a reliable one by source, NVIDIA’s upcoming RTX 50 GB200 series graphics cards will not have a 512-bit memory bus. Instead, the flagship GPU in this series, called GB202, will continue to use the 384-bit memory bus configuration in the ADA architecture. In fact, the entire GB200 GPU family is expected to have similar memory bus features to the RTX 40 series.
Despite everything, it has been officially approved by JEDEC that there will be an upgrade from GDDR6/GDDR6X to GDDR7 in memory technology. Thanks to this new standard, memory bandwidth per module will increase and allow larger memory capacities such as 3 GB per module. Of course, this offers new opportunities for NVIDIA, but no definitive information has yet been given about the memory capacity features of the RTX 50 series. Micron, NVIDIA’s memory modules partner, announced plans to release 24Gb (3 GB) modules by 2025. NVIDIA’s GB200 series GPUs will be produced on TSMC’s 3nm manufacturing process and have DisplayPort 2.1 support.
Possible memory bus configurations for GB200 series GPUs are as follows:
- GB202: 384-bit / 32 Gbps / 24 GB (Maximum Memory) / 1536 GB/s (Maximum Bandwidth)
- GB203: 256-bit / 32 Gbps / 16 GB (Maximum Memory) / 1024 GB/s (Maximum Bandwidth)
- GB204: 192-bit / 32 Gbps / 12 GB (Maximum Memory) / 768.0 GB/s (Maximum Bandwidth)
- GB206: 128-bit / 32 Gbps / 8 GB (Maximum Memory) / 512.0 GB/s (Maximum Bandwidth)
- GB207: 128-bit / 32 Gbps / 8 GB (Maximum Memory) / 512.0 GB/s (Maximum Bandwidth)
Follow our site to read more technology news!
