TOP Fugaku About Fugaku
The system configuration of the supercomputer Fugaku is described below.
Total number of nodes
Peak Theoretical Performance
Fugaku can be run in a normal mode (CPU clock speed of 2 GHz) and a boost mode (CPU clock speed of 2.2 GHz), and the peak theoretical performance in each mode is summarized in the table below.
The high memory bandwidth is also one of the features of Fugaku.
Normal mode（CPU clock speed: 2 GHz）
Boost mode (CPU clock speed: 2.2 GHz）
|Total memory||4.85 PiB|
|Total memory bandwidth||163 PB/s|
Single node performance
The single node performance is summarized below. The unique chip is based on the Arm instruction set architecture.
|Instruction set architecture||Armv8.2-A SVE 512 bit
Fujitsu extension: hardware barrier, sector cache, prefetch
|Number of core||48 + 2 assistant cores
4 CMG (Core Memory Group, NUMA node)
Normal mode （CPU clock speed: 2 GHz）
|Double precision: 3.072 TF; single precision: 6.144 TF; half-precision: 12.288 TF|
Boost mode（CPU clock speed: 2.2 GHz）
|Double precision: 3.3792 TF; single precision: 6.7584 TF; half-precision: 13.5168 TF|
|Cache *1 *2||L1D/core: 64 KiB, 4way, 256 GB/s (load), 128 GB/s (store)
L2/CMG: 8 MiB, 16way
L2/node: 4 TB/s (load), 2 TB/s (store)
L2/core: 128 GB/s (load), 64 GB/s (store)
|Memory||HBM2 32 GiB, 1024 GB/s|
|Interconnect||Tofu Interconnect D (28 Gbps x 2 lane x 10 port)|
|I/O||PCIe Gen3 x16|
|Technology||7 nm FinFET|
- 1 Cache performance is with the CPU clock speed of 2 GHz
- 2 Please refer to GitHub for details
CPU-Die (Image courtesy of Fujitsu)
Tofu Interconnect D
About 200,000 Tofu cables connect the CPUs of Fugaku. About half of them are fiber optic cables that run between racks, the total length of which is about 900 km.
|Number||Total length (m)|
|Fiber optic cables||
Another approx. 10,000 cables connect the racks to the storage, etc.
- LLIO (Lightweight Layered IO-Accelerator)
LLIO is a file system dedicated to the job execution area.
Providing the following three types of areas to the job.
・Node Temporary Area
・Shared Temporary Area
・Cache Area of 2nd Layer Storage
- FEFS (Fujitsu Exabyte File System)
FEFS is a large shared area used by users and jobs.
Luster-based file system.
- Commercial cloud storage services
|Compiler||Fortran 2008 and Fortran 2018
C11 with GNU and Clang extensions
C++14 and C++17 with GNU and Clang extensions
OpenMP 4.5 and OpenMP 5.0
|Parallel programing||XcalableMP [Details of XcalableMP (PDF 535 KB)]
FDPS [Details of FDPS (PDF 260 KB)]
|Script language||Python / Numpy / Scipy, Ruby|
|Numerical library||BLAS, LAPACK, ScaLAPACK
SSL II (Fujitsu)
EigenExa, Batched BLAS, 2.5D-PDGEMM
|Open-source management tool||Spack [Details of Spack (PDF 355 KB)]|
|Container, virtual machine||Singularity, KVM|
|OS||Red Hat Enterprise Linux 8
McKernel [Details of McKernel (PDF 641 KB)]
|MPI||Fujitsu MPI (Based on OpenMPI),
MPICH-Tofu (Based on MPICH) [Details of MPICH-Tofu (PDF 404 KB)]
DTF (Data Transfer Framework) [Details of DTF (PDF 220 KB)]