Supercomputers/Systems/Nano 4 GB200
Taiwan · operational · mixed
Nano 4 GB200
Operated by National Center for High-Performance Computing .
Researched with AI assistance, checked by scripts. These rows were researched by AI agents reading public operator, laboratory and vendor sources, appended by scripts and checked by automated validation; no subject-matter expert has reviewed them (how this site is made). A mechanical second-reader pass has since fetched every cited source: 1350 of 1540 systems have a readable citation that names them, and 64 are genuinely weakly sourced. Every claim carries its source and a confidence tier. Treat anything below verified as a lead, not a citation.
Measured performance
- Rmax
- 5.16 PFlop/s
- Rpeak
- 5.62 PFlop/s
- Rmax ÷ Rpeak
- 91.8%
- Cores
- 26,496
- Power
- 164.16 kW
- Per watt
- 31.4 GF/W
Figures are as last publicly reported for the configuration described below, not a live measurement. Where a system was upgraded in place, the post-upgrade configuration is shown and the earlier one appears in the timeline.
What this machine is made of
One row per supplier relationship. “Supplier at build” is the company that shipped the part at the time; where that company has since been acquired, the parent it rolls up to today is shown beside it. That distinction is what makes ticker-level aggregation possible across a thirty-year dataset.
| Role | Part | Supplier at build | Quantity | Confidence | Source |
|---|---|---|---|---|---|
| Integrator | - ASUS ESC NM2N721-E1, with Taiwan AI Cloud and Taiwan Mobile | ASUS | - | Reported | top500.org |
| Accelerator | NVIDIA GB200 NVL72 Accelerators Arm CPU + CUDA GPU · TSMC 4NP NVIDIA B200 180 GB in GB200 NVL72 racks (NCHC: two GB200 NVL72) | NVIDIA | - | Reported | top500.orgnchc.org.tw |
| CPU | NVIDIA Grace CPUs Armv9 (Neoverse V2) · 72C · 3.1 GHz · TSMC 4N NVIDIA Grace 72C 3.1GHz | NVIDIA | - | Reported | top500.org |
| Interconnect | NVIDIA NVLink 5 / NVSwitch Interconnect All-to-all within an NVLink domain · 1.8 TB/s per GPU NVLink 5 and NDR 400 InfiniBand | NVIDIA | - | Reported | top500.org |
| Interconnect | NVIDIA InfiniBand NDR (Quantum-2) Interconnect Fat tree or dragonfly+ · 400 Gb/s per port · Under 0.6 microseconds NVIDIA NDR 400 InfiniBand | NVIDIA | - | Reported | top500.org |
| Operating system | Red Hat Enterprise Linux Operating systems Linux distribution · Open source, commercial support subscription RHEL | Red Hat | - | Reported | top500.org |
The read
The Grace-Blackwell part of NCHC's Nano 4 (Chinese name 晶創26) installation, listed separately on the June 2026 TOP500 as Nano4-b at rank 298: ASUS ESC NM2N721-E1 nodes (ASUS, Taiwan AI Cloud and Taiwan Mobile named as builders) with NVIDIA Grace 72-core CPUs, B200 180 GB GPUs, NVLink 5 plus NDR 400 InfiniBand and RHEL, 26,496 cores, 5.16 PFlop/s Rmax, 5.62 PFlop/s Rpeak and 164.16 kW. NCHC's website says Nano 4 combines 220 H200 compute nodes with two GB200 NVL72 racks, the first in Taiwan, and opened a free GB200 NVL72 trial on 1 September 2026; the 26,496 cores in the TOP500 entry equal about 368 Grace CPUs at 72 cores each, more than the 72 that two NVL72 racks (36 Grace each) would hold, so the exact hardware boundary of the listed run is not confirmed. The H200 portion is the separate row nano-4 (81.55 PFlop/s). Ranked 0 here because the June 2026 list edition is already full in the dataset.
Timeline
- 2025 First operational source
Change history
- 2026-10-18 Added Nano 4 GB200 (National Center for High-Performance Computing)
Further reading
Operating a supercomputer
Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.
Inside a node
Latency-optimised versus throughput-optimised processors, NUMA, and the end of the host-to-device copy.
The interconnect
Topologies, why latency and tail behaviour matter more than bandwidth, and how the fabric market consolidated.