Ask about N100 mini PC power consumption and you’ll get two confident answers that contradict each other: “it’s a 6W chip” from the spec sheet crowd, and “mine pulls 20 watts” from someone staring at a wall meter. Both are sort of right, which is exactly why the topic deserves a straight answer. The real numbers, pulled from the published reviews cited below: roughly 6 to 12 watts at the wall idling with a full software stack, and 20 to 30 watts flat out. For a box running 24/7, that idle number is the one that matters, and it’s the one the spec sheet never tells you.
The short answer, with the caveats attached
A complete N100 system, measured at the wall outlet:
| State | Typical draw | Notes |
|---|---|---|
| Idle, tuned, headless | 6–8 W | C-states working, no display, efficient PSU |
| Idle, typical | 8–12 W | Full service stack loaded, nothing actively working |
| Average mixed load | 12–15 W | A few containers doing real work |
| Full load | 20–30 W | All four cores pinned, brief peaks at the top of that |
The spread is real, not sloppy measurement. Notebookcheck’s review of the Minisforum UN100L measured 3.8 to 9.8 W at idle, just under 13 W under average load, and around 29 W in short bursts. ServeTheHome’s review of a fanless N100 firewall appliance landed at 10.5 to 12 W idle and 22 to 23 W maximum, high enough at idle that the reviewers called it out, but with a ceiling low enough that they ran the whole unit off a PoE+ splitter. That appliance is one of the fanless N100 firewall boxes covered in the pfSense hardware guide. Same CPU, different boxes, meaningfully different idle floors.
Two units with the same chip can sit 5 watts apart at idle. That’s the single most useful thing to understand about this platform, and it’s why measuring your own idle wattage beats trusting any table, including this one.
Why “6W TDP” is not what your meter shows
Intel lists the N100, a 4-core E-core-only chip boosting to 3.4 GHz, at a 6 W base TDP. Three things turn that 6 into the 8-to-12 you’ll actually see:
TDP is a thermal budget for the CPU package, not a system power figure. Vendors also set their own power limits per unit. In the UN100L, for example, the N100 runs a 7 W sustained limit (PL1) with an 18 W short-burst limit (PL2), per Notebookcheck’s review. Other vendors configure the same silicon anywhere from 7 to 20 W, which is part of why identical-looking N100 boxes benchmark and draw differently.
The CPU is not the whole computer. RAM, an NVMe drive, the NIC (2.5GbE ports idle hungrier than gigabit), a fan, Wi-Fi and Bluetooth radios all draw continuously. On a chip this frugal, the peripherals can out-draw the processor at idle.
The power brick taxes everything. Small external adapters are typically least efficient exactly where a homelab box lives, at low load. A few watts of your wall reading is the brick warming the air.
None of this is a scandal. It’s just the difference between a component spec and a system measurement, and the wall meter is the only spec sheet that matters for a 24/7 machine.
What a 24/7 N100 costs to run
Use your own electricity tariff rather than a national average. The energy-cost calculation multiplies power by operating hours, converts watt-hours to kilowatt-hours, and applies the unit rate:
kWh/year = average watts × 24 × 365 / 1000
cost/year = kWh/year × price per kWhFor continuous operation, each watt uses 8.76 kWh per year. That is about $1.60 a year per continuous watt at an illustrative $0.1834/kWh, or $1.75 at an illustrative $0.20/kWh. These rates are examples, not current utility prices.
| Average wall draw | Annual energy | At $0.10/kWh | At $0.20/kWh | At $0.40/kWh |
|---|---|---|---|---|
| 6 W | 52.56 kWh | $5.26 | $10.51 | $21.02 |
| 8 W | 70.08 kWh | $7.01 | $14.02 | $28.03 |
| 10 W | 87.60 kWh | $8.76 | $17.52 | $35.04 |
| 15 W | 131.40 kWh | $13.14 | $26.28 | $52.56 |
| 30 W | 262.80 kWh | $26.28 | $52.56 | $105.12 |
| 70 W | 613.20 kWh | $61.32 | $122.64 | $245.28 |
This table is calculated from assumed constant loads; it is not a set of hardware measurements. Enter a measured value and your tariff to calculate what it costs to leave running.
