How to Actually Measure Mini PC Idle Wattage (and Why Spec Sheets Lie)
Spec-sheet TDP isn't what a mini PC actually pulls from the wall. Here's how I measure idle wattage with a $25 Kill A Watt, what to expect from common N100/N305/Ryzen boxes, and the BIOS settings that cut 2–4W.
Every spec sheet I see for an N100 mini PC says “6W TDP.” That number is meaningless for homelab planning. The chip itself might use 6W under load, but the box around it — the SSD, the RAM, the NIC, the chipset, the power supply efficiency — adds another 4–10W on top before the system even idles.
I run nine mini PCs and thin clients 24/7. The single most useful $25 I’ve spent on this hobby is a Kill A Watt P3 P4400 ↗. Here’s how I measure each one, what the numbers actually look like, and the BIOS tweaks that save real watts.
What “idle wattage” actually means
There are three states worth measuring:
- Cold idle. Booted, no services running, nothing logged in. This is the lowest the system will ever go and it’s mostly useful as a sanity check.
- Warm idle. Your full Docker stack is up, every container is running, but nothing’s actively doing work. This is what your 3 a.m. wattage actually looks like.
- Average draw. Measured over 24 hours via the Kill A Watt’s kWh counter. This is what shows up on your power bill.
I report warm idle in my posts because it’s the most useful number. Spec sheets give you a CPU TDP, which only describes the chip. Reviewers sometimes quote cold idle, which understates your actual draw.
The measurement setup
You need three things:
- A wall-pluggable energy meter. Kill A Watt P3 P4400 (analog readout, $25), or any of the smart plugs with energy monitoring (TP-Link Tapo P110 or Kasa KP125M are common; both work and integrate with Home Assistant). The Kill A Watt is more accurate at low draw.
- The mini PC, fully configured with your actual workload running.
- A few minutes of patience. Don’t trust the first reading — let the system idle for 10 minutes before recording.
Plug the energy meter into the wall. Plug the mini PC into the meter. That’s it. Watts read in real time; cumulative kWh accumulates over time. After 24 hours, divide kWh by 24 to get average draw — that captures the brief CPU spikes that warm idle misses.
What I actually measured
These are warm idle numbers (full Docker stack up, no active load) from my fleet, measured with the same Kill A Watt over the course of a month:
| Machine | CPU | RAM | Storage | Warm idle |
|---|---|---|---|---|
| HP T620 Plus | AMD GX-415GA | 16GB DDR3L | mSATA + 1TB SATA SSD | 7.2W |
| HP T630 | AMD GX-420GI | 16GB DDR4 | 500GB M.2 SATA | 9.4W |
| HP T640 | AMD Ryzen R1505G | 16GB DDR4 | 1TB NVMe | 11.1W |
| Beelink S12 Pro | Intel N100 | 16GB DDR4 | 500GB NVMe | 8.6W |
| Trigkey N100 mini | Intel N100 | 16GB DDR4 | 500GB NVMe | 7.9W |
| GMKtec Nucbox G3 | Intel N100 | 16GB DDR4 | 1TB NVMe | 9.1W |
| Beelink EQ12 | Intel N100 | 16GB DDR4 | 500GB NVMe + 2TB SATA | 10.3W |
| Minisforum UN305 | Intel N305 | 32GB DDR5 | 1TB NVMe | 14.2W |
| GMKtec K10 (Ryzen 7) | Ryzen 7 8845HS | 32GB DDR5 | 2TB NVMe | 21.5W |
A few things that surprised me when I gathered these numbers:
- The T620 Plus is still the lowest-idle x86 box I own. Old AMD APUs on DDR3L pull less than modern N100 systems, mostly because the DDR3L SODIMM uses less power than DDR4 or DDR5.
- N305 systems are not as efficient as N100. The N305 has 8 cores vs the N100’s 4. Even at idle, that’s more power-management overhead. Expect 4–6W more than an equivalent N100 system.
- DDR5 costs you about 2W vs DDR4 in this class of system. Worth knowing if you’re choosing between two otherwise-similar boxes.
