The Aurora supercomputer utilizes 21248 Xeon processors and 63744 graphics processors and is equivalent in processing power to the world’s fastest supercomputer. Intel has finally announced the complete specifications of the Aurora supercomputer, which was designed for the Argonne National Laboratory in the United States. The Intel Aurora supercomputer has long been delayed, but construction is finally nearing completion. The system, now powered by Intel’s Xeon CPU Max and Xeon GPU Max series, has reached an incredible processing power of two exaflops, while its initial goal was equivalent to one exaflop. Thus, Aurora is equivalent in power to the current world’s fastest supercomputer, Frontier, equipped with AMD processors. Intel announced that Aurora will have a total of 10,624 nodes, including 2,124 Xeon processors based on the Sapphire Rapids-SP family, and 3,744 graphics processors based on Ponte Vecchio. This extremely powerful supercomputer utilizes 2.12 petabytes per second of remarkable bandwidth and a maximum dual-bandwidth of 0.69 petabytes, which is a mind-boggling figure.
According to WccfTech, the Aurora supercomputer is equipped with 10.9 petabytes of system DRAM, 1.36 petabytes of processor bandwidth memory, and 8.16 petabytes of graphics processor bandwidth memory. The system DRAM achieves a maximum bandwidth of 9.95 petabytes per second, processor DRAM achieves a maximum bandwidth of 30.5 petabytes per second, and graphics processor DRAM achieves a maximum bandwidth of 208.9 petabytes per second. The system also utilizes 230 petabytes of storage memory, which achieves a maximum bandwidth of 31 terabytes per second. In total, it is configured with 1024 nodes.
The Aurora supercomputer will begin operating later this year with a maximum processing power of 2 exaflops. The system also utilizes the latest Aurora artificial intelligence model and provides one trillion parameters for practical applications.