Create separate Organizational Profiles for a facility's IT, OT and building management systems (for example electrical and cooling SCADA/BMS) and state each Profile's scope.
The Core has 22 Categories with Subcategories; the outcomes are not a checklist.
Keep embodied-carbon amortisation (TE x T/EL) in procurement, refresh and design-option comparisons, and use operational carbon only for runtime placement and scheduling recommendations.
Scheduling by SCI raised datacenter carbon by nearly 30%; the penalty was over 15% at 40-80% utilisation.
When the user raises the inlet setpoint, include server fan power and airflow increase (flat/low/moderate/high CFM-vs-inlet curve) and CRAH fan and count changes; warn when PUE improves but total energy rises.
TCO unchanged, 6% better and 6% worse respectively.
Define AI load not only by energy but by peak/average (about 1 in training), peak/idle ratio and ramp/decline rates; make sub-second dynamics inputs for UPS, generator and grid-connection studies.
Traditional computing is 10-15 kW/rack versus 30+ kW/rack for AI workloads.
Add passive and active ducted racks and ducted HACS to the containment selector, with density limits tied to a ceiling plenum vacuum input (default 0.03 inH2O / 7.5 Pa, warning above 0.06 inH2O / 15 Pa).
Cooling-to-IT airflow ratio: 1 for HACS, 1.2 or higher for ducted racks.
Compare HVAC options not only by on-site water use but by total water footprint including the grid's water intensity (WUEsource); the ranking can flip with the grid.
Outbound data traffic carries a water footprint of 1-205 litres per gigabyte; the high end is close to the 214 L footprint of 1 kg of tomatoes.
In a design tool, set CFD accuracy expectations from this review's ranges (0.1-4 °C absolute, relative mostly below 20%) and do not limit validation scope to temperature; ask for velocity and pressure too.
Of the 58 CFD studies, Fluent was used in 24 (41.4%) and 6SigmaDC in 12; standard k-epsilon in 41 (about 70.7%).
For paralleled modules in a modular UPS, do not look only at the N+1 redundancy count; add to the design check that module-to-module filter and control parameter differences cause circulating current and thermal imbalance.
The authors recommend keeping overshoot or sag on module switching within 10% of nominal critical-bus voltage.
In preliminary cooling selection, make climate (humidity, haze/dust) and the free-cooling switchover point decisive inputs; direct airside economisers may be unsuitable in haze- and dust-prone regions.
An indirect evaporative cooling unit in Shenzhen achieved a PUE of 1.249 at 5.3 kW per cabinet.
Calculate the permissible time for a cooling outage with a pipe-volume energy balance and set the backup-cooling start time accordingly (here 300 s); high-temperature, contained designs give less time.
From 18 °C chilled water and 25 °C supply air, coil capacity fell below the load and inlet temperatures rose rapidly.
When evaluating turning liquid-cooled data-centre waste heat into electricity, weigh RTE (76-83%) against specific cost (767-1125 USD/kWh); higher RTE can come with a cost increase.
System cost is about 767-796 USD/kWh depending on the cycle combination.
In large-load interconnection negotiation, present AI load flexibility (a deferral window of up to about 3 hours) as a lever for interconnection capacity and timing; returns diminish beyond 3 hours.
In the IEEE 14-bus study, data centres on buses connected to congested lines benefit most from geographic load shifting.
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.