Japan · operational · mixed

mdx I

Also known as mdx

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 40GB Accelerators CUDA (Ampere GA100) · 6,912C · 1.41 GHz · 7 nm · 400 W 40 GPU nodes of 8 NVIDIA A100 40GB SXM each	NVIDIA NVDANASDAQ	320 reported	Reported	mdx.jp arxiv.org

The read

The first-generation mdx data-science cloud platform, hosted by the Information Technology Center of the University of Tokyo and jointly operated by nine national universities and two national research institutes. Its system page lists 40 GPU nodes of eight NVIDIA A100 40 GiB GPUs each (320 GPUs, recorded here as 40 nodes times 8) and 368 CPU nodes, with the platform in operation since autumn 2021 per the platform's own paper (arXiv 2203.14188). It matters for AI because the LLM-jp project trained its Japanese open models on it: the LLM-jp paper describes a 16-node, 128-GPU virtual cluster on mdx, cut to 12 nodes for model v1.0 because of network problems and run at the full 16 nodes for v2.0. mdx is a virtualised multi-tenant cloud, so a training run sees a slice of the 320 GPUs, not the whole system. The mdx I portal carried a notice dated 26 August 2026 about an emergency shutdown for a cooling failure that was later resolved, which is why status is dated August 2026. The site is left unlocated because the pages we fetched name the University of Tokyo but no campus. The second-generation mdx II at the University of Osaka (15 nodes of four H200 GPUs) is not recorded here. The CPU model is not recorded because the platform's pages disagree on whether it is the 8368 or 8369.

Timeline

2021 First operational source

Change history

2026-10-02 mdx I added (University of Tokyo data-science cloud used to train LLM-jp, 320 A100)

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 40GB · counts: 320
mdx.jp	does not name this system	NVIDIA A100 SXM4 40GB

Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.