Supercomputers/Systems/CARA
Germany · operational · hpc
CARA
Also known as Computer for Advanced Research in Aerospace
Operated by DLR at TU Dresden ZIH .
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
- 1.75 PFlop/s
- Rpeak
- 2.57 PFlop/s
- Rmax ÷ Rpeak
- 68.1%
- Cores
- 145,920
- Power
- 1.20 MW
- Per watt
- 1.5 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 | - NEC LX cluster built by NEC for DLR | NEC | - | Reported | nec.comtop500.org |
| Accelerator | - NVIDIA A100 GPUs in 10 nodes of four GPUs each | NVIDIA | 40 reported | Reported | dlr.de |
| CPU | - AMD EPYC 7601 32C 2.2GHz | AMD | - | Reported | nec.comtop500.org |
| Interconnect | NVIDIA InfiniBand HDR Interconnect Fat tree or dragonfly+ · 200 Gb/s per port · About 0.6 microseconds Mellanox HDR InfiniBand fabric in a Dragonfly+ setup | NVIDIA | - | Reported | nec.comtop500.org |
| Cooling | Direct liquid cooling, warm water Cooling Direct-to-chip cold plates, facility water loop · Removes 90+ percent of node heat to liquid Direct liquid cooling at 35 C warm water inlet temperature | NEC | - | Reported | nec.comdlr.de |
The read
DLR's German Aerospace Center cluster, the Dresden counterpart to CARO in Gottingen, hosted and administered by the ZIH in the Lehmann Data Centre at TU Dresden. An NEC LX cluster of AMD EPYC 7601 32C 2.2GHz nodes on HDR InfiniBand with direct liquid cooling. NEC's June 2019 award announcement describes 2,300 Gigabyte compute nodes and up to 147,200 cores in a budget above 20 million euros, a Dragonfly+ HDR fabric, 116 large-memory nodes and 20 visualisation nodes; DLR's page describes 2,280 nodes plus 10 nodes with four NVIDIA A100 GPUs, up to 1.2 MW and a PUE near 1.2, which is the power figure recorded. TOP500 lists 145,920 cores, Rmax 1.75 PFlop/s and Rpeak 2.57 PFlop/s, ranking 221st on the November 2019 list and last appearing at rank 494 in November 2022. Node and core counts differ between sources; TOP500's core count is recorded.
Timeline
- 2019 First operational source
- 2019-06 Announced source
- 2019-11 List debut at #221 (November 2019) source
Public list appearances
Pointers only: rank and edition, linking to the canonical entry. We do not reproduce list tables. See sourcing policy.
Site & facility
Change history
- 2026-10-07 Added CARA, DLR's EPYC cluster at TU Dresden
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 |
|---|---|---|
| top500.org | names system and part | NVIDIA InfiniBand HDR |
| dlr.de | names system only | - |
Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.
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.