Skip to content
ServerCalc ServerCalc
Power & Energy

Server Power Consumption Calculator

Pick your server model, set the load, and get real watts, annual kWh, heat output, circuit draw and running cost.

Inputs

Server

Loads that model's chassis, socket, memory and PSU defaults.

Totals scale across the rack or fleet.

2U chassis · 2 socket(s) · up to 24 drive bays

Processors & memory

TDP drives most of the dynamic power.

W

Empty slots draw nothing.

512 GB installed, about 50 W

Storage & accelerators
W

NICs, HBAs, DPUs and RAID controllers not already counted.

Power supply & load
W

Rating of the supplies actually carrying load: one unit in a 1+1 hot-spare pair, the whole active bank on a GPU node. Nameplate, not draw.

Sets utilization. Pick Custom to drag the slider.

Three-phase carries the same watts at 1/√3 of the line current.

Real power over apparent power. Sets the watts-to-VA gap.

40%
W

Read it off iDRAC, iLO, XCC, IPMI or a metered PDU. Overrides the estimate and recalibrates the whole curve. Leave at 0 to estimate.

Operating profile & cost

24 for always-on production.

$/kWh

Your blended rate per kWh, taxes included.

Facility overhead multiplier. 1.0 counts IT load only.

Grams of CO2e per kWh on your grid.

Rack & facility

Defaults to a full 42U rack minus 2U for switching.

The breaker feeding the rack PDU.

sq ft

Optional. Gives watts per square foot for the room.

kg

kgCO2e to build one server, from the vendor's Product Carbon Footprint report.

Years the manufacturing carbon is amortized over.

Dell PowerEdge defaults follow the assumptions in Dell's Enterprise Infrastructure Planning Tool: Performance Per Watt BIOS profile, no chassis power cap applied.

  • A full rack peaks at 17949 W, more than the 4892 W a single circuit delivers. Budget 4 circuits per rack.
Results update live as you type.

Results

Estimated power draw
507 W
PowerEdge R760 at 40% CPU utilization
Idle draw
272 W
Powered on, no workload.
Max draw (100% load)
897 W
Sustained worst case, not a boost spike.
PSU efficiency here
95.0%
34% of the 1400 W nameplate
Annual energy (IT)
4,437 kWh
24 h/day × 365 days/year
Energy cost per month
$55.46
Energy cost per year
$666
Facility power (PUE 1.50)
760 W
Facility cost per year
$998
IT load plus cooling and distribution overhead.
Heat output
1,728 BTU/hr
Cooling required
0.14 tons
One ton of cooling is 12,000 BTU/hr.
Current draw
2.5 A
At 208 V, power factor 0.98.
Apparent power
517 VA
Minimum UPS size
646 VA
Apparent power plus 25% headroom.
Operational carbon
2,556 kg CO2e/yr
384 g CO2e/kWh applied to facility energy.
Manufacturing carbon
263 kg CO2e/yr
1,313 kg per server over 5 years.
Total carbon
2,818 kg CO2e/yr
Operational plus manufacturing. 9% of it is manufacturing.
Servers per circuit
5
30 A at 208 V Single-phase, 80% derate, 4,892 W usable.

Per rack

Rack draw at this load
10.1 kW
20 servers per rack.
Rack draw at 100%
17.9 kW
The figure the PDU and breaker have to survive.
Circuits per rack
4
Rack space used
40U
Of 42U available.

Where the watts go

507 W from the wall

At 40% CPU utilization, per server.

CPUs
268 W · 53%
Memory
50 W · 10%
Storage
74 W · 15%
Board, BMC & I/O
58 W · 11%
Fans
32 W · 6%
PSU conversion loss
26 W · 5%

Power vs CPU utilization

Idle is 30% of max draw

Server power is not proportional to work done. A box at 0% utilization still burns its idle floor, which is why consolidation beats tuning: switching a server off saves the whole curve, tuning it saves only the sloped part.

02505007501.0k0%25%50%75%100%Your setting: 507 WCPU utilizationWall power
Wall powerIdle draw: 272 WMax draw (100% load): 897 W

About this calculator

Server power consumption is the number everyone needs and nobody has. The sticker on the back of the chassis says 1400 W, the vendor quote says 'up to 2.5 kW', and the machine actually pulls about 500 W. Those are three different quantities, and only the last one shows up on your electricity bill.