Warm idle, average load and time-of-use rates
A machine that spends 20 hours at 8 W and four hours at 20 W averages 10 W over the day: (20 × 8 + 4 × 20) / 24. Using only the 8 W idle reading would understate that example’s annual energy by 17.52 kWh. A plug-in meter’s accumulated kWh over representative days also captures scheduled backups and library scans.
For time-of-use billing, calculate kWh in each tariff window and multiply each window by its own price. Add those costs; do not average peak and off-peak prices unless the energy consumption in both windows is equal. Keep fixed supply charges outside an incremental server comparison when those charges remain payable without the server.
When replacing hardware pays back
At $0.20/kWh, replacing an assumed 70 W average tower with an 8 W average box saves 62 × 8.76 × 0.20 = $108.62 per year. If replacement costs $180, simple electricity-only payback is about 1.66 years. Add RAM, storage and adapter purchases before comparing total ownership costs.
By contrast, paying $60 extra to save 3 W saves just $5.26 a year at that same tariff: about 11.4 years to recover the premium. Free configuration changes can make sense; replacing working hardware for a tiny wattage difference may not.
A smaller tuning example illustrates the scale: at an assumed $0.19/kWh, reducing average draw from 8.6 W to 7.4 W changes yearly cost from $14.31 to $12.32, a $2.00 saving after rounding. These are calculation inputs, not measurements made by MiniLabHQ.
Solar and batteries still need an energy budget
An assumed 10 W continuous load needs 240 Wh each day before conversion losses. Lowering it by 3 W saves 72 Wh daily, which matters when sizing overnight battery capacity as well as the electricity bill. Solar equipment and battery costs also belong in the budget; a low annual grid bill does not establish solar payback.
For generation and battery planning, see running a homelab off solar. For short outages on grid power, use the separate UPS sizing guide.
For alternatives, compare HP thin clients using the same tariff and a measured wall reading. The low-power 24/7 server plan shows which components belong in the complete load budget.
Getting your box to the low end of the range
If your N100 idles at 12 W and you’d like it closer to 7, the levers are known and mostly free:
- Let the C-states work. Deep package sleep states are where the idle savings live. Check BIOS settings first; some vendors ship with C-states limited or ASPM off. The BIOS checklist for 24/7 mini PCs covers the specific toggles.
- Run headless. An attached display keeps parts of the iGPU pipeline awake. Notebookcheck’s 3.8 W idle floor is a best-case reading; a homelab box with a monitor plugged in won’t see it.
- Disable what you don’t use. Wi-Fi and Bluetooth radios, audio, unused SATA controllers. Each is small; together they’re a watt or two.
- Mind the OS.
powertop --auto-tuneon Linux gets most of the software-side savings. A hypervisor with a dozen idle containers still lets the package sleep; one chatty service polling every 100 ms does not. - Don’t over-optimize. The gap between 8 W and 11 W is about $5.26 a year at the illustrative $0.20/kWh rate. Spend an evening on it because tuning is satisfying, not because it’s an investment.
One non-power note that belongs in any 24/7 discussion: a box that’s always on and often internet-reachable is also always exposed. Keep it patched.
The bottom line
Budget 8 to 12 watts at the wall for a realistic N100 homelab setup, roughly $14 to $21 a year at an illustrative $0.20/kWh, with 20 to 30 watts as a ceiling you’ll rarely touch. The spec sheet’s 6 W is a component number, not a promise. If the difference between an N100 and its slightly hungrier siblings is what’s keeping you up at night, the N100 vs N305 vs N97 comparison settles it, but the honest summary is that every chip in this class is cheap enough to run that workload fit, not wattage, should pick your box.
Related across the network
- Firewall Hardware Tiers: Celeron vs N100 vs i5 Mini-PCs — firewallcompare.com
- Best Mini PC for a Docker Homelab in 2026: 3 Tiers Compared — dockerhomelab.com
- Best Mini PC for OPNsense Homelab 2026: NIC and CPU Sizing — opnsenselab.com
- Best Docker Containers for Your Home Server in 2026 — dockerhomelab.com
