Fugaku node architecture
Fugaku is operated by RIKEN Center for Computational Science in Japan. This profile covers the documented compute inventory. Specifications are listed below; anything the operator has not published is itemised in the documented-fields list.
Published benchmark (TOP500 · June 2026)
- HPL rank
- #9
- HPL Rmax
- 442.01 PFlop/s
- Theoretical peak Rpeak
- 537.21 PFlop/s
- Ranked core count
- 7,630,848
- Measured power
- 29899 kW
- Energy efficiency
- 14.78 GFlop/s per watt
Cluster · A64FX compute
- Operator / site
- RIKEN Center for Computational Science
- Country
- Japan
- Profile scope
- Documented compute inventory
- Data checked
- 13 September 2026
- Manufacturer
- Fujitsu
- Ranked core count
- 7630848
- Theoretical peak (PFlop/s)
- 537.21
- Measured power (kW)
- 29899
- Energy efficiency (GFlop/s per watt)
- 14.783437573162981
- Operating system
- Red Hat Enterprise Linux 8 · McKernel lightweight kernel
- Network topology
- 6D mesh / torus
A64FX compute
- Compute nodes of this type
- 158976
- CPU model
- Fujitsu A64FX · 48 compute cores
- CPU sockets per node
- 1
- Cores per socket
- 48
- Physical CPU cores
- 48
- CPU note
- 48 compute cores plus 2 assistant cores, arranged in 4 Core Memory Groups (NUMA nodes)
- Host / unified memory
- 32 GiB HBM2
- Network interfaces per node
- 6
- Interface model
- Tofu Network Interface (Tofu interconnect D)
- Interface speed
- 6.8 GB/s per TNI · 40.8 GB/s per node (28 Gbps × 2 lanes × 10 ports)
- Form factor
- 2 nodes per CPU Memory Unit · 8 CMUs per Bunch of Blades (16 nodes) · 3 BoBs per shelf (48 nodes) · 8 shelves per rack (384 nodes) · 432 racks
Scope and uncertainty
The overview is a 2D projection of the 6D network, with omitted dimensions and wraparound links. Nodes connect through integrated network interfaces.
Documented fields: 18 / 22
Every field this catalog checks. Fields listed above have a value; the remainder are shown here. A field marked absent is not publicly available in the sources we could verify, which does not mean the hardware is missing. Fields that cannot apply, such as benchmark results for an unranked system, are excluded from the total.
Cluster — 7 / 7
- ✓Manufacturer
- ✓Ranked core count
- ✓Theoretical peak (PFlop/s)
- ✓Measured power (kW)
- ✓Energy efficiency (GFlop/s per watt)
- ✓Operating system
- ✓Network topology
A64FX compute — 11 / 15
- ✕System / node model
- ✓Compute nodes of this type
- ✓CPU model
- ✓CPU sockets per node
- ✓Cores per socket
- ✓Physical CPU cores
- ✓CPU note
- –Accelerator model
- –Physical accelerators per node
- ✓Host / unified memory
- –Accelerator memory
- –Accelerator ↔ accelerator link
- –On-node fabric
- –CPU ↔ accelerator link
- ✓Network interfaces per node
- ✓Interface model
- ✓Interface speed
- ✕Interface-to-accelerator topology
- –Aggregate scale-out per node
- –GPUDirect RDMA
- ✕Management / storage networking
- ✕Local storage
- –NVMe data disks
- –Max remote data disks
- –Uncached disk IOPS
- –Uncached disk throughput (MBps)
- ✓Form factor
Sources
Browse all cluster architecture pages →
CLUSTER