The silicon industry has arrived at a consensus that the HBM pathway to unlock incremental AI compute is stalling, especially as converting silicon wafers into an HBM stack yields far less usable capacity than commodity DRAM, prompting the wider industry to start experimenting with SRAM-only decode, Processor-In-Memory (PIM) within LPDDR, CXL pooling, and 3D DRAM. Of course, it is the 3D DRAM architecture that is considered somewhat of a panacea at the moment, as highlighted by AMD’s recent admission as to its phenomenal energy efficiency. Even so, odious hurdles continue to prevent the commercialization of this promising technology. AMD has […]
Read full article at https://wccftech.com/amd-says-a-hybrid-bonded-3d-dram-is-17x-more-energy-efficient-than-an-hbm-stack-that-relies-on-microbumps-but-thermal-hurdles-block-commercialization/