JEDEC Solid State Technology Association has released the JESD209-6 Low Power Double Data Rate 6 (LPDDR6) standard, designed to enhance memory speed, power efficiency, and security for mobile devices and AI applications. The standard, available for download on the JEDEC website, introduces features to support high-performance workloads and power-conscious systems.
LPDDR6 features a dual sub-channel architecture with 2 sub-channels per die, each with 12 data signal lines and 4 command/address signals to optimize channel performance and reduce ball count. It supports a 32-byte access granularity, flexible data access, and on-the-fly burst length control for 32B and 64B access. A static efficiency mode supports high-capacity memory configurations, while dynamic write NT-ODT adjusts on-die termination based on workload demands to improve signal integrity.
For power efficiency, LPDDR6 operates with a lower voltage VDD2 supply compared to LPDDR5, mandating two VDD2 supplies. It includes alternating clock command inputs, Dynamic Voltage Frequency Scaling for Low Power (DVFSL) to reduce power during low-frequency operation, a dynamic efficiency mode for low-bandwidth use cases, and support for partial self and active refresh to minimize refresh power usage.
Security and reliability enhancements include Per Row Activation Counting (PRAC) for data integrity, a carve-out Meta mode for allocating memory regions to critical tasks, a programmable link protection scheme, on-die error correction code (ECC), and support for command/address parity, error scrubbing, and memory built-in self-test (MBIST).
Mian Quddus, JEDEC Chairman, noted that LPDDR6 balances power efficiency, security, and performance for mobile devices and AI applications. Industry leaders, including Advantest, Cadence, Keysight Technologies, MediaTek, Micron Technology, Qualcomm Technologies, Samsung, SK hynix, and Synopsys, expressed support for the standard, highlighting its role in enabling faster, more efficient, and reliable memory solutions for mobile, automotive, edge AI, and data center applications.
