- Kioxie develops a 5 TB flash memory prototype with a bandwidth of 64 GB / s for IT Edge
- The prototype memory module demonstrates the design of the marguerite chain by maintaining the bandwidth to higher capacities
- The new module uses PAM4 signaling and controller pre-festive to improve flash performance
Kioxia has developed a new type of flash prototype memory module which offers both a large capacity and a high bandwidth.
The module was created as part of the R&D project for improving Japan information and communication systems.
The module offers 5 to storage and a transfer rate of 64 GB / s, positioning it as a potential alternative or a supplement to the conventional dram in data processing. Kioxia said the design addresses the long -standing compromise between bandwidth and the capacity found in DRAM modules by organizing flash memory fleas in a chain of marguerites rather than a bus connection.
Flash pearls
Each flash unit, described as a pearl, is connected in sequence with its own controller. The design of the marguerite chain allows the ability to develop without degrading the bandwidth, which was a limitation in previous architectures.
The company has also developed high-speed transcenders capable of flow of 128 Gbit / s. These use the PAM4 modulation, which codes for two bits of information per signal, doubling the effective bandwidth compared to traditional binary signaling.
In addition to this, the new controller features such as pre-sample help reduce reading latency, while low-amplitude signaling and distortion correction increase the speed of the memory interface to 4.0 Gbit / s.
Kioxia reported that the prototype consumes less than 40 W when it operates a full bandwidth on a connection to eight PCIe Gen 6.
The combination of powerful efficiency and flow is presented as a step towards the more practical meeting in roles with high memory intensity usually filled by DRAM.
The work is targeted on servers in mobile IDGE IT environments, where 5G and 6G networks create the need to process information closer to the final devices.
The workloads IoT and IA in real time arouse the interest of memory systems that balance the cost, capacity and performance.
Kioxia said it planned to continue to develop this technology towards marketing, with potential uses in IoT, megadroned analysis and advanced AI models, including generative AI.
