— The legacy processor-centric solutions are fundamentally limited by the data bus thoughput that does not scale with the memory demand/capacity ; demonstrating dominant data movement costs, the « memory wall » issue, resulting in a dramatic loss in performance, energy and money.
Upmem PIM thousands of proprietary innovative data processors (DPU) are uniquely positionned in the memory, next to the data, and execute the data-intensive operations. Those DPUs are under the control of the high level application running on the main CPU, that ensures tasks orchestration.
20x faster/cheaper computing on multiple workloads, 15x more energy efficient, 17x better TCO on average
The number of PIM units grows proportionnally with the memory capacity, with no change in memory processes.
Easy integration with high level applications. PIM units are programmable with standard tools and skillset.
Compatible with existing servers.
First product based on UPMEM technology for the datacenter
Best performance per area processor
UPMEM DPU processor is delivering the best performance per mm2, compare to the latest RISC and CISC processors of the market, and is the only processor that can be routed in a memory chip.
A technical exploit
Designing a RISC processor that can be fabbed with a memory process is a true technical exploit. It is opening the door to a new generation of applications leveraging the power of PIM.
Co-processing model/smooth integration
The model of programmable co-processors is versatile and enables a large number of use cases. The PIM units can be programmed one by one or by group, and process their local data independently from each other.
C coding and no OS
The UPMEM DPU processor is programmable in C, thanks to its Software Development Kit (SDK), and does not require an OS.