NVIDIA has confirmed its adoption of the newly ratified GDDR7 memory standard by JEDEC for its upcoming consumer-grade graphics cards codenamed “Blackwell” generation. However, only the top three models of the GeForce RTX 50-Series will utilize GDDR7.
This implies that only the GB202, GB203, and GB205 chips will feature GDDR7 specifications. These chips are expected to power the RTX 5070 to RTX 5090, built on TSMC’s 4N fabrication process, equivalent to 4nm.
Meanwhile, lower-tier models in the “Blackwell” generation are likely to employ older memory standards such as GDDR6 or GDDR6X. These models are successors to the current AD106 and AD107, which are the core components of RTX 4060 Ti and below.
GDDR6X itself is an exclusive standard developed in collaboration between NVIDIA and Micron, boasting speeds of up to 23Gbps. On the other hand, the newly ratified GDDR7 standard by JEDEC offers speeds of up to 32 Gbps or even 36 Gbps.
However, initial implementations of GDDR7 are believed to be limited to 28 Gbps, hence NVIDIA’s choice to opt for GDDR6X in lower-tier models.
Moreover, NVIDIA is expected to utilize standard GDDR6 with speeds of up to 20 Gbps for entry-level graphics cards in the upcoming generation. The RTX 50-Series is anticipated to launch either at the end of 2024 or early 2025.
