New 3D-stacked memory tech seeks to dethrone HBM for AI inference — d-Matrix claims 3DIMC will be 10x faster and 10x more efficient

Memory startup d-Matrix is introducing 3D stacked memory technology, 3DIMC, as a faster and more efficient alternative to HBM for AI inference tasks. The company's Pavehawk 3DIMC is designed to perform computations within the memory itself, offering improved latency, bandwidth, and efficiency gains. d-Matrix is already looking ahead to its next-gen Raptor, claiming it will outpace HBM by 10x in inference tasks while using 90% less power. The move towards specialized hardware for specific computational tasks, like AI inference, is gaining traction in the tech industry. This new memory technology could provide a cost-effective alternative to HBM, which is currently produced by a limited number of companies and expected to see rising prices.