Samsung plans to unveil GDDR memory in a month. With an impressive speed of 37 Gbps, it is the fastest in the world. It promises to significantly improve GPU performance, while optimizing their energy efficiency and implementation costs.
The advent of memory GDDR7 by Samsung risks marking a revolution in the field of graphics technologies by promising unrivaled performance for future devices. This development is particularly significant for NVIDIA’s RTX 5000 series, where expectations for speed and efficiency are considerable. Although speculation about the speed of these upcoming graphics cards varies, the integration of GDDR7 could be the deciding factor in reaching previously undreamed-of performance heights.
The RTX 5000 models, eagerly anticipated from 2024, are at the heart of discussions regarding future advancements in graphics andartificial intelligence. The RTX 5090, in particular, is expected to offer twice the power of its predecessor, the 4090, thanks in particular to the adoption of the GDDR7 video memory.
Samsung plans to unveil the world’s fastest GDDR7 memory next month
As part of the constant evolution of memory technologies, Samsung is preparing to take a new step with its GDDR7 memory. Scheduled to be presented at the IEEE International Solid-State Circuit conference in February 2024 in San Francisco, this innovation is distinguished by its use of PAM3 coding. This technological choice makes it possible to achieve significantly higher data transmission rates per cycle compared to NRZ, without requiring high memory bus frequencies. It thus makes it possible to obtain higher speeds without increasing the energy consumption.
Thus, GDDR7 is positioned not only as an improvement over GDDR6 in terms of performance but also as a solution more energy efficient And less expensive to implementpromising a speed of 37 Gbps per pin. Compared to the current 24 Gbps model, it offers 1.4x performance improvement and an energy efficiency increased by 20%.
Samsung’s introduction of GDDR7 memory is eagerly awaited, not only for its potential impact on the next generation of GPUs from AMD and NVIDIA, but also for its broader implications in the graphics and intelligence sectors artificial. With innovations such as four different reading modes and the ability toissue two independent commands in parallelit promises to optimize memory efficiency and power consumption.
Sources: techradar; Samsung