CEA-HE node architecture
CEA-HE is operated by Commissariat a l'Energie Atomique (CEA) in France. This profile covers Ranked system / partition. Specifications are listed below; anything the operator has not published is itemised in the documented-fields list.
Published benchmark (TOP500 · June 2026)
- HPL rank
- #29
- HPL Rmax
- 90.79 PFlop/s
- Theoretical peak Rpeak
- 171.26 PFlop/s
- Ranked core count
- 548,352
- Measured power
- 1770 kW
- Energy efficiency
- 51.29 GFlop/s per watt
Cluster · BullSequana XH3000 · public profile
- Operator / site
- Commissariat a l'Energie Atomique (CEA)
- Country
- France
- Profile scope
- Ranked system / partition
- Data checked
- 14 September 2026
- Manufacturer
- Bull
- Ranked core count
- 548352
- Theoretical peak (PFlop/s)
- 171.26
- Measured power (kW)
- 1770
- Energy efficiency (GFlop/s per watt)
- 51.293785310734464
- Operating system
- Linux
BullSequana XH3000 · public profile
Items marked ▧ come from the documented NVIDIA GH200 Grace Hopper Superchip platform specification.
- System / node model
- BullSequana XH3000
- Compute nodes of this type
- 672
- CPU model
- GH Superchip 72C 3GHz
- Cores per socket
- 72
- Accelerator model
- NVIDIA GH200 Superchip
- CPU ↔ accelerator link
- NVLink-C2C · 900 GB/s coherent CPU ↔ GPU
- Interface model
- Atos BXI v2 (quad-rail)
- Form factor
- 28 BullSequana XH3000 racks · direct liquid cooled
Scope and uncertainty
Node count and rack count follow CEA's October 2025 site update, which states 672 Grace Hopper nodes in 28 BullSequana XH3000 racks. The April 2024 Eviden/CEA launch announcement gave 477 nodes and 104 PFlop/s peak against the later 672 nodes and 180 PFlop/s; CEA does not reconcile the two, so the later CEA-authored figure is used. Superchips per node are not published. CEA publishes a fat-tree topology only for the separate EXA1-HF partition, which is not carried across. Node specifications follow the documented NVIDIA GH200 Grace Hopper Superchip platform (Superchip module — nodes per system and per-node superchip counts are operator-specific); operator-specific deviations are not published.
Documented fields: 14 / 27
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 — 6 / 7
- ✓Manufacturer
- ✓Ranked core count
- ✓Theoretical peak (PFlop/s)
- ✓Measured power (kW)
- ✓Energy efficiency (GFlop/s per watt)
- ✓Operating system
- ✕Network topology
BullSequana XH3000 · public profile — 8 / 20
- ✓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 →
CLUSTERATLAS