SK Hynix, the South Korean memory and semiconductor maker, recently announced that it has begun mass production of 192GB SOCAMM2 memory. The product is described by the maker as a next generation memory module standard, based on the 1cnm process (6th generation 10nm technology), LPDDR5X low-power DRAM.
SOCAMM2, short for Small Outline Compression Attached Memory Module 2, is an AI-specific, server-optimised memory module based on the LPDDR memory standard. The new memory is designed primarily for the NVIDIA Vera Rubin platform, which is slated for a Q3 2026 launch and expected to replace the GPU brand’s current Blackwell architecture. In practice, the new memory standard also offers a slim form factor and high scalability, while its compression connector enhances siignal integrity, as well as allow for easy module replacement.
“By supplying the 192GB SOCAMM2, SK Hynix has established a new standard for AI memory performance,” Justin Kim, President & Head of AI Infra, CMO, SK Hynix said. “We will solidify our position as the most trusted AI memory solution provider, through close collaboration with our global AI customers.”
This is not the first time SK Hynix has used its 1cnm process. Back in March this year, the memory maker also revealed the world’s first 1c-based LPDDR6 RAM chip. Performance-wise, the memory manufacturer says that its new LPDDR6 DRAM is reportedly 33% faster and 20% more power efficient than the current LPDDR5X memory standard, with operating speeds of up to 10.7Gb/s and beyond. Again, it is designed for mobile devices, including smartphones, tablets, and PC gaming handhelds.
On a related note, SK Hynix isn’t the only memory maker getting into the SOCAMM2 space; Samsung is as well. The rival South Korean company says that it is already supplying its customers samples of the memory, with the brand saying that its own SOCAMM2 can deliver more than 70% better power efficiency and up to 2.6x higher bandwidth, compared to DDR-based server memory.
(Source: SK Hynix, Videocardz)
