Feature: Power management
Low power double data rate (LPDDR),
which was originally developed for battery- operated mobile devices, off ers compelling
power effi ciency advantages for bandwidth- intensive data processing
SOCAMM2 brings LPDDR’s power
efficiency advantages into AI servers through ultra-compact, low-profile modules that preserve short signal paths and maintain signal margins necessary for optimal performance at LPDDR5X data rates. The “compression attach” technology means the back of the SOCAMM2 module provides the contact area for the module and provides the signal integrity for LPDDR’s low-voltage operation. In addition, the module lies flush on the motherboard, which supports liquid cooling implementations. From a performance standpoint, SOCAMM2 is capable of supporting
Figure 1: SOCAMM2 chips
high data throughput required by AI workloads, helping alleviate the growing bottlenecks between memory and compute. Importantly, the architecture provides modularity and serviceability that data centres demand in servers. This is possible because SOCAMM2 is a removable module unlike traditional soldered-down LPDDR solutions. This means it can be upgraded, repaired, or reconfigured to reduce downtime and improve lifecycle flexibility without replacing the entire mainboard. Power delivery is another key
differentiator incorporated into the module itself. With on-module voltage
regulation, SOCAMM2 can provide stable power at low voltage to the LPDDR memory devices. An SPD Hub IC on the SOCAMM2 module supports telemetry, identification and configuration.
The future of AI workloads Approaches like SOCAMM2 reflect a broader shift in how memory is integrated and deployed in next- generation AI infrastructure. As AI workloads continue to grow in complexity and scale, data centres are under increasing pressure to deliver higher bandwidth, improved energy efficiency, and more scalable memory subsystems capable of supporting evolving compute demands. Bringing LPDDR into server
environments represents an important step toward addressing many of these challenges. By enabling more modular and flexible AI server designs, emerging architectures like SOCAMM2 support long-term demands of AI training and inference workloads and help bridge the gap between mobile memory innovation and AI server requirements.
www.electronicsworld.co.uk September 2026 21
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44
