In Li-ion UPS and BESS areas, set up detection with combined smoke-heat detectors; do not rely on heat detectors alone.
In DNV-GL tests, F500, FireIce, PyroCool foam, Stat-X aerosol and sprinklers all extinguished the fires, but the water-based systems cooled better over time.
For high-heat-flux chips, weigh the pump-power increase of complex microchannel cold-plate geometry against the thermal-resistance gain (for example 11.8%).
Chip heat flux is described as reaching ultra-high levels of 10^6-10^7 W/m².
Set WUE targets from the heat-rejection topology (dry, free-cooling, hybrid, evaporative) and not only the server-side cooling technology; WUE depends on the share of heat rejected by evaporation.
Dry facilities with closed-loop liquid cooling need essentially no water, while evaporative systems typically report WUE up to 2.5 L/kWh.
In efficiency reporting do not make PUE the only metric; knowing its weaknesses (manipulation, optimal-condition reporting, ignoring IT utilisation), complement it with WUE, CUE and IT-side metrics (SPUE, PpW, ITEE).
Google reported a 2023 fleet-average annual PUE of 1.10 against an industry average of 1.58; Microsoft reports PUE and WUE together.
In the power-chain efficiency calculator multiply UPS, transformer/PDU, distribution and PSU efficiencies at the actual load fraction rather than nameplate, and offer architecture presets for the five options with Table 1 values as defaults.
The paper calls the North American 480 V AC system with PDU transformers inefficient and obsolete for new designs.
For racks where GPU/accelerator thermal design power exceeds 1,200 W, take liquid cooling as a mandatory design input, and weigh the initial cost (about 1.9-2.2 times air cooling) against TCO (up to 20% lower after 4 years, vendor figure).
Cooling systems in air-cooled data centres consume around 40% of total data-centre energy.
Add UPS placement (dedicated_ups_room, white_space) to the TCO module and charge room construction ($/ft2 or $/m2, editable by building type) plus occupancy cost for a dedicated room.
Base rents of $19-50/ft2 plus taxes, utilities and common areas.
In data-centre carbon accounting, state the method (Tier 1, 2, 3) and factor source (IPCC or national) when choosing a grid emission factor; simple methods can underestimate generation emissions by about 73%.
AEF for the North zone by Method 5 rose from 0.1627 tCO2/MWh in 2016 to 0.4709 in 2022, then fell to 0.3936 in 2023.
In the cooling-unit catalog record compressor modulation type (fixed staged, digital, inverter), fan type (EC/AC/VFD), valve type (EEV/TXV, PICV) and control mode (supply or return air).
Electronic expansion valves can save up to 15% versus thermostatic expansion valves.
Add an AI load-profile object to IT loads (step amplitude %, repetition Hz, duty cycle, peak % of nominal) and use it in UPS and battery sizing alongside autonomy.
Lab test on a large Li-ion UPS: load steps of 31-97% and 63-97% at 1 Hz kept input current nearly constant.
For layout iteration and add-load what-ifs, a surrogate that predicts only cabinet-inlet temperature (the compliance quantity) is cheaper and more accurate than full-field surrogates.
As a practical guideline, an initial dataset is about 500 simulation cases with around ten randomised parameters.
Include server fan power rising with inlet temperature in the cooling energy tool (for example about +13% from 68 to 104 F for A3) so set-point optimisation minimises total energy, not PUE.
If server fan energy is unknown, assume about 10% of operating load.
All content and rights belong to the respective authors and publishers. SoSo CAD provides only the citation, its own summary and a link to the source, and is not responsible for the sources' content.