CLUSTERATLAS
A field guide to supercomputing

MareNostrum 5 ACC

EuroHPC/BSC

CPU Accelerator Interconnect
i
Reading this map

MareNostrum 5 ACC node architecture

MareNostrum 5 ACC is operated by EuroHPC/BSC in Spain. 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
#16
HPL Rmax
175.3 PFlop/s
Theoretical peak Rpeak
249.44 PFlop/s
Ranked core count
663,040
Measured power
4159 kW
Energy efficiency
42.15 GFlop/s per watt

Cluster · ACC compute

Operator / site
EuroHPC/BSC
Country
Spain
Profile scope
Ranked system / partition
Data checked
14 September 2026
Manufacturer
Bull
Ranked core count
663040
Theoretical peak (PFlop/s)
249.44
Measured power (kW)
4159
Energy efficiency (GFlop/s per watt)
42.14955518153402
Operating system
Red Hat Enterprise Linux 9.1
Network topology
Three-layer fat-tree · 324 QM9790 switches · 3 GPP islands, 1 storage island and 7 ACC islands

ACC compute

System / node model
BullSequana XH3000
Compute nodes of this type
1120
CPU model
Intel Xeon Platinum 8460Y+ · 40 cores
CPU sockets per node
2
Cores per socket
40
Physical CPU cores
80
Accelerator model
NVIDIA H100
Physical accelerators per node
4
Host / unified memory
512 GB DDR5
Accelerator memory
64 GB HBM2 per GPU
CPU ↔ accelerator link
PCIe 5.0
Network interfaces per node
4
Interface model
NVIDIA ConnectX-7 (InfiniBand NDR200)
Interface speed
200 Gb/s per adapter · 800 Gb/s per node
Local storage
480 GB NVMe (/scratch)

Scope and uncertainty

Accelerated partition (ACC) only; the 6,408-node general-purpose partition, the GPP-HBM nodes and the AI upgrade partitions are outside this profile. BSC documents a three-level fat-tree with 324 QM9790 switches across seven ACC islands. The machine-wide storage totals 248 PB of SSD/flash plus disk and a 402 PB tape archive.

Documented fields: 22 / 29

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

ACC compute — 15 / 22

  • 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 →

Data & sources