Samsung GDDR7: 32Gbps High-Bandwidth Memory
Samsung has made an exciting announcement regarding its first-generation GDDR7 memory, signaling a significant leap in high-bandwidth technology. The global tech giant is leading the way as one of the first memory suppliers to develop and introduce this cutting-edge memory type.
The GDDR7 memory is expected to achieve an impressive bandwidth of 32Gbps/pin, surpassing the current best GDDR6 components by an impressive 33%. Samsung plans to achieve this remarkable performance boost by adopting PAM-3 signaling for data transfer.
Although the final specifications from JEDEC are yet to be released, Samsung's competitor, Micron, has already revealed its plans to launch its GDDR7 memory around the same time in 2024. Additionally, Cadence, a leading EDA tools company, has shared technical insights into GDDR7, showcasing its high-speed data transfer capabilities of up to 36Gbps/pin using PAM-3 signaling.
Another significant advantage of Samsung's GDDR7 lies in its improved energy efficiency. Operating at a lower voltage of 1.2v compared to its predecessor, it delivers a 20% increase in energy efficiency, making it a promising choice for high-bandwidth memory technology.
With a capacity of 16Gb for its first GDDR7 chip, Samsung matches the current density of top-tier GDDR6(X) chips. As a result, assuming the same memory bus width, the memory capacity of the final product is not expected to differ significantly from current offerings.
Moreover, Samsung has confirmed that it will offer a low-voltage version of GDDR7 memory, running at 1.1v and operating at a clock speed close to that of GDDR6. This marks a significant step forward in energy-efficient solutions.
To achieve the impressive energy efficiency boost, Samsung will employ the D1z process, the same technology used in their latest GDDR6, which utilizes EUV technology to improve efficiency by 20% without relying on newer nodes.
Samsung's breakthrough GDDR7 memory is set to revolutionize the market, offering faster performance and improved energy efficiency for various computing needs.
