COMPUTE ATLAS Supercomputer supply-chain graph
1581 systems 558 sites Sourced data

Supercomputers/Systems/Dalia

France · operational · mixed

Dalia

RSS for this system Your watchlist

Also known as IDRIS Dalia GB200 NVL72

Operated by IDRIS Computing Centre at IDRIS .

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 1581 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
Integrator -
Supplied by NVIDIA through its French partner Bull
 Bull
ATOEPA rolls up to Atos
- Reported idris.fr
Accelerator NVIDIA GB200 NVL72
Accelerators
Arm CPU + CUDA GPU · TSMC 4NP
NVIDIA GB200 NVL72, 18 nodes with four GB200 GPUs each (186 GB HBM3e per node)
 NVIDIA
NVDANASDAQ
- Reported idris.fr
CPU NVIDIA Grace
CPUs
Armv9 (Neoverse V2) · 72C · 3.1 GHz · TSMC 4N
Two 72-core Arm Neoverse-V2 Grace CPUs per node across 18 nodes
 NVIDIA
NVDANASDAQ
36
reported
 Reported idris.fr
Interconnect -
InfiniBand single-level non-blocking fat tree
 NVIDIA
NVDANASDAQ
- Reported idris.fr
Storage -
DDN AI400X2 Turbo array, 500 TB total
 DDN
500
reported
 Reported idris.fr

The read

Dalia is the NVIDIA GB200 NVL72 prototype installed at IDRIS (Orsay) in summer 2025, which IDRIS calls the first GB200 NVL72 supercomputer put into service in Europe. NVIDIA supplied it through its French partner Bull, and it is meant for scientific AI and HPC benchmarking and 'Grand Challenge' projects. The IDRIS hardware page lists 18 compute nodes, each with two 72-core Grace CPUs (144 cores per node), 1.7 TB of memory and four GB200 GPUs with 186 GB HBM3e per node, so 18 x 4 = 72 GPUs, a full NVL72. The same page's headline says 64 GPUs, which conflicts with its own node table; the 72 total is derived and the 64 is probably a typo, but IDRIS does not say so, so no accelerator count is recorded. Networking is a single-level non-blocking InfiniBand fat tree; storage is 500 TB on a DDN AI400X2 Turbo array over converged Ethernet (eight 200 GbE links, two per compute node). Compute nodes can reach the internet, unlike Jean Zay, and login nodes are Intel Xeon Gold 6530 systems. It sits beside the larger Jean Zay machines at the same centre (jean-zay-h100). No power or FLOPS figure is published on the pages fetched. Both pages belong to IDRIS, so the tier is reported.

Timeline

No dated events recorded.

Site & facility

IDRIS

Location
Orsay, France
Commissioned
2024

Change history

Same site

Same operator

Same accelerator

Further reading

Inside a node

Why accelerators exist, what the memory hierarchy costs, and how the CPU and accelerator merged onto one package.

Operating a supercomputer

Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.

The interconnect

Topologies, why latency and tail behaviour matter more than bandwidth, and how the fabric market consolidated.