Dell Server Power Consumption Calculator

Dell publishes the Enterprise Infrastructure Planning Tool for exactly this job, and it is a good tool. It is also behind a login, it only knows Dell hardware, and it will not tell you what any of its numbers mean. This calculator does the same arithmetic in the open, on the PowerEdge models people actually deploy.

Twenty-three PowerEdge chassis are preloaded with their real socket counts, DIMM slots, drive bays and power supply banks. That covers the 17th generation R770, R670, R7725 and R6725, the 16th generation R760, R660, R750 and R650, the older R740 and R730 still everywhere in the field, the 4-socket R860 and R960, and the entry R360. Pick one, adjust the CPU, memory, drives and accelerators to match your build, and you get wall watts rather than nameplate watts.

The AI line is here too, and it is a different order of magnitude. The XE9680 and the liquid-cooled XE9680L and XE9685L carry eight SXM accelerators; the XE8640 and XE9640 carry four; the XE7745 takes eight double-wide PCIe cards; the R760xa takes four. A mainstream R760 peaks near 900 W. An XE9685L with eight B200s peaks past 10 kW, which is more than a whole rack of traditional servers, in 4U. If you are planning an AI row, that ratio is the entire problem.

If you have the hardware in front of you, skip the estimate entirely. iDRAC already knows. Under Power Management, the current and peak system draw are right there. Put that reading in the Measured draw field and the calculator calibrates its whole curve to your machine instead of to a model.

Wall power = (CPUs + memory + drives + GPUs + board + fans) ÷ PSU efficiency at that load

Each component contributes an idle floor plus a share that scales with CPU utilization. A PowerEdge CPU at idle sits near 16% of its TDP with C-states active, so a dual Xeon Gold 6430 R760 idles around 270 W and only approaches 900 W when both sockets are genuinely pinned.

Dell-specific things that move the number:

• System Profile in BIOS is the big one. Performance disables C-states and can add 40 to 80 W at idle over Performance Per Watt (DAPC). The defaults here assume a per-watt profile, which is what Dell's own planning tool assumes.
• Redundancy policy. In a 1+1 configuration Dell's Hot Spare mode parks the second supply so the active one stays in its efficient band. The efficiency curve here is modelled against a single supply for that reason. Turn Hot Spare off and both units share the load at half the load ratio each, where efficiency is worse.
• Power cap. iDRAC can enforce a chassis-level ceiling. Nothing here models a cap; if you have one set, your measured figure will be lower than the maximum shown.
• Fresh Air / high ambient operation. Running warm to save on cooling costs more in fan power. The fan figures here assume a normal 22-25 °C cold aisle.
• Power supply banks on AI nodes. A mainstream PowerEdge runs one active supply with a hot spare. The XE nodes run banks: the XE9680 has six 2,800 W units in 4+2, the XE7745 has eight 3,200 W units in 4+4. The PSU capacity field here holds the capacity of the units actually carrying load, not one unit's rating, which is why it reads 11,200 W rather than 2,800 W on an XE9680.
• Direct liquid cooling. The XE9640, XE9680L and XE9685L move heat with coolant instead of a wall of 40 mm fans. That is worth 150 to 250 W per node in fan power alone, and it is why the XE9640 fits four SXM accelerators in 2U. It does not reduce the heat, it just moves it somewhere a CDU can deal with.

Facility numbers follow from the wall figure: BTU/hr = watts × 3.412, amps = watts ÷ (volts × PF) single-phase or ÷ (√3 × volts × PF) three-phase, and kWh/year = watts × hours × days ÷ 1000.