Samsung's HBM-PIM to address memory/power challenges for HPC, AI computing

Samsung's HBM-PIM integrates into existing systems and does not require any hardware or software changes. The new architecture, when applied to the company's existing HBM2 Aquabolt solution, is claimed to deliver over twice the system performance whilst reducing energy consumption by over 70 percent.

Samsung has developed High Bandwidth Memory (HBM), the industry's first processing-in-memory (PIM) solution with artificial intelligence (AI) processing power, the South Korean technology giant announced on Wednesday.

The HBM-PIM solution will accelerate large-scale processing in data centres, high performance computing (HPC) systems and AI-enabled mobile applications.

While most of today's computing systems are based on the von Neumann architecture, which uses separate processor and memory units to carry out millions of intricate data processing tasks, the news solution brings processing power directly to where the data is stored, enabling parallel processing and minimizing data movement. This is achieved by placing a DRAM-optimized AI engine inside each memory bank, a storage sub-unit.

Samsung's HBM-PIM integrates into existing systems and does not require any hardware or software changes. The new architecture, when applied to the company's existing HBM2 Aquabolt solution, is claimed to deliver over twice the system performance whilst reducing energy consumption by over 70 percent.

"Our groundbreaking HBM-PIM is the industry's first programmable PIM solution tailored for diverse AI-driven workloads such as HPC, training and inference. We plan to build upon this breakthrough by further collaborating with AI solution providers for even more advanced PIM-powered applications," said Kwangil Park, senior vice president of Memory Product Planning at Samsung Electronics.

Samsung's paper on the HBM-PIM will be presented at the 2021 edition of the International Solid-State Circuits Conference (ISSCC), an annual conference for the presentation of advances in solid-state circuits and systems-on-a-chip, held virtually through February 22.