“The Matrix Core moves into hardware the matrix operations for supercomputing workloads – that provides a 7x performance improvement,” said McCready.
The chips use the latest PCI Express connections to servers and also bring out AMD’s on-die Infinity fabric to allow up to four GPU accelerator cards can be connected together in a topological cube.
”Four chips can be connected using our Infinity architecture with coherency rather than PCIe Gen4 for a fully connected cube. We physically implement it with a bridge card that goes across the top of the rack with 576Gb/s,” he said. This provides up to 340 GB/s of aggregate bandwidth per card.
Memory bandwidth is also important. “We also get 20 percent memory improvement using HBM2,” said McCready. The accelerator cards support 32GBytes of HBM2 memory at a clock rate of 1.2 GHz and delivers 1.23 TB/s of memory bandwidth to support large data sets and help eliminate bottlenecks in moving data in and out of memory.
The chip is being used by Dell, Gigabyte, HPE and Supermicro for cards alongside AMD’s EPYC Gen2 processor.
“Customers use HPE Apollo systems for purpose-built capabilities and performance to tackle a range of complex, data-intensive workloads across high-performance computing (HPC), deep learning and analytics,” said Bill Mannel, vice president and general manager, HPC at HPE. “With the introduction of the new HPE Apollo 6500 Gen10 Plus system, we are further advancing our portfolio to improve workload performance by supporting the new AMD Instinct MI100 accelerator, which enables greater connectivity and data processing, alongside the 2nd Gen AMD EPYC™ processor. We look forward to continuing our collaboration with AMD to expand our offerings with