Kioxia has developed a prototype flash memory module with a capacity of five terabytes and a bandwidth of 64 gigabytes per second, designed to meet the needs of large-scale artificial intelligence models and edge data centers. Kioxia reports that the prototype was developed as part of the Post-5G Information and Communication Systems Infrastructure Enhancement R&D Project commissioned by Japan’s New Energy and Industrial Technology Development Organization.
Kioxia’s new module targets the increasing demand for large-capacity, high-bandwidth memory in post-5G and 6G mobile edge computing (MEC) servers, which are essential for applications such as Internet of Things, big data analytics, and advanced AI processing. The company specifically addresses the traditional trade-off between capacity and bandwidth found in conventional DRAM-based modules by using a daisy-chain connection of flash memory controllers. This architecture enables the addition of more flash units without bandwidth degradation, supporting large capacities beyond previous limits.
The module features high-speed transceiver technology based on pulse amplitude modulation with four levels (PAM4), providing 128 gigabits per second signaling and reducing the number of connections needed. Kioxia has also implemented flash prefetch technology to minimize read latency during sequential data accesses. The memory bandwidth between controller and flash memory is improved to 4.0 gigabits per second through low amplitude signaling and distortion correction. The resulting memory module and controller prototypes use a PCI Express 6.0 interface at 64 gigabits per second across eight lanes, while maintaining power consumption under 40 watts.
According to Kioxia, practical applications for this module are expected in data center and edge use cases that require high-speed, high-capacity flash memory, particularly for advanced AI workloads and real-time data processing closer to the user in MEC architectures. The prototype demonstrates the combination of 5 TB capacity, 64 GB/s bandwidth, and energy-efficient operation, which is positioned to support new AI and edge computing demands.
Source: Kioxia
