Wccftech reports that, The EuroHPC LUMI supercomputer, known as the "large unified modern infrastructure ", has chosen to configure AMD new generation of Xiaolong (EPYC) processors Radeon Instinct GPU acceleration cards.Lumi supercomputing, located in Kajaani, Finland, is scheduled to be fully operational in 2021 and is expected to achieve a peak performance of 550 p-flops.
HP said Lumi will use HPE's Cray ex architecture to support half of its performance. As part of the accelerated supercomputing platform of eurohpc GPU, it is expected to be fully operational by the end of 2021 and become the top 5 in the world.
Like Lumi, there are also El Capitan and frontier supercomputers, whose computing power is about 1.5 ~ 2.0 exaflops.
Forrest norrod, senior vice president and general manager of AMD data center and embedded systems division, said: through cooperation with HPE, Lumi supercomputing will promote scientific research in AI, weather forecast and drug development.
In addition to achieving powerful AI performance with the help of AMD Xiaolong CPU and radon insight GPU accelerator, HPE's Cray clusterstor E1000 scalable storage system also provides support for modeling, simulation and complex workload in AI.
In terms of network, Lumi is equipped with HPE slingshot scheme to meet the demand of data intensive workload for higher speed / congestion control. It is worth mentioning that this large-scale system also adopts advanced liquid cooling scheme to improve the overall energy saving and water use efficiency.
In addition, we speculate that Lumi supercomputing will use a new generation of AMD Xiaolong Milan processor based on Zen 3 architecture (a total of 200000 CPU cores), 30pb of extended storage, 7pb of flash memory, up to 2TB / s bandwidth, and 80pb of parallel file system capacity.
In terms of power consumption, the next generation of the cDNA architecture based GPU accelerator card will occupy the majority of Lumi supercomputing platform (performance 550 p-flops).
Hydropower accounts for 200 megawatts of about 8.5 megawatts of power, making it a carbon-neutral supercomputing system. According to these figures, the peak efficiency per watt of the system is about 44.1 G-Flops.. The
Finally, Lumi supercomputing covers more than 150 square meters and is expected to cost 144.5 million euros.
Still, in the news, HPE will build a strong Pawsey overcalculation for Australia based on the Cray EC architecture and the AMD new generation of Xiaolong acceleration programs. A $4.8 billion contract, with a 50 P-Flops. peak performance