PIM Improves GPU Performance
Samsung Electronics has succeeded in demonstrating the performance of a processing in memory (PIM) chip by installing it on a graphics processing unit (GPU) accelerator card from AMD.
PIM refers to the integration of a processor with random access memory (RAM) on a single chip. This technology is expected to contribute to improving the performance of mammoth-sized artificial intelligence (AI).
Samsung announced pn Oct. 20 that it has succeeded in operating an MI-100 accelerator card, a commercial GPU from AMD, which was loaded with a HBM-PIM chip. HBM refers to a high bandwidth memory.
The HBM-PIM chip doubled the performance of AMD’s GPU accelerator card and cut energy consumption by about 50 percent on average compared to other GPU accelerators without an HBM-PIM chip, Samsung Electronics said. One-year energy consumption by the GPU accelerator equipped with an HBM-PIM dropped by about 2,100 GWh compared to GPU accelerators equipped with an HBM only.
“HBM-PIM technology is the industry’s first customized memory solution for the huge AI sector,” said Park Chul-min, managing director of Samsung Electronics. “We will integrate HBM-PIM technology with CXL-PNM solutions through an integrated software standardization process."