NIPA (Kakao) node architecture
NIPA (Kakao) is operated by Kakao in South Korea. 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
- #39
- HPL Rmax
- 69.1 PFlop/s
- Theoretical peak Rpeak
- 78.65 PFlop/s
- Ranked core count
- 36,720
Cluster · Supermicro SYS-A22GA-NBRT · public profile
- Operator / site
- Kakao
- Country
- South Korea
- Profile scope
- Ranked system / partition
- Data checked
- 14 September 2026
- Manufacturer
- Niagara Computers / Supermicro
- Ranked core count
- 36720
- Theoretical peak (PFlop/s)
- 78.65
- Operating system
- Ubuntu 24.04.3 LTS
- Network topology
- 8-rail optimised fat-tree
Supermicro SYS-A22GA-NBRT · public profile
Items marked ▧ come from the documented Supermicro SYS-A22GA-NBRT platform specification.
- System / node model
- Supermicro SYS-A22GA-NBRT
- Compute nodes of this type
- 255
- CPU model
- Intel Xeon 6960P 72C 2.7GHz
- CPU sockets per node
- 2
- Cores per socket
- 72
- Physical CPU cores
- 144
- Accelerator model
- NVIDIA B200
- Physical accelerators per node
- 8
- Accelerator memory
- 180 GB HBM3e per GPU
- Accelerator ↔ accelerator link
- NVLink 5 · all GPU pairs
- On-node fabric
- NVLink 5 through NVLink Switch — 1.8 TB/s GPU-to-GPU, 14.4 TB/s aggregate across the HGX B200 baseboard
- Interface model
- InfiniBand NDR
- Interface-to-accelerator topology
- PCIe 5.0 x16 slots 2–9 are the GPU-linked east-west backend network adapter slots
- Form factor
- 10U rackmount · 6 × 5250 W Titanium (3+3 redundant)
Scope and uncertainty
Korea's national AI computing portal lists three 510- and 255-server B200 clusters without naming their providers, one of which is this system; every one is listed with 2,304 GB per server and a 72-core CPU complement. The 72-core figure conflicts with this system's own TOP500 submission, whose core count implies 144 cores per node across two 72-core sockets, so neither the memory nor the core figure is carried over from the portal. Node count and fabric follow Kakao Cloud's own build report for this cluster. Kakao does not name a CPU, host memory, NVLink generation or adapter model; accelerators per node follow the documented Supermicro SYS-A22GA-NBRT chassis. Node specifications follow the documented Supermicro SYS-A22GA-NBRT platform (10U rackmount, two sockets, NVIDIA HGX B200 8-GPU); operator-specific deviations are not published.
Documented fields: 19 / 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 — 5 / 7
- ✓Manufacturer
- ✓Ranked core count
- ✓Theoretical peak (PFlop/s)
- ✕Measured power (kW)
- ✕Energy efficiency (GFlop/s per watt)
- ✓Operating system
- ✓Network topology
Supermicro SYS-A22GA-NBRT · public profile — 14 / 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 →
CLUSTERATLAS