Instead of consuming massive power to move, and wait to move, data to the computing unit, UPMEM is bringing the computing unit to the data. An explanation brief and to the point of Processing in-Memory (PIM). Thanks Bryon Moyer for this write up, In-Memory Computing, No Fewer than Four Approaches, on EEjournal, and for not seeing « any red flag ».

UPMEM Processing in-Memory accelerator includes standard PIM-DIMM modules, which contain DPUs (UPMEM Patented microprocessors) system-based PIM-DRAM chips, and SDK for smooth integration. UPMEM PIM on DRAM technology provides the first ultra-efficient scalable and programmable acceleration solution for Big Data & AI applications. Use cases such as genomics, search, database… have demonstrated PIM accelerators’ reliability to boost 20x data-intensive apps and improve 10x energy efficiency at marginal hardware and application porting costs.




Welcome to contact us for more insights !

Contacts: contact@localhost