NVIDIA RTX 50 GPUs will use Samsung’s GDDR7 memory chips

As reported by Korean website Greened.kr, Samsung’s GDDR7 memory will be used in NVIDIA’s new products, scheduled for release early next year. Samsung was among the top three manufacturers developing GDDR7 memory, with SK Hynix and Micron.

Previously, NVIDIA opted for Micron memories in its RTX 40 and RTX 30 series GPUs such as GDDR5X and GDDR6X, but for GDDR7, NVIDIA opted for Samsung, which emerged the winner to power the GeForce RTX 50 “Blackwell” GPU lineup. ”.

The reason for Samsung’s victory in the GDDR7 segment is due to its early verification and validation when connected to the product.

For now, the GeForce RTX 50 “Blackwell” GPUs are expected to be launched at CES 2025, and the first products should hit shelves in January.

GDDR7 modules

Samsung’s GDDR7 models should be adopted in both desktop and laptop GPUs from the GeForce RTX 50 “Blackwell” line. For laptops, NVIDIA can utilize GDDR7 solutions from all three vendors, but Samsung will be prioritized.

It is known that NVIDIA should use 28 Gbps and 32 Gbps GDDR7 memory dies for its next line. The RTX 5080 in particular will use the fastest memory dies, while the rest of the lineup will stick with a 28 Gbps pin speed.

For the RTX 5090, the 28 Gbps solution still offers a significant performance increase, reaching 1792 GB/s (+78%) compared to the 1008 GB/s of the RTX 4090.

Another option for NVIDIA would be models with 32 Gbps, but the 5090 uses many more memory chips than the 5080 due to its 512-bit interface, which could affect the power consumption of the entire card.

Samsung has already announced 16Gb and 24Gb memory modules based on the GDDR7 standard, with its roadmap showing speeds of up to 42.5 Gbps for the future. A 24Gb solution could lead to greater capacities, but these modules are expected to be introduced later or as an upgrade.

Fonte: Wccftech e Greened.kr.