Supercomputers/Systems/Hawk
Germany · decommissioned · hpc
Hawk
Operated by HLRS at HLRS, University of Stuttgart .
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: 1263 of 1345 systems have a readable citation that names them, and 56 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
- 19.33 PFlop/s
- Rpeak
- 26.00 PFlop/s
- Rmax ÷ Rpeak
- 74.4%
- Cores
- 720,896
- Power
- 3.91 MW
- Per watt
- 5.0 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 | - Apollo 9000 | Hewlett Packard Enterprise | - | Reported | hlrs.de |
| CPU | AMD EPYC 7742 (Rome) CPUs x86-64 · 64C · 2.25 GHz · 7 nm · 225 W AMD EPYC 7742 64C 2.25GHz | AMD | - | Reported | hlrs.de |
| Interconnect | NVIDIA InfiniBand HDR Interconnect Fat tree or dragonfly+ · 200 Gb/s per port · About 0.6 microseconds Infiniband HDR200 | NVIDIA | - | Verified | hlrs.detop500.org |
| Storage | DDN EXAScaler Storage Lustre appliance Storage: DDN EXAScaler | DDN | - | Reported | hlrs.de |
| Cooling | Direct liquid cooling, warm water Cooling Direct-to-chip cold plates, facility water loop · Removes 90+ percent of node heat to liquid HPE Apollo 9000: direct liquid cooled | Hewlett Packard Enterprise | - | Reported | hlrs.de |
The read
CPU-only at a moment when almost every comparable procurement went heterogeneous - 720,896 EPYC cores and no accelerator at all. HLRS's engineering user base runs codes that do not vectorise onto GPUs, which is a reminder that the accelerator share chart measures supply, not suitability. TOP500 independently confirms HLRS, HPE, EPYC 7742 and Rmax 19.33, but disagrees with the two figures recorded here: it lists Rpeak 25.16 against the 25.999 recorded, and 698,880 cores against 720,896. Both are left as recorded and unresolved, which is why the tier stays reported.
Timeline
- 2019 Announced source
- 2020-02 First operational source
- 2020-06 List debut at #16 (June 2020) source
- 2025-04-27 Decommissioned 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
HLRS, University of Stuttgart
- Location
- Stuttgart, Germany
- Cooling
- Direct liquid cooling
- Commissioned
- 1996
Where sources disagree
We record conflicts instead of picking a winner quietly. All open questions.
- Rpeak: 25.999 PFlop/s as recorded against 25.16 PFlop/s. TOP500 and HLRS disagree; the recorded figure is left unchanged until one is confirmed.
- Core count: 720,896 as recorded against 698,880. The recorded figure equals 5,632 nodes times 128 cores; no fetched page states the node count.
Change history
- 2026-10-02 Marked Hawk decommissioned
- 2026-09-19 Second source added, but TOP500 disagrees on Rpeak and cores
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 | AMD EPYC 7742 (Rome), NVIDIA InfiniBand HDR |
| hlrs.de | names system and part | AMD EPYC 7742 (Rome), NVIDIA InfiniBand HDR |
Checked 2026-10-13 by pnpm verify. Re-run it and the table changes with the web.
Further reading
Power and cooling
Why air ran out at 20 kW a rack, how warm-water cooling removes the chillers, and what PUE conceals.
Inside a node
Latency-optimised versus throughput-optimised processors, NUMA, and the end of the host-to-device copy.
Operating a supercomputer
Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.