Micron is now mass producing next generation 8-gigabit (Gb) GDDR6 memory chips that will be used in a broad range of applications, the company announced today. The significance of this is that it means Micron is right on schedule with its next-gen graphics memory rollout—Micron in December of last year stated that its initial progress was "very promising" and that it was aiming to ramp up production in the first half of 2018.
These new chips are Micron's fastest and most powerful graphics memory to date. Micron envision GDDR6 being used in high performance applications such as artificial intelligence (AI), networking, automotive, and of course paired with graphics processing units (GPUs). While not specifically mentioned, it is widely believed that NVIDIA's next generation GeForce 20 series GPUs (Turing) will feature GDDR6 memory.
"Micron is a pioneer in developing advanced high bandwidth memory solutions and continues that leadership with GDDR6. Micron demonstrated this leadership by recently achieving throughput up to 20GB/s on our GDDR6 solutions," said Andreas Schlapka, director, Compute Networking Business Unit, Micron. "In addition to performance increases, Micron has developed a deep partner ecosystem to enable rapid creation of GDDR6 designs, enabling faster time to market for customers looking to leverage this powerful new memory technology."
We are most interested in graphics when it comes go GDDR6 memory, and as it pertains to upcoming graphics cards, Micron is prepping chips rated to run at 12Gbps and 14Gbps. Both run at 1.35V, which is where the power savings come from—GDDR5 runs at 1.5V to 1.6V. The 12Gbps memory is capable of delivering 48GB/s of memory bandwidth per chip, while the 14Gbps packages kick things up to 56GB/s.
"Enabling significant performance improvements for today's top GPUs, GDDR6 delivers enhanced graphic memory speeds to enable higher application bandwidth. Micron GDDR6 will be a core enabling technology of advanced GPU applications, including acceleration, 4K video and improved rendering, VR/AR and crypto mining applications," Micron added.
Compared to previous generations memory architectures, GDDR6 provides higher densities than either GDDR5 and the less frequently found GDDR5X. It also doubles the bandwidth, extending speeds past GDDR5X, and is based on a dual-channel architecture (2 x 16-bit instead of 1 x 32-bit). According to Micron, switching to a dual-channel architecture "enables a huge performance increase while still providing backward compatibility to GDDR5 memory access size."
One thing to keep in mind is that GDDR6 is not a direct replacement for GDDR5/GDDR5X due to package size differences. This means you aren't likely to see current-generation graphics cards refreshed with GDDR6. Instead, it will be found on cards featuring next generation GPUs.