[Intel N100] Can a processor's power limits by bypassed or its consumption misrepresented?

Let’s see if you can help me make some sense of this, I must preface it all saying I don’t have a technological background, but I am always learning, so there are bound to be some misconceptions here.

I have two mini-PCs, both using Intel’s N100 processor, but they are different (Firebat T8 Plus and TexHoo QN10). I don’t use the first one a lot, it’s connected to the TV, but I wanted to experiment with adding a firewall to my home network, so I got the other one.

From what I could gather, Power Level 1 (PL1) defines the sustained power consumption of the processor and PL2 the maximum power draw for short bursts, there is another setting for how long the processor can be in that PL2 state before going back to PL1; PL3 and 4 also exist, but I don’t think they are being used.

The thing is, the processor has a TDP of 6 W, but I think both manufacturers decided to operate it off-spec; and in the case of the QN10, I don’t even think the power reported is correct.

The T8 has both PL1 and PL2 set to 15 W by default, shouldn’t PL1 be set to 6 W or thereabouts? I suppose they wanted to give users the impression of having a fast system, but I’d rather it lasted instead; that enclosure is small enough already. TexHoo set the PL1 to 6 W and PL2 to 25 W for the QN10, I understand the PL1 value being 6 W but driving it at 25 W seems insane even if only on short bursts.

Intel doesn’t state the reference values for those power states, do they? The only thing I could find is the processor TDP, but I also saw that it doesn’t equate to power drawn… so I lost track there, it seemed a bit over my head.

I configured them both in the BIOS to 7 W for PL1 and 12 W for PL2, and I used a power meter before the power adapter (I used the same for both) and then HWiNFO and ThrottleStop to monitor test things and see how they go.

Firebat T8 Plus:

(alternative link to the recording on Vimeo, set it to 1080p to read values well)

Things seem to work as I thought they would, for around 8 or 9 seconds the processor draws 12 W, then the time window ends and goes back to PL1 drawing 7 W for the remainer of the test.

The power meter registered a peak of around 15 W, falling back to 10.x when the PC was in that second stage power capped by the PL1 limit.

TexHoo QN10:

(alternative link to the recording on Vimeo, set it to 1080p to read values well)

I don’t understand this, I have both configured the same in the BIOS for a fair comparison, but this one is operating differently, as if limits didn’t exist, or reporting wrong power consumption. As soon as I started the test, ThrottleStop and HWiNFO showed the package power being around 8 W, which was already higher than PL1, but it never was power limited at any point, so it never went back to 7 W, and the power meter clocked 23 W of power throughout.

Is it possible for package power reported to be wrong (intentionally or not)? I see a lot of watts unaccounted for, there were 14 W of difference between idling and running the test, while the reported package power difference was only 5.5 W.

If it is not possible, why the different behaviors?

I would rather have the machines running within spec, I thought setting PL1 to 7 W was already overdriving it (QN10 has PL1 being 6), but I am worried that this second machine is misreporting things allowing the processor to run unrestricted (except for the turbo multipliers I think).

Because that is the one that would be used as a firewall, I appreciate that it can go faster, but it would be in operation 24/7 and probably seeing high CPU usage.

Thanks anyway!

There should be a thing in HWinfo64 that’s called power discrepancy or something like that, they build mobos power delivery to sometimes cheat the power limit and that should tell you how accurate it is

1 Like