Parts/CPUs/AMD 4th Gen EPYC 24C (in MI300A)
Why this part
Not a socket at all: three eight-core Zen 4 chiplets sit inside the MI300A package alongside six CDNA 3 accelerator chiplets, and both address the same 128GB of HBM3. El Capitan is built on it. The distinction between host memory and device memory, which has shaped every heterogeneous machine since Roadrunner, simply does not exist here, and neither does the copy between them. It also collapses the supply chain: one part number covers what used to be a CPU vendor, a GPU vendor and the link between them.
Systems using this part
- Systems
- 10
- Rmax underneath
- 2.75 EFlop/s
- Units recorded
- 168
1 of 10 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| HPC7 2026-06 | CPU | Italy | - | 571.5 PFlop/s | AMD | #6 → |
| Sirius (Tsukuba) 2026 | CPU | Japan | - | 5.63 PFlop/s | AMD | - |
| Tuolumne 2025-01 | CPU | USA | - | 208.1 PFlop/s | AMD | #14 → |
| Hunter 2025-01 | CPU | Germany | - | 31.68 PFlop/s | AMD | #54 → |
| Cosmos (SDSC) 2025 | CPU | USA | 168 | 5.18 PFlop/s | AMD | #272 → |
| El Capitan 2024-11 | CPU | USA | - | 1.81 EFlop/s | AMD | #2 → |
| El Dorado 2024-11 | CPU | USA | - | 68.02 PFlop/s | AMD | #41 → |
| rzAdams 2024 | CPU | USA | - | 24.38 PFlop/s | AMD | #87 → |
| AMD Ouranos 2024 | CPU | France | - | 2.99 PFlop/s | AMD | #9 → |
| Portage 2024 | CPU | USA | - | 24.50 PFlop/s | AMD | #10 → |
“Units recorded” sums only the edges where a public source states a quantity. It is a floor, not a total, and should never be read as installed base.