Ekaterina Alexandrova | 02/18/2021
Samsung Electronics today announced the development of the industry’s first high-bandwidth memory (HBM) integrated with artificial intelligence computing power — HBM-PIM.
The new processing architecture (PIM) leverages the power of AI to accelerate the processing of large amounts of data in data centers, high performance computing (HPC) systems, and
AI-enabled mobile applications.
“We plan to develop this technology through further collaboration with AI vendors to create even more advanced PIM-based products,” said Kwangil Park, Sr.
VP of Memory Product Planning at Samsung Electronics.
“The HBM-PIM design has demonstrated impressive performance and efficiency gains in critical types of AI computing, so we look forward to evaluating it.
the ability to solve additional problems of interest to Argonne National Laboratory, “commented Rick Stevens, Deputy Director of Argonne National Laboratory
US Department of Energy Computing, Environment, and Life Sciences.
Most modern computing systems are based on the von Neumann architecture, which uses separate blocks of processor and memory to perform millions of complex data processing tasks.
This approach to sequential processing requires data to constantly move back and forth, which leads to system slowdowns, especially when processing ever-growing volumes.
Instead, HBM-PIM transfers processing power directly to where the data is stored, since a DRAM-optimized AI engine is placed in each memory bank (storage sub-block).
providing parallel processing and minimizing data movement. With Samsung’s HBM2 Aquabolt solution, the new architecture can double system performance while reducing
energy consumption by more than 70%. HBM-PIM also does not require any hardware or software changes, allowing for faster integration into existing systems.
Samsung’s paper on HBM-PIM will be presented at the International Solid State Circuits Conference (ISSCC) on February 22nd. Samsung HBM-PIM is currently being tested by leading partners
on AI solutions in artificial intelligence accelerators. This work is expected to be completed in the first half of 2021.
Samsung Artificial Intelligence AI