Supercomputers/Systems/Kalos (Shanghai AI Lab)
China · operational · ai training
Kalos (Shanghai AI Lab)
Also known as Kalos
Operated by Shanghai AI Laboratory .
Phase 1 seed dataset, compiled by hand. These rows were built from public operator, laboratory and vendor sources. A mechanical second-reader pass has since fetched every cited source: 1092 of 1153 systems have a readable citation that names them, and 40 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
- -
- Rpeak
- -
- Rmax ÷ Rpeak
- -
- Cores
- -
- Power
- -
- Per watt
- - 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 |
|---|---|---|---|---|---|
| Accelerator | NVIDIA A100 SXM4 80GB Accelerators 302 nodes of 8 NVIDIA A100-SXM 80GB GPUs | NVIDIA | 2,416 reported | Reported | arxiv.org |
| Interconnect | NVIDIA InfiniBand HDR Interconnect Fat tree or dragonfly+ · 200 Gb/s per port · About 0.6 microseconds NVIDIA Mellanox 200Gbps HDR InfiniBand, four adapters per node for application traffic plus one for storage | NVIDIA | - | Reported | arxiv.org |
The read
The newer of the two LLM-development clusters in Shanghai AI Laboratory's private Acme GPU datacenter, described in the March 2024 paper 'Characterization of Large Language Model Development in the Datacenter'. Kalos has 2,416 NVIDIA A100 GPUs in 302 nodes (eight A100-SXM 80GB per node, two Xeon Platinum 8358P CPUs, 2 TB of host memory), joined by NVLink and NVSwitch inside a node and 200 Gb/s HDR InfiniBand between nodes, with four InfiniBand adapters per node for application traffic plus a dedicated storage adapter, an all-NVMe shared parallel file system and Kubernetes scheduling. The paper's workload trace runs from March to August 2023, so status is dated August 2023 and no later confirmation exists. Together with Seren it makes up Acme's 4,704 A100s. The paper gives no location, power or start date, so the site is left unlocated. The paper is the only source and is the operator's own research output, so the tier is reported.
Timeline
- 2024-03 Announced source
Change history
- 2026-10-02 Kalos (Shanghai AI Lab) added (2,416 A100, LLM development)
Source check
We fetched this system's own citations and recorded whether each page actually mentions it. This is a corroboration signal, not a fact check, and it is published so you can see how well the sourcing holds up rather than take it on trust.
| Cited source | Result | Found on the page |
|---|---|---|
| arxiv.org | names system, part and count | NVIDIA A100 SXM4 80GB, NVIDIA InfiniBand HDR · counts: 2416 |
Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.