The 57 th edition of the TOP500 saw little change in the top 10. The only new entry in the top 10 systems is the Perlmutter system at NERSC at the DOE Lawrence Berkeley National Laboratory. The machine is based on the HPE Cray ‘Shasta’ platform and a heterogeneous system with both GPU-accelerated and CPU-only nodes. Perlmutter achieved 64.6 Pflop/s, putting the supercomputer at number 5 in the latest version of the Top500 list.
The Japanese supercomputer Fugaku held onto the top spot on the list. A system co-developed by Riken and Fujitsu, Fugaku has an HPL benchmark score of 442 Pflop/s. The performance of Fugaku is approximately three times that of the Number two machine Summit. The machine is based on Fujitsu's custom ARM A64FX processor. What's more, in single or further reduced precision, which is often used in machine learning and AI, Fugaku's peak performance is actually above an exaflop. Such an achievement has caused some to introduce this machine as the first ‘Exascale’ supercomputer. Fugaku already demonstrated this new level of performance on the new HPL-AI benchmark with 2 Eflop/s.
During his keynote speech at ISC High Performance 2021 Digital conference Erich Strohmaier, senior scientist at the TOP500 discussed how the first two systems, Fugaku and Summit have such a lead over the other systems that they sweep the awards across all benchmarks excluding the Green500.
‘Fugaku has a comfortable lead over the next biggest systems in any of these lists. It shouldn't be a surprise when you look at the other benchmark rankings, that Fugaku is also the number one on the HPCG ranking as well as on the HPL AI ranking. Fugaku has achieved a clean sweep of the three size base rankings,’ states Strohmaier. Looking at the number two system, on the Top500, it's again the summit system, Nothing has changed there… It [Summit] is separated, well enough from the following systems, as well, that it also managed to do a clean sweep of all of the size based rankings, the HPCG and HPL AI benchmarks. Congratulations Summit.’
As Strohmaier notes during his presentation, ‘the third position on these lists is where the competition starts.’ On the HPCG benchmark the new system Perlmutter took the third spot. The Perlmutter system installed at the NERSC computing Centre at Lawrence Berkeley National Labs scored 1.91 HPCG-petaflops. Perlumtter is the number 5 system on the Top500.
The HPL-AI benchmark seeks to highlight the convergence of HPC and artificial intelligence (AI) workloads based on machine learning and deep learning by solving a system of linear equations using novel, mixed-precision algorithms that exploit modern hardware.
Much like the TOP500, RIKEN's Fugaku system is leading the pack here. As stated, this machine is considered by some to be the first "Exascale" supercomputer. Fugaku demonstrated this new level of performance on the new HPL-AI benchmark with 2 Exaflops. Once again Summit took the number two spot on this benchmark.
‘The HPL AI benchmark again has a different system listed as number three,’ notes Strohmaier. ‘The Selene system, an NVIDIA built system, has been on the number three spot before, was powerful enough on the HPL AI benchmark to actually hold off Perlumtter which is number four on this benchmark grading.’
When it comes to the Green500 there is a lot more turnover and Strohmaier puts this down to the fact that the list is more technology-based rather than size-based. This is because the Green500 looks at performance/power rather than sheer computational might in a specific benchmark.
The system to take the number one spot for the Green500 was MN-3 from Preferred Networks in Japan. Knocked from the top of the last list by NVIDIA DGX SuperPOD in November 2020, MN-3 is back to reclaim its crown in June 2021. This system relies on the MN-Core chip, an accelerator optimised for matrix arithmetic, as well as a Xeon Platinum 8260M processor. MN-3 achieved a 29.70 gigaflops/watt power efficiency and has a TOP500 ranking of 337.
HiPerGator AI of the University of Florida in the USA is now number two on the Green500 with a 29.52 gigaflops/watt power efficiency. The Nvidia machine boasts 138,880 cores, much more than any other machine in the Top5 of the Green500. Like many other systems on the list, this supercomputer utilises an AMD processor – specifically the AMD EPYC 7742. With an overall performance that outpaces most of the other competition on the Green500, it's no surprise this machine holds the 22nd spot on the TOP500 list.
Strohmaier noted that the HiPerGator AI system ‘missed the number one spot by less than 1 per cent power efficiency. It achieved 29.5 gigaflops/per watt.’
The Wilkes-3 system out of the University of Cambridge in the UK has achieved the number three spot. A Dell EMC machine, this supercomputer had an efficiency of 28.14 gigaflops/watt. Like HiPerGator AI, this system relies on an AMD EPYC processor – the AMD EPYC 7763. This system is ranked 101 on the TOP500 list.
The MeluXina machine of LuxProvide holds the number four spot on the Green500. It's also unique in that it is the only machine on the list from Luxembourg. Like many other machines, MeluXina is also using an EMD EPYC processor. This system's efficiency was clocked at 26.96 gigaflops/wat, and it is ranked 37 on the TOP500.
Rounding out the Top5 of the Green500 is NVIDIA DGX SuperPOD. Created by NVIDIA in the USA, this machine also relies on an EMD EPYC processor. Ranked 216 on the TOP500, this system achieved 26.20 gigaflops/watt of efficiency. Again, this machine has an AMD EPYC processor.
Outside of the Top5, an incredible development is the Perlmutter system. This supercomputer was the only new machine in the Top10 in the TOP500 rankings; it also claims the number 6 spot on the Green500.