- Adding a 2.5” SATA drive adds 1.0–1.5W, even when idle. Worth it for storage, but factor it in.
- NVMe drives vary by 0.5–1.5W at idle. The Samsung 980 Pro is on the high end; the WD SN570 and Crucial P3 are on the low end. For a 24/7 box, the lower-power drives matter more than the speed difference.
BIOS settings that actually save power
After measuring stock idle, I tweak BIOS settings and measure again. These are the changes that consistently shave 1–4W:
1. Disable Intel/AMD turbo
Counterintuitive but real. Turbo lets the CPU spike to higher clocks under load; even at idle, the higher turbo ceiling slightly increases voltage and power management overhead. Disabling turbo on an N100 saves 0.5–1.0W idle. For homelab workloads that don’t need bursty performance, it’s free wattage.
2. Enable all C-states
Some BIOS configurations default to limiting CPU C-states (deep sleep states). C7 and C10 are where the real idle savings come from. Look for “C-State Control” or “Package C-State Limit” — set to maximum (C10 on modern Intel chips, similar on AMD).
This can save 1–2W on N100/N305 systems. On older AMD GX chips, less; they don’t have as many deep idle states.
3. Disable unused peripherals
Audio, secondary serial ports, parallel ports, secondary NICs you’re not using — each one shaves 0.1–0.5W. The savings are small but free.
4. Disable RGB
If your mini PC has front-panel LEDs you can disable in BIOS, do it. Most are 0.2–0.5W. The GMKtec models in particular have notable LED lighting.
5. Disable WiFi/Bluetooth radios
If you’re wired and have no plans to use the WiFi card, disable both in BIOS. 0.5–1.0W on most boards.
After applying all five tweaks on a Beelink S12 Pro, I dropped warm idle from 9.1W to 7.4W — a 19% reduction. That’s about 15 kWh per year, or roughly $2–3 in electricity. Over five years of 24/7 operation, it adds up to one nice meal.
What it costs at the wall
A few quick numbers for budgeting. At the US average residential rate of $0.16/kWh (2026), continuous 24/7 operation costs:
- 5W system: $7/year
- 10W system: $14/year
- 20W system: $28/year
- 50W system: $70/year
A typical N100 mini PC running a full self-hosted stack costs $10–$14 per year to power. A repurposed thin client comes in around $9. A Ryzen 7 mini PC running the same stack costs $25–$30 per year. None of these are dealbreakers, but if you’re running multiple boxes, the savings add up.
The right machine for your homelab depends mostly on what you’re running. The team at SelfhostRealm ↗ has a guide on sizing your hardware around your service list — start there if you’re not sure whether you need an N100 or can get away with a thin client. And if storage is a big part of your plan, TrueNASGuide’s hardware notes ↗ cover how disk choices interact with idle power on a NAS-focused build.
TL;DR
- Spec-sheet TDP is not your real idle wattage. Add 4–10W for everything else.
- $25 for a Kill A Watt is the best money you’ll spend on this hobby.
- Disable turbo, enable deep C-states, and turn off radios — easy 1.5–3W savings on any mini PC.
- Expect 7–10W warm idle from a well-tuned N100 system; 11–15W from an N305; 8–10W from a thin client.
Measure first, plan after. The spec sheet is the starting point, not the answer.
Related
The Intel N100 Mini PC Buying Guide for Homelabs
What to look for when buying an Intel N100 mini PC for a 24/7 homelab — RAM, storage layout, NIC quality, BIOS quirks, and the brands worth (and not worth) buying.
My $200 N100 Homelab: 11 Services, 8W Idle, Eight Months Uptime
A first-person walkthrough of my current N100 mini PC homelab — every part, every service, every BIOS tweak, and what the whole thing costs to run per year. Built and documented so you can copy it.
HP T620, T630, T640 Thin Clients as Homelab Servers
Used HP thin clients are still the cheapest entry into an x86 homelab. Here's what each generation actually idles at, what you can run on them, and the upgrade gotchas worth knowing before you buy.