Accelerating AI with
Samsung is accelerating the capabilities of AI by becoming the first company in the industry to integrate Processing-in-Memory (PIM) in High Bandwidth Memory (HBM) configurations. PIM is able to process some of the logic functions by integrating an AI engine called the Programmable Computing Unit (PCU) in the memory core. PIM will stimulate growth in the use of AI applications that require continuous performance improvements, such as mobile, data centers and HPC.
2X performance increase
with parallel processing
Samsung’s PIM can theoretically improve performance up to 4 times compared to existing memory solutions through the Programmable Computing Unit (PCU). Like multi-core processing in the CPU, the PCU enables parallel processing in memory to enhance performance. As an example of the benefits that PIM can bring, in AI applications such as speech recognition, PIM(Processing-In-Memory) showed a 2 times increase in performance compared to existing HBM.
Uses 70% less energy, even for intensive tasks
Power consumption is an important issue in AI applications that need to process huge amounts of data quickly. PIM responds to this need by reducing energy consumption by 70% in systems applied to PIM(Processing-In-Memory), compared to existing HBM. This solution is therefore suited to AI applications with high power consumption and offers the right conditions for demanding tasks.
Easy to adopt and offers
PIM can be applied without the need to change existing memory ecosystem environments, and can be integrated with HBM as well as LPDDR and GDDR memory. When it comes to AI applications, the incorporation of PIM enables a higher level of performance in a range of capabilities, especially those of speech recognition, translation and recommendation. PIM unlocks the power of AI, changing business and everyday life for the better.