Liquid Cooling System
As accelerator racks move from 15 kW to 60 kW and beyond, air can no longer carry the heat away.
HDmann designs and manufactures the wetted loop that sits between the facility water and the GPU cold plates:
coolant distribution units, rack manifolds, EPDM hose assemblies with UQD quick couplings, and the stainless
valves and fittings that complete the secondary circuit.
Every component is built for a 15 to 25 year design life: 316L wetted parts, orbital welds, pickled and passivated
internal surfaces, factory pressure and helium leak testing, and full material traceability with EN 10204 3.1 certificates
on request. Loops are delivered as pressure-tested modules, so site work is limited to connection and commissioning.
Coolant Distribution Unit
HD-CDU series in-row and in-rack coolant distribution unit, 60 to 600 kW, redundant pumps, plate heat exchanger, Modbus/SNMP to DCIM.
Liquid Cooling Manifold
HD-MF 316L supply/return manifolds with 4 to 12 branches, UQD and UQDB blind-mate couplings, integrated balancing valves.
EPDM Hose & Quick Coupling
EHW194 grade peroxide cured EPDM hose assemblies with dripless UQD quick couplings, 100% pressure tested at 25 bar.
Valves & Fittings
Stainless butterfly and ball valves, strainers, elbows, tees and adaptors for the secondary cooling loop, PN16.Typical Loop Arrangement
| 1 | Facility water loop (cooling tower / chiller), 32 to 45 °C |
| 2 | Coolant distribution unit (CDU) - plate heat exchanger, pumps, filtration |
| 3 | Rack manifold (supply / return) with balancing valves |
| 4 | EPDM hose assembly with UQD quick coupling to each cold plate |
| 5 | Cold plate on CPU / GPU, return to CDU |
Why HDmann
- Wetted materials 316L / EPDM, compatible with PG25 and deionized water
- Rack-level and row-level scope from one supplier, one interface
- Pressure tested modules with FAT report and material certificates
- UQD / UQDB blind-mate interface, OCP ORV3 rack compatible
- Commissioning support and spare parts kits for the first fill
Applications
- AI training clusters (GPU)
- HPC and supercomputing
- Hyperscale and colocation data centers
- Edge and containerized compute
- Immersion and hybrid cooling
- Battery energy storage cooling
- Power electronics and converters
- Industrial process cooling