The historical past of the computing business is certainly one of fixed progress. Processors get sooner, storage will get cheaper, and reminiscence will get denser. We see the repercussions of this development by means of all facets of society, and that extends to the highest as nicely, the place nationwide governments proceed to put money into larger and higher supercomputers. One half technological necessity and one half technological race, the exascale period of supercomputers is about to start, as orders for the primary exaFLOP-capable at the moment are going out. It’s solely becoming then that this morning the United States Department of Energy is saying the contract for his or her quickest supercomputer but, the Frontier system, which will likely be constructed by Cray and AMD.
Frontier is deliberate for supply in 2021, and when it’s activated it would turn out to be the second and strongest of the US DOE’s two deliberate 2021 exascale methods, with efficiency anticipated to succeed in 1.5 exaFLOPS. The bold system received’t come cheaply, nonetheless; with a price ticket of over 500 million {dollars} for the system alone – and one other 100 million {dollars} for R&D – Frontier is among the many costliest supercomputers ever ordered by the US Department of Energy.
The new supercomputer is being constructed as a part of the US DOE’s CORAL-2 program for supercomputers, with Frontier scheduled to exchange Oak Ridge National Laboratory’s present Summit supercomputer. Summit is the present reigning champion within the supercomputer world, with 200 petaFLOPS of efficiency, and accordingly the US DOE and Oak Ridge are aiming to considerably enhance on its efficiency for the brand new laptop. All informed, Frontier ought to be capable of ship over 7x the efficiency of Summit, and is anticipated to be the quickest supercomputer on the planet as soon as it’s activated.
Like Summit (and Titan earlier than it), Frontier is an open science system, that means that it’s out there to educational researchers to run simulations and experiments on. Accordingly, the lab is anticipating the supercomputer for use for a variety of tasks throughout quite a few disciplines, together with not solely conventional modeling and simulation duties, but additionally extra data-driven strategies for synthetic intelligence and information analytics. In reality the latter is a bit of latest floor for the lab and the system’s eventual customers; simply as we’ve seen within the enterprise area over the previous few years, neural network-based AI is changing into an more and more common method to unravel issues and extract evaluation from massive datasets, and now researchers are taking a look at methods to refine these strategies from the current-generation methods and apply them to exascale-level tasks.
US Department of Energy Supercomputers | |||||
Frontier | Aurora | Summit | |||
CPU Architecture | AMD EPYC (Future Zen) |
Intel Xeon Scalable | IBM POWER9 | ||
GPU Architecture | Radeon Instinct | Intel Xe | NVIDIA Volta | ||
Performance (RPEAK) | 1.5 EFLOPS | 1 EFLOPS | 200 PFLOPS | ||
Power Consumption | ~30MW | N/A | 13MW | ||
Nodes | 100 Cabinets | N/A | 3,400 | ||
Laboratory | Oak Ridge | Argonne | Oak Ridge | ||
Vendor | Cray | Intel | IBM | ||
Year | 2021 | 2021 | 2018 |
Frontier: Powered by Cray & AMD
Officially, the prime contractor for Frontier will likely be Cray. But wanting on the specs, you may be excused for considering it was AMD. Cray for its half is partnering with the chipmaker for the system, and consequently AMD is offering many of the core {hardware} for the brand new supercomputer. Designed as a next-generation CPU + accelerator system, with a mixture of CPUs and GPUs doing the heavy compute work, AMD will likely be supplying each the CPUs and GPUs for Frontier. And because the precept processor supplier, AMD may even be taking over quite a lot of the accountability for growing the software program stack as nicely, with the corporate working with Cray to develop an enhanced model of their ROCm setting to greatest extract efficiency from the huge cluster of CPUs and…