M720q: Idle Manners | Suite Despair

This test started with a simpler question than “what is the absolute lowest idle number?” The real question was whether a used Lenovo ThinkCentre M720q could earn a place in the kind of home infrastructure this site cares about: small, tidy, always on, and not quietly wasteful.

Why Test It At All

The M720q is already a popular homelab box, and not by accident. It is cheap on the used market, small enough to disappear into a shelf or cupboard, and unusually flexible for its size, including a proper PCIe slot for people who want a little more room to adapt the machine to a job. If one of these boxes can also idle sensibly, run headless, and stay out of the way, it starts to look like exactly the sort of neat home-infrastructure component that deserves to be kept around.

Lenovo did publish idle figures for the M720 Tiny family back on 2018-04-18: 230 V numbers between 9.23 W and 16.88 W, depending on category, plus an ErP idle-state demand of 12.92 W. That is useful context, but it lives in the sort of eco-declaration PDF that is better at satisfying compliance than answering a homelab question. It also almost certainly reflects a very different test shape from a headless Debian 13.5 box with one NVMe drive and Ethernet.

So this page tests the narrower question that matters here: what this exact M720q did at the wall in a tidy Linux setup, and whether that result is good enough to fit the site rather than just flatter the hardware. The cleanest run was the post-BIOS tuned window: after the first reading, the remaining six points sat between 4.16 W and 4.41 W, with the full 30-minute run ranging from 4.16 W to 4.84 W.

Hardware And Method

The measured system used:

Lenovo ThinkCentre M720q
Intel Core i5-9400T
16 GB DDR4
Samsung 980 500 GB NVMe SSD
Lenovo 60 W PSU
wired Ethernet
Sonoff S60TPG smart plug at the wall

Measurement method:

boot the system
wait for SSH to become available
allow a 10-minute settle period
record wall-power readings every five minutes

The tuned runs used:

powertop --auto-tune
cpupower frequency-info
cpupower frequency-set -g powersave

The governor was already powersave before tuning. Linux also showed intel_idle, deep idle states through C10, and PCIe ASPM active on the NVMe path.

Lenovo’s PSREF confirms the broad platform shape used here: i5-9400T support, two DDR4 SODIMM slots up to 32 GB, M.2 2280 NVMe storage, and onboard Intel I219-V gigabit Ethernet. Intel’s official specifications cover the CPU basics retained here, SONOFF identifies the S60TPG meter, and Lenovo’s eco declaration supplies the family-level idle figures above. Those sources pin down the hardware context. The wall-power result still comes from the measured setup on this page.

Measurements

Session	Lowest	Highest	Notes
Soft off	0.00 W	0.00 W	single reading after shutdown; likely rounded down by the meter
Default idle, pre-BIOS	4.26 W	5.74 W	extended to 45 minutes; too noisy for a clean benchmark claim
Tuned idle, pre-BIOS	4.17 W	5.42 W	improved, but still spiky early in the run
Tuned idle, post-BIOS	4.16 W	4.84 W	cleanest run; last six readings sat between `4.16 W` and `4.41 W`

The interesting part was not that the box briefly touched 4.16 W. It was that the first passes were messier than the headline number wanted them to be, and only the cleaned-up post-BIOS run looked calm enough to trust.

The post-BIOS tuned readings were:

4.84 W
4.39 W
4.41 W
4.25 W
4.16 W
4.33 W
4.39 W

BIOS Changes

The BIOS review was modest rather than dramatic:

Smart Power On disabled
Bluetooth disabled
Max C-states already enabled
ASPM already on Auto
sound card already disabled

Linux-side inspection before the BIOS visit had already shown:

intel_idle active
deep idle states available through C10
NVMe link reporting ASPM L1 enabled
L1 substates enabled on the root port and NVMe device

That is not enough to hand out a neat little prize to each BIOS toggle. It is enough to say the calmest run arrived after a few unnecessary wake and device paths were trimmed back.

Does It Fit The Brief?

For this measured setup, yes. The M720q did not need a fantasy bare-board setup, a stripped benchmark image, or one decorative minimum to look respectable. Once the system had settled, it behaved like a small always-on node that could quietly do a job in a cupboard, on a shelf, or beside a router without feeling absurd.

In this measured Debian 13 setup, the Lenovo M720q settled into the low-4 W range at idle.

That is the useful summary. The lowest reading matters less than the settled band, and that band is what makes the result worth keeping. In this setup, the box behaved less like a cheap old office machine that happened to boot Linux and more like the sort of calm, low-drama infrastructure component the site is trying to champion.

Why It Matters

This is why the benchmark is worth keeping. It shows that a common used Tiny PC can fit the site’s low-power brief without stage-managing the test into nonsense. The point is not one magic minimum. The point is a believable low-4 W idle band in an ordinary headless Debian setup.