The Department of Energy, Intel, HPE, and Argonne National Laboratory are working together to bring the most potent supercomputer online by the end of this year.
The Lab claims that “Aurora will theoretically be capable of delivering more than two exaflops of computing power, or more than 2 billion billion calculations per second.”
The astounding performance numbers are the result of thousands of chips, but with quantum computing appearing to be close at hand, this may very well be a one-of-a-kind supercomputer that opens the door to the future.
The supercomputer is made up of 10,624 rectangular blades weighing 70 pounds, with the final one being successfully placed on June 22. 166 racks, each holding 64 blades, are utilized to store the blades. The Argonne claims that Aurora takes up a footprint the size of two varsity basketball courts in its ALCF data center, spread out over eight rows.
Each blade contains memory, networking, cooling, two Intel Xeon Max Series CPUs, six Intel Max Series GPUs, and other components.
According to Susan Coghlan, project director for Aurora at the ALCF, “We’re looking forward to putting Aurora through its paces to make sure everything works as intended before we turn the system over to the broader scientific community.”
63,744 Intel Data Centre GPU Max Series “Ponte Vecchio” and 21,248 Intel Xeon CPU Max Series “Sapphire Rapids” processors are also included in the top numbers, along with more than 1,024 storage nodes with a combined capacity of 220 petabytes and a maximum bandwidth of 31 terabytes per second.
All of this, according to Intel and its partners, should make supercomputing more effective and potent when it comes to running models to address issues like combating climate change and other anthropogenic issues.
While we work towards acceptance testing, we’re going to be using Aurora to train some large-scale open-source generative AI models for science, said Rick Stevens, associate laboratory director at Argonne National Laboratory.