This calculator works the number out from the bottom up. Pick your server model and it loads that chassis's real socket count, memory layout, drive bays and power supply. Adjust the CPU, DIMMs, drives and accelerators to match your build, set how hard the box actually works, and you get the watts it draws from the wall, what that costs per year, how much heat it dumps into the room, and how many of them fit on a circuit.

Three things separate this from the wattage-adder calculators you find elsewhere. It models the idle floor, because a server at 0% utilization still burns 30 to 45% of its maximum. It models PSU efficiency at the actual load point, not the single headline percentage on the 80 PLUS badge, because an oversized supply on a lightly loaded box is a real and avoidable cost. And it separates TDP from draw: TDP is a thermal ceiling the CPU only reaches under sustained all-core load, not the power it burns while your web server waits for requests.

The formula

The calculation runs in two stages. First, add up what the components pull on the DC side of the power supply, each with an idle floor plus a part that scales with utilization u (0 to 1):

P_cpu = sockets × TDP × (0.16 + 0.84u)
P_mem = dimms × W_per_dimm × (0.55 + 0.45u)
P_storage = Σ drives × (idle + (active − idle) × u)
P_gpu = cards × TDP × (0.12 + 0.88u)
P_fans = fan_idle + (fan_load − fan_idle) × u²
P_platform = board + BMC + NICs + controller (constant)

Second, divide by the power supply's efficiency at that load, because the PSU burns some of what it converts:

P_wall = (P_cpu + P_mem + P_storage + P_gpu + P_fans + P_platform + P_other) ÷ η(load)

Efficiency η is not one number. It follows the 80 PLUS curve for the supply's grade: it peaks near 50% load and falls off at both ends, so a 1600 W Titanium supply carrying 180 W runs at about 90%, not 96%.

Everything else is arithmetic on that wall figure:

kWh/year = P_wall × hours_per_day × days_per_year ÷ 1000
cost = kWh × price_per_kWh
facility kWh = IT kWh × PUE
BTU/hr = watts × 3.412 and tons of cooling = BTU/hr ÷ 12,000
kg CO2e/year = facility kWh × grid intensity (g/kWh) ÷ 1000

Current, phase and circuit sizing

Current depends on how the rack is fed. Single-phase and three-phase move the same watts, but three-phase does it at 1/√3 of the line current, which is why dense rows are wired that way:

single-phase: amps = watts ÷ (volts × PF)
three-phase: amps = watts ÷ (√3 × volts × PF)
VA = watts ÷ PF and UPS VA = VA × 1.25

PF is power factor, real power over apparent power. Server supplies with active PFC sit at 0.98 to 0.99, which is why watts and VA are nearly the same number on modern kit; capacity planners often use 0.95 for a mixed rack, and legacy non-PFC gear can be 0.8. It is an input here rather than a constant.

A branch circuit may only carry 80% of its rating for a continuous load, so:

usable watts = √3 (if three-phase) × volts × amps × 0.8 × PF
servers per circuit = floor(usable watts ÷ max draw per server)

Note the max draw, not the typical draw. After a power event everything spins up at once.

Rack and room

rack kW = P_wall × servers_per_rack ÷ 1000
racks needed = ceil(total servers ÷ servers_per_rack)
circuits per rack = ceil(rack max watts ÷ usable circuit watts)
watts per sq ft = total IT watts ÷ white space area

Lifecycle carbon

Electricity is only part of a server's footprint. Building it has a cost too, and the Green Software Foundation's SCI counts both:

embodied kg/year = manufacturing kgCO2e × servers ÷ service life
total kg/year = operational + embodied

Dell's published life cycle assessment for the PowerEdge R740 puts manufacturing at 1,313 kgCO2e of an 8,640 kgCO2e lifecycle total, and that is the anchor used for the model defaults. The share matters more than the number: on a US grid manufacturing is under 10% of the annual footprint, but on the French grid it is over 40%. The cleaner your electricity, the more the answer is 'keep the server longer' rather than 'buy a more efficient one'.

