South Korean memory giant (and Samsung’s archrival) SK Hynix has announced the arrival of what it says is the "industry's best" GDDR7 graphics memory, promising a big leap in performance.

First developed in March 2024, this new DRAM product is designed to meet the increasing demand for high-speed, specialized memory in the AI sector. The company says it plans to begin volume production in the third quarter of 2024.

The GDDR7 memory boasts an operating speed of 32Gbps, which represents a 60% improvement over its predecessor. Depending on the usage, speeds can reach up to 40 Gbps. This new memory chip can process more than 1.5Tbps - equivalent to 300 Full-HD movies (5GB each) - when used in high-end graphics cards.

More efficient

SK Hynix says it has also improved power efficiency by over 50% compared to the previous generation thanks to advanced packaging technology that addresses heat issues associated with ultra-fast data processing.

The company has increased the number of heat-dissipating substrate layers from four to six and introduced Epoxy Molding Compound (EMC) for packaging, which has reduced thermal resistance by 74% while keeping the product size the same.

Sangkwon Lee, Head of DRAM Product Planning and Enablement at SK Hynix, highlighted that the GDDR7 is expected to find applications in high-spec 3D graphics, AI, high-performance computing, and autonomous driving. “We will continue to work towards enhancing our position as the most trusted AI memory solution provider by strengthening the premium memory lineup further,” he said.

As The Korea Economic Daily notes, SK Hynix has major competition in this field. In July 2023, Samsung announced it had developed the industry's first GDDR7 chip and was in the process of verifying its products on customer systems, while Micron also created its version of the GDDR7 chip last year and is now preparing for mass production.

More from TechRadar Pro



source https://www.techradar.com/pro/fastest-ever-graphics-memory-launches-but-still-lags-hbm-sk-hynix-gddr7-ram-tops-15tbps-and-could-find-its-way-in-more-affordable-ai-gpu-cards