Supercomputers/Systems/Medusa (Dell)
USA · operational · mixed
Medusa (Dell)
Also known as Dell Medusa H100 cluster
Operated by Dell Technologies .
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
- 3.08 PFlop/s
- Rpeak
- 4.33 PFlop/s
- Rmax ÷ Rpeak
- 71.1%
- Cores
- 9,216
- 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.
How it compares
Ranked against the other machines in this dataset that report the same figure: those that first came into service within two years of Medusa (Dell), and those in USA. Computed from our tables; a metric is ranked only when at least eight peers report it.
| Metric | Value | Against its era | Against its country |
|---|---|---|---|
| Rmax | 3.08 PFlop/s | #207 of 234 · bottom 13% | #115 of 332 · top 34% |
| Rmax ÷ Rpeak | 71.1 % | #135 of 231 · bottom 43% | #194 of 331 · bottom 43% |
Every metric is ranked largest-first, so a high power-draw rank means a big power draw, not a good one. Rmax ÷ Rpeak uses HPL-measured results only. Rankings move as machines are added; they describe this dataset, not the world.
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 | - Dell PowerEdge XE9680 | Dell Technologies | - | Reported | top500.org |
| Accelerator | NVIDIA H100 SXM5 Accelerators CUDA (Hopper GH100) · 16,896C · 1.98 GHz · TSMC 4N · 700 W NVIDIA H100 | NVIDIA | - | Reported | top500.org |
| CPU | Intel Xeon Platinum 8468 CPUs x86-64 · 48C · 2.1 GHz · Intel 7 · 350 W Intel Xeon Platinum 8468 48C 2.1GHz | Intel | - | Reported | top500.org |
| Interconnect | NVIDIA InfiniBand NDR (Quantum-2) Interconnect Fat tree or dragonfly+ · 400 Gb/s per port · Under 0.6 microseconds Infiniband NDR400 | NVIDIA | - | Reported | top500.org |
| Operating system | Ubuntu Server Operating systems Linux distribution · Open source, free with paid support tiers Ubuntu 22.04.2 LTS | Canonical | - | Reported | top500.org |
The read
NVIDIA GPU system in Dell Technologies' HPC & AI Innovation Lab, listed by TOP500 as Medusa: PowerEdge XE9680 with Xeon Platinum 8468 48-core 2.1 GHz, NVIDIA H100 and InfiniBand NDR400 under Ubuntu 22.04.2 LTS (nvcc and gcc, cuBLAS and MKL, HPC-X MPI), installed 2024. It has 9,216 cores, Rmax 3.08 and Rpeak 4.33 PFlop/s (71 percent efficiency) and 29.2 TFlop/s on HPCG, and no power figure; it debuted at 346th in November 2024 and was 440th in June 2026. Medusa is the InfiniBand H100 counterpart of the lab's MI300X Ethernet machine IronMan (dell-ironman) and follows the earlier A100 Rattler (rattler-dell). TOP500 is the only source and no Dell page naming Medusa was found, so no site or in-service date is recorded and the tier is reported.
Timeline
Public list appearances
Pointers only: rank and edition, linking to the canonical entry. We do not reproduce list tables. See sourcing policy.
Change history
- 2026-10-18 Added Medusa (Dell)
Nearby in scale
The machines closest to Medusa (Dell) by Rmax, smaller to larger.
- NIMS Materials Simulator · Japan, operational · 3.08 PFlop/s
- Hera · USA, operational · 3.08 PFlop/s
- Camphor 2 · Japan, operational · 3.06 PFlop/s
- Komondor · Hungary, operational · 3.1 PFlop/s
- ASEAN-Korea HPC Facility · Indonesia, operational · 3.1 PFlop/s
- Lichtenberg II Phase 2 · Germany, operational · 3.1 PFlop/s
Related systems
Same operator
- IronMan (Dell) · USA, operational, 2024
- Zenith (Dell EMC) · USA, operational, 2016
- Rattler · USA, operational
Same accelerator
- ARF-ACC · Turkey, operational, 2024
- Arka AI/ML · India, operational, 2024
- ASPIRE 2A+ · Singapore, operational, 2024
- BioHive-2 · USA, operational, 2024
- CHIE-2 · Japan, operational, 2024
- CHIE-3 · Japan, operational, 2024
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.