In an electrifying revelation, Intel has unfolded the full specifications of its highly anticipated Aurora supercomputer, destined for the Argonne National Laboratory in the United States.
After a suspense-filled wait, the tech titan has delivered a system that leapfrogs expectations, and quite possibly, the competition.
The Aurora supercomputer has been the stuff of legends for quite some time now, with its long-anticipated arrival marked by tantalizing leaks and speculation.
With its unveiling at the ISC High Performance 2023 event in Hamburg, Germany, it is now clear that the wait was worth it.
Today at #ISC23, @codenative announced that Intel has completed the physical delivery of more than 10,000 blades to @argonne for the Aurora supercomputer! The 2 ExaFLOP supercomputer will house more than 20,000 CPUs and more than 60,000 GPUs. https://t.co/rZMuy2SOTb pic.twitter.com/LORSeJOcgn
— Intel Graphics (@IntelGraphics) May 22, 2023
Monumental Collaboration: The Power Behind Aurora
Intel has armed the Aurora supercomputer with a staggering 21,248 Xeon CPUs and 63,744 GPUs, surpassing its initial 1 Exaflop target and entering the 2 Exaflop territory, placing it shoulder-to-shoulder with AMD’s Frontier supercomputer.
Under the leadership of Argonne, this colossal project is the outcome of an extensive collaboration that spans continents.
The collaboration includes industry heavyweights such as Intel, HPE, and a slew of Department of Energy laboratories, universities, non-profits, and international partners like RIKEN.
As of now, Intel has physically delivered over 10,000 blades for the Aurora supercomputer, with the system being assembled using HPE Cray EX supercomputers.
Aurora’s early performance results have been extraordinary – the supercomputer demonstrated up to 2x performance over AMD MI250 GPUs, a 20% improvement over the H100 on the QMPACK quantum mechanical application, and near-linear scaling up to hundreds of nodes.
This puts the supercomputer in a league of its own, offering more than 2 exaflops of peak double-precision compute performance, a feat unmatched till now.
Intel did disclose some more details on Aurora.
– 10,624 Nodes
– 21,248 CPUs
– 63,744 GPUs
– lots of CPU/GPU memory BW
– lots of fabric BW
– lots of storage
The hardware has been built, now it’s time to make sure the applications scale well on it. pic.twitter.com/ZxCxEKLyeX
— Andreas Schilling (@aschilling) May 22, 2023
Intel Aurora: Staggering Storage and Memory Capabilities
The Aurora is more than just about processing power – as the system boasts 10.9 PB of DDR5 system DRAM, 1.36 PB of HBM capacity through the CPUs, and 8.16 PB of HBM capacity through the GPUs.
Furthermore, the system DRAM reaches a peak bandwidth of 5.95 PB/s, while the CPU HBM and the GPU HBM achieve a peak bandwidth of 30.5 PB/s and 208.9 PB/s, respectively.
In terms of storage, Aurora is equipped with an impressive 230 PB DAOS capacity that runs at a peak bandwidth of 31 TB/s, configured in a total of 1024 nodes.
As the cherry on top, Intel announced its brand-new Data Center GPU Max Subsystem, which will be incorporated in the Aurora.
The Max Series outperforms Nvidia’s H100 PCIe card by an average of 30% on diverse workloads, showcasing Intel’s commitment to pushing boundaries in supercomputing.
The Aurora supercomputer is expected to launch later this year, bringing with it a new era of unprecedented computational power.
As it prepares to step into the limelight, the Aurora gen AI model, offering 1 trillion parameters for scientific applications, adds to the growing excitement.
As the boundaries between science fiction and reality blur, the Aurora supercomputer stands as a testament to the limitless potential of human innovation.
Love Hate Inu – Next Big Meme Coin
- First Web3 Vote to Earn Platform
- Latest Meme Coin to List on OKX
- Staking Rewards
- Vote on Current Topics and Earn $LHINU Tokens