Supercomputers/Systems/Theia-cmh
USA · operational · ai training
Theia-cmh
Also known as NVIDIA Theia-cmh GB300 NVL72 cluster
Operated by NVIDIA at NVIDIA (undisclosed US site) .
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
- -
- 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 GB300 NVL72 Accelerators 512 nodes x 4 NVIDIA Blackwell Ultra GB300 GPUs (32x NVIDIA GB300 NVL72 per system name), 2,048 GPUs submitted to MLPerf Training v6.0 | NVIDIA | 2,048 reported | Reported | github.com |
| CPU | NVIDIA Grace CPUs Armv9 (Neoverse V2) · 72C · 3.1 GHz · TSMC 4N Two NVIDIA Grace (Neoverse-V2, 72 cores) host processors per node | NVIDIA | - | Reported | github.com |
| Interconnect | - 4x ConnectX-8 Ethernet NDR (2x400 Gb/s) per node | NVIDIA | - | Reported | github.com |
AI datacenter metrics
Figures beyond an accelerator count, each with who stated it. What can and cannot be compared across AI datacenters.
| Metric | Value | As of | Basis | Note | Source |
|---|---|---|---|---|---|
| Accelerators | 2,048 accelerators | 2026-06 | Operator stated | MLCommons v6.0 NVIDIA submission file: 512 nodes x 4 GPUs; rack count in the name implies 2,304 installed | github.com |
The read
NVIDIA's GB300 NVL72 cluster in the MLPerf Training v6.0 files (repository created 15 June 2026), named 'Theia-cmh (32x NVIDIA GB300 NVL72)' with 512 nodes of four Blackwell Ultra GPUs, so 2,048 GPUs in the largest run. Its fabric is listed as four ConnectX-8 Ethernet NDR (2x400 Gb/s) adapters per node, where the smaller 'Theia (8x NVIDIA GB300 NVL72)' entry in v5.1 lists ConnectX-8 InfiniBand XDR 800 Gb/s, so the two are described with different fabrics. By the rack count in the name the installed total is 32 x 72 = 2,304 GPUs against 2,048 submitted. The entry status is 'Available cloud' and no page gives a location or explains the 'cmh' suffix, so the site is NVIDIA's undisclosed one. NVIDIA's MLPerf Training 6.0 blog does not mention Theia, so this rests on the MLCommons file alone. Sibling of nvidia-tyche-hsg, which is the GB200 equivalent. It has no HPL result.
Timeline
No dated events recorded.
Site & facility
Change history
- 2026-10-17 Added NVIDIA Theia-cmh GB300 NVL72 cluster (MLPerf Training v6.0)
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 |
|---|---|---|
| github.com | names system, part and count | NVIDIA GB300 NVL72 · counts: 2048 |
Checked 2026-10-18 by pnpm verify. Re-run it and the table changes with the web.