAM4, four levels and 2 bits per cycle. Micron says GDDR7 is not backward compatible with GDDR6 or GDDR6X and needs a new memory controller.

Samsung lists GDDR6 at up to 24 Gbps per pin. Micron lists GDDR7 at up to 32 Gb/s, 1.2 V, and up to 60 percent more bandwidth than GDDR6. Shipping cards sit lower. NVIDIA's RTX PRO 6000 Blackwell carries 96 GB of 28 Gbps GDDR7 on a 512-bit bus for 1,792 GB/s. The RTX A6000 carries 48 GB of GDDR6 at 16 Gbps on a 384-bit bus for 768 GB/s.

Compare cards on bandwidth in GB/s, which is bus width times data rate divided by 8. The memory generation alone misleads: by that formula a 192-bit GDDR7 bus at 28 Gbps gives 672 GB/s, under the 768 GB/s of the 384-bit GDDR6 card above.

Sources

Last verified October 30, 2026.