Arm’s Fugaku tops supercomputer rankings again
Fugaku, the Arm technology-based supercomputer jointly developed by RIKEN and Fujitsu, has been awarded the number one spot on a Top500 list for the second time in a row.
The 56th edition of the Top500 supercomputer listings project, which began in 1993, saw improvements in performance flatten this year, even as speeds have reached astonishing new peaks.
Fugaku grew its benchmark performance to 442 petaflops, a modest increase from the 416 petaflops the system achieved when it debuted in June – but still three times ahead of the number two in the list, the IBM Summit.
A petaflop is one thousand trillion, or one quadrillion, operations per second. Petaflops are rarely used to measure a single computer's performance, since only the fastest supercomputers run at more than one petaflops. Therefore, petaflops are more often used when calculating the processing power of multiple computers. The Fugaku’s 442 petaflops means it can perform 442 quadrillion computations per second.
The entry level to the list moved up to 1.32 petaflops on the High Performance Linpack (HPL) benchmark, a small increase from the 1.23 petaflops recorded in the June 2020 rankings. In a similar vein, the aggregate performance of all 500 systems grew from 2.22 exaflops in June to just 2.43 exaflops on the latest list. Likewise, average concurrency per system barely increased at all, growing from 145,363 cores six months ago to 145,465 cores in the current list.
The notable developments in the top 10 include two new systems, as well as a new highwater mark set by the top-ranked Fugaku supercomputer, which uses Arm chips. Thanks to additional hardware, Fugaku grew its HPL performance to 442 petaflops, up from the 416 petaflops the system achieved when it debuted in June. More significantly, Fugaku increased its performance on the new mixed precision HPC-AI benchmark to 2.0 exaflops, besting its 1.4 exaflops mark recorded six months ago. These represents the first benchmark measurements above one exaflop for any precision on any type of hardware.
The current top three systems are:
- Fugaku remains at the top spot, growing its Arm A64FX capacity from 7,299,072 cores to 7,630,848 cores. A CPU can contain one or more processing units: each unit is called a core. It is common for computers to have two (dual), four (quad) or even more cores. CPUs with multiple cores have more power to run multiple programs at the same time. The additional hardware enabled its new world record 442 petaflops result on HPL. This puts it three times ahead of the number two system in the list. Fugaku was constructed by Fujitsu and is installed at the RIKEN Center for Computational Science in Kobe, Japan.
- Summit, an IBM-built system at the Oak Ridge National Laboratory (ORNL) in Tennessee, remains the fastest system in the US with a performance of 148.8 petaflops. Summit has 4,356 nodes, each one housing two 22-core Power9 CPUs and six NVIDIA Tesla V100 GPUs.
- Sierra, a system at the Lawrence Livermore National Laboratory in California, is ranked third with an HPL mark of 94.6 petaflops. Its architecture is very similar to that of Summit, with each of its 4,320 nodes equipped with two Power9 CPUs and four NVIDIA Tesla V100 GPUs.
NVIDIA’s $40bn acquisition of Arm in September is currently awaiting regulatory approval.
Arm’s Brent Gorda, senior director of HPC, infrastructure line of business, said: “The innovation and momentum around Arm in HPC is incredibly gratifying to see.
“HPC is undergoing a fundamental transformation and will require more innovation inside the socket and inside the system to meet today’s pressing problems. Our goal is to give our partners technology that is both powerful and flexible so that they can optimize their systems to pursue a broad spectrum of workloads - from research around Covid-19 to dynamic climate models - within their budgetary constraints.”