Samsung develops industry’s first high-bandwidth memory with AI processing power – Samsung Global Newsroom

The new architecture delivers more than twice the system performance
and reduces energy consumption by more than 70%

Samsung Electronics, the world leader in advanced memory technology, announced today that it has developed the industry’s first High Bandwidth Memory (HBM) integrated with artificial intelligence (AI) processing power – the HBM-PIM. The new PIM (memory processing) architecture provides powerful, high-performance memory computer capabilities to accelerate large-scale processing in data centers, high-performance computing (HPC) systems, and mobile AI applications.

Kwangil Park, senior vice president of memory product planning at Samsung Electronics, said: “Our groundbreaking HBM-PIM is the industry’s first programmable PIM solution designed for diverse AI-driven workloads such as HPC, training and diversion. We plan to build on this breakthrough by further partnering with AI solution providers for even more advanced PIM-powered applications. ”

Rick Stevens, co-laboratory director for computer, environmental and life sciences at Argonne, said: “I am delighted to see that Samsung is addressing the memory bandwidth / power challenges for HPC and AI. HBM-PIM design has shown impressive performance and power gains on key classes of AI applications, so we look forward to working together to evaluate its performance on additional issues of interest to the Argonne National Laboratory. ”

Most of the current computer systems are based on the von Neumann architecture, which uses separate processing and memory units to perform millions of complex data processing tasks. This sequential processing approach requires that data be constantly moved back and forth, resulting in a system that slows the bottleneck, especially when handling the increasing amount of data.

Instead, the HBM-PIM brings processing power directly to where the data is stored by placing a DRAM-optimized AI engine within each memory bank – a storage unit – that enables parallel processing and minimizes data movement. When applied to Samsung’s existing HBM2 Aquabolt solution, the new architecture can deliver more than twice the system performance, while reducing power consumption by more than 70%. The HBM-PIM also does not require any changes to hardware or software, enabling faster existing systems.

Samsung’s article on the HBM-PIM was chosen to be presented at the famous International Solid-State Circuits Virtual Conference (ISSCC) which will be held until 22 February. Samsung’s HBM PIM is now being tested in AI accelerators by leading AI solution partners, with all validations expected to be completed within the first half of this year.

Source