Calibration: a 2-socket HPE ProLiant DL380 Gen11 with two Xeon Platinum 8592+, 256 GB of DDR5 and a 1000 W Titanium supply comes out at 228 W idle here. Its published SPECpower_ssj2008 result measured 224 W. The loaded end runs a few percent high, which is the direction you want when the number is about to size a breaker.

Better still, don't estimate. Every vendor's answer to 'what does my server draw' is the same: read it off the baseboard controller. iDRAC, iLO, XCC, an IPMI sensor or a metered PDU will all tell you the real figure. Put that number in the Measured draw field and the calculator stops guessing: it calibrates the entire utilization curve to pass through your measurement, so idle, maximum and everything between stay consistent with the meter rather than with the model. The component breakdown is then apportioned using the modeled shares, which is the one part a single wall reading cannot tell you.

Common use cases

  • Sizing a UPS and PDU before a rack build, using max draw rather than PSU nameplate
  • Working out how many servers fit on a branch circuit, single- or three-phase, at the 80% continuous-load derate
  • Turning a single iDRAC or iLO reading into a full year of cost, heat and carbon
  • Planning rack density: kW per rack, circuits per rack, and whether the U actually fit
  • Putting a real annual electricity figure into a refresh business case, old model versus new
  • Estimating the CRAC or in-row cooling load in BTU/hr and tons for a new room
  • Checking a colocation quote: whether the kW commit you are being sold matches what your kit draws
  • Comparing the running cost of one big consolidated host against several small ones
  • Estimating the lifecycle carbon of a fleet, operational plus embodied, for an ESG or CSRD disclosure
  • Deciding whether to refresh hardware or keep it longer, once embodied carbon is in the picture
  • Deciding whether an 80 PLUS Titanium supply pays for itself at your actual load and tariff

