NVIDIA commissions DRAM giants to develop SoCAMM: new DRAM for next-gen AI GPUs

NVIDIA commissioned Samsung Electronics, SK hynix, and Micron to develop SoCAMM: a new memory module standard conceived by NVIDIA, with Micron reportedly the first to receive approval for mass production... putting the US-based memory manufacturer ahead of its South Korean rivals.

In new South Korean media reports picked up by insider @Jukenlosreve, we're hearing that the new SoCAMM standard is a memory module conceived by NVIDIA, consisting of 16-stacked LPDDR5X chips bundled in groups of four.

HBM is a DRAM that supports GPUs in AI accelerators, but SoCAMM is a new DRAM that attaches to the CPU, and while it connects to the CPU, the primary role of SoCAMM is to provide auxiliary support to make sure AI accelerates reach their peak performance. We should expect to see SoCAMM make an appearance in NVIDIA's next-gen AI GPUs -- Rubin -- which will be dropping in 2026.

Unlike HBM, which connects DRAM by vertically drilling through it, SoCAMM is manufactured using a wire bonding method, connecting the 16 chips with copper wires. Copper has high thermal conductivity, with its main advantage in minimizing heat generation from each individual DRAM chip. Micron has said that the power efficiency of its latest low-power DRAM is 20% higher than its competitors.

NVIDIA's next-gen AI server (powered by Rubin AI GPUs and Vera CPUs) will be using 4 x SoCAMM memory modules, which calculated by the number of LPDDR5X memory chips, will add up to 256 chips. The industry points to Micron's late adoption of EUV lithography equipment as to why it could supply SoCAMM memory chips earlier than SK hynix or Samsung, adds the new report.

This means that unlike its DRAM manufacturer competitors, which boost DRAM performance with EUV, US-based Micron used low-heat generation tech through innovations through its design structure while improving memory performance along the way... a win-win.