Frequently Asked Questions

How do you calculate server power consumption?
Add the draw of each component at your working load, then divide by the power supply's efficiency at that load. Component draw is CPU (sockets × TDP, scaled by utilization), memory (about 3 W per DDR4 DIMM and 4 to 6 W per DDR5 DIMM), drives (roughly 6 to 9 W per 3.5-inch HDD, 2 to 4 W per SATA SSD, 8 to 15 W per enterprise NVMe), any GPUs at their TDP, fans, and a constant 35 to 60 W for the motherboard, BMC, NICs and storage controller. Divide that sum by 0.85 to 0.96 depending on the 80 PLUS grade and how loaded the supply is. Multiply the result by hours of operation for kWh, and by your tariff for cost.
How much power does a typical server use?
A modern 1U or 2U dual-socket server draws roughly 150 to 250 W at idle and 600 to 950 W flat out, landing near 350 to 550 W in normal production. Entry single-socket boxes run 60 to 220 W. Four-socket and dense storage chassis run 800 W to 1.5 kW. GPU nodes are a different category entirely: a 4-card inference server is 1.5 to 2.5 kW, and an 8-way H100 or MI300X training node is 8 to 12 kW on its own, which is more than a whole traditional rack used to be.
Is TDP the same as power consumption?
No, and conflating the two is the most common sizing error. TDP is the sustained heat load a cooling solution must handle, which for Intel and AMD server parts equals the sustained all-core power ceiling. Your CPU only reaches it under a genuine all-core workload. At idle a modern server CPU sits near 15 to 20% of TDP thanks to C-states and per-core frequency scaling, and at typical production utilization it is somewhere in between. This calculator scales CPU draw with utilization for exactly that reason.
Why does the calculator show less power than my PSU rating?
Because a power supply's rating is a maximum it can deliver, not a rate it consumes. A 1400 W supply in a server drawing 500 W delivers 500 W. Vendors ship generous supplies so the same chassis can take a full complement of GPUs and drives later, and so a redundant pair can carry the whole load if one fails. Sizing your UPS or PDU off nameplate rather than measured draw is how people end up paying for two or three times the capacity they need.
Does PSU efficiency actually matter?
At scale, yes, and the load point matters as much as the badge. Going from 80 PLUS Gold to Titanium is about 4 percentage points at half load. On a server pulling 500 W running 24/7 that is roughly 175 kWh a year, plus the cooling energy to remove the heat it no longer makes. Across a 40-server room at typical commercial tariffs it is real money. The subtler trap is oversizing: efficiency peaks near 50% load and falls off sharply below 20%, so a 1600 W Titanium supply carrying 180 W performs worse than a 800 W Gold one at the same load.
What is PUE and should I include it?
Power Usage Effectiveness is total facility energy divided by IT energy. At PUE 1.5, every watt reaching your server costs another half watt for cooling, UPS losses and distribution. Include it whenever you are asking what the machine costs to run, because that is the bill the facility pays. Leave it at 1.0 when you only want the IT load, for example when sizing a PDU. Hyperscale sites run 1.1 to 1.2, well-run enterprise rooms 1.4 to 1.6, and a converted office closet with a domestic split-system can exceed 2.0.
How do I convert server watts to BTU and cooling tons?
Multiply watts by 3.412 for BTU/hr, then divide by 12,000 for tons of refrigeration. A server pulling 500 W produces about 1,706 BTU/hr, or 0.14 tons. Essentially all electrical energy a server consumes leaves as heat, so IT load and heat load are the same number. Size cooling on maximum draw, not typical draw, and add margin: the day you need the headroom is the hot day when the chiller is already struggling.
How many servers fit on a 30 A 208 V circuit?
A 30 A 208 V circuit is 6,240 VA nameplate, but electrical code treats servers as a continuous load and derates to 80%, leaving 4,992 VA usable. At a power factor of 0.98 that is about 4,892 W. Divide by each server's maximum draw, not its typical draw, because you must survive everything spiking at once after a power event. At 900 W max that is five servers per circuit. This calculator reports that figure directly.
What is the difference between watts and VA for UPS sizing?
Watts is real power, the energy actually consumed. VA (volt-amperes) is apparent power, volts times amps, which is what a UPS and its wiring must carry. They differ by power factor. Modern server supplies with active PFC run at 0.95 to 0.99, so watts and VA are within a few percent. Older equipment at 0.6 to 0.8 power factor needs substantially more VA than watts. Size the UPS on VA with about 25% headroom for growth and inrush; this calculator reports both figures plus a recommended minimum.
Do redundant power supplies double the power consumption?
No. A redundant pair carries the same total load, either shared between both units or with one parked in hot-spare mode. Total draw is essentially unchanged, apart from a few watts of standby overhead on the idle unit. What redundancy does change is efficiency: when two supplies share a load, each runs at half the load ratio, pushing both into the less efficient part of the curve. That is why Dell, HPE and Lenovo all ship a hot-spare mode that parks one unit to keep the other in its efficient band. This calculator models efficiency against a single supply, matching that behavior.
Why is my measured power different from this estimate?
Common causes, roughly in order of impact: your CPUs are not at the utilization you assumed (check the BMC, not the OS load average); the BIOS power profile is set to Performance rather than a balanced or per-watt profile, which disables deep C-states and can add 40 to 80 W at idle; ambient temperature is high, so fans are working much harder than the model's normal cold-aisle assumption; or there is hardware in the chassis the inputs do not cover, like a 100 GbE NIC, a DPU or a battery-backed RAID cache. Add those under Other devices. The estimate targets ±10% on the machines it has published data for.
How do I calculate power consumption for a whole rack?
Work out one server's draw, multiply by the servers in the rack, then check it against the circuit feeding that rack. Rack kW = (server watts × servers per rack) ÷ 1000. The check that matters is the maximum, not the average: a rack of twenty 2U servers might sit at 10 kW in production but peak near 18 kW, and a 30 A 208 V single-phase circuit only delivers about 4.9 kW continuous. That rack needs four circuits, or a three-phase feed. Set servers per rack and the branch circuit rating in the Rack & facility section and the calculator reports rack draw, circuits per rack, rack units consumed, and watts per square foot.
What is the difference between single-phase and three-phase for a rack?
They deliver the same watts at different currents. Single-phase is watts = volts × amps × power factor. Three-phase is watts = √3 × volts × amps × power factor, so the same load draws about 42% less current per line. A 30 A 208 V circuit gives roughly 4.9 kW usable single-phase but about 8.5 kW three-phase, after the 80% continuous-load derate. That is why anything above about 5 kW per rack is fed three-phase, and why high-density and AI rows go further still to 415 V or 480 V. Select the phase in the calculator and the current and circuit figures follow the right formula.
Should I include the carbon from manufacturing the server?
If you are reporting a footprint rather than an electricity bill, yes. The Green Software Foundation's Software Carbon Intensity specification counts embodied emissions alongside operational ones, amortized over the hardware's service life. Dell's life cycle assessment for the PowerEdge R740 puts manufacturing at 1,313 kgCO2e out of an 8,640 kgCO2e lifecycle total. Whether that matters depends entirely on your grid: on a US grid at 384 g/kWh it is under 10% of a server's annual footprint, but in France at 56 g/kWh it is over 40%. On a clean grid, extending refresh cycles cuts more carbon than buying more efficient hardware. Your vendor publishes a Product Carbon Footprint report per model; use their figure when you have it.
Should I use the PSU nameplate or the measured draw for BTU and cooling?
Measured draw, always. Nameplate is what the supply can deliver, and vendor spec sheets often publish a maximum heat dissipation derived from maximum input power, which is higher again. Both overstate a real machine by a wide margin. A server actually pulling 400 W sheds about 1,365 BTU/hr and one pulling 700 W about 2,388 BTU/hr, regardless of whether the supply is rated 800 W or 1,400 W. Size cooling on the real maximum draw with margin, and get that maximum from a meter or from this calculator, not from the sticker.
Where do the model defaults come from?
Chassis figures come from vendor technical guides and QuickSpecs: socket counts, DIMM slots, drive bays and the power supply options actually offered on each platform. CPU and GPU TDPs are the manufacturers' published ratings. The PSU efficiency curves are the 80 PLUS internal-redundant targets at 230 V. The idle and load behavior is calibrated against published SPECpower_ssj2008 results. Every model default is a starting point you can override; if your build differs, change the fields and the math follows.

Spot an error? Have feedback?

Tell us what is wrong with the math, what is missing, or which server model you would like added. We read everything.

Related Calculators

Power & Energy

AI Server Rack Power Consumption Calculator

AI rack power calculator: enter how many GPUs you need and get racks, kW per rack, total megawatts, cooling method, heat load and cost per accelerator per year.

Open calculator
Power & Energy

NVIDIA Server Rack Power Consumption Calculator

NVIDIA rack power calculator for GB200 and GB300 NVL72, DGX B200, H200, H100 and HGX nodes. Get kW per rack, amps, circuits, liquid cooling load, weight and annual cost.

Open calculator
Power & Energy

42U Server Rack Power Consumption Calculator

Build a 42U rack from real servers, switches and storage and get its power draw in kW, amps, circuits, heat in BTU/hr, weight and running cost. Space, power and weight budgets in one view.

Open calculator
Cooling & Thermal

Server Room BTU Calculator

Server room BTU calculator for air conditioning sizing. Work out BTU/hr heat load from your servers, UPS, lighting and walls, then the AC capacity in BTU and tons to buy.

Open calculator
Virtualization

Terminal Server Calculator

Terminal server calculator for RDS session hosts: users per server, hosts needed, vCPU and RAM, profile storage, bandwidth and running cost. Built on Microsoft's sizing guidance.

Open calculator
Power & Energy

Dell Server Power Consumption Calculator

Dell PowerEdge power consumption calculator covering R770, R670, R7725, R760, R750, R740 and the AI line: XE9680, XE9680L, XE9685L, XE8640, XE9640, XE7745, R760xa.

Open calculator
Power & Energy

HP Server Power Consumption Calculator

HP / HPE ProLiant power consumption calculator for DL360 and DL380 Gen10 and Gen11, DL325, DL345, ML350 and DL20. Watts, kWh, cost, BTU/hr and amps.

Open calculator
Power & Energy

Supermicro Server Power Consumption Calculator

Supermicro power consumption calculator: 22 SuperServers from X11 and X12 through Hyper, CloudDC and the Blackwell AI line. Watts, kWh, cost, BTU/hr and rack amps.

Open calculator