Random power measurements on three OPNsense installs

[Linwood]

I was doing some testing on various boxes and had these numbers handy so if of interest.

Relatively vanilla install of opnsense, no IPS or other heavy CPU utilization.

Code Select Expand

Beelink EQi12 Intel Core 1220P with 1g NIC's --- 13 watts

GMKTec M5 AMD Ryzen 7 5825U with 2.5g NIC's   --- 25 watts

(Vintage 2016) I7-6700k in Z170WS with (new) X550-T2 dual 10g (@ 10g) -- 60 watts

I am NOT recommending either of the above mini-PC's, especially the GMKtec where the NIC's are not even recognized out of the box, you have to get the plugin onto it (hard without a working NIC).  The Beelink does work quite well, though has Realtek NIC's as well, but they seem to work without the plugin (and fail with it).  That it's Realtek is troublesome, but that they are not multi-gig makes it not very future proof.  But it does work and has been stable. 

But I found the power interesting.  I also did some math -- that old huge PC (which was a monster in its day) may draw 60watts, but to buy a MiniPC to replace it at, say, $400, has a power-payback of almost 6 years (and maybe not that really as 10g NIC's probably are drawing more power by their nature, same with 2.5 vs 1g).

Anyway... just some random numbers for anyone interested.  Does make me more likely to reuse an old PC rather than buy new for OPNsense.

[BrandyWine]

What's the power with no ethernet cables attached?
What about BIOS differences?

Is it just a matter of software configuration (bios, OS, etc)?

[Linwood]

I don't know some of those answers, I was saying "random".

The old big PC was convenient and had the 10g X550-T2 card. Leaving the card in but having the ethernet disconnected dropped the power (once things stabalized) from 60 to 48 watts. I would think the X550-T2 is going to be a worst case compared to 2.5 or 1g cards.  The 2.5g GMKtec is on its way back so can't test, but the Beelink is only 13 watts anyway, so it can't make much difference.

Also, unplugging the NIC's causes OPNsense to basically do nothing (once it starts up) as there is no traffic, so the drop in power may well be partially CPU not NIC's.

I assume there are settings in bios that could change this. By and large mine is set at defaults, not over-clocked, though if there are explicit power saving modes I see I usually turn them off.

This was not scientific data.  It was just numbers I had handy.  What I found most interesting is the MUCH higher draw of the old, huge CPU still has a 6 year payback in power (at my rates in USA, North Carolina) costs.  Kind of kills my justification of saving money by saving energy.

[pfry]

Interesting. I have a fanless i3-9300t (TL;DR: 32GB RAM, 800GB NVME SSD, X710-DA2, i210, i219) that only consumes around 30W idle. This thing (TL;DR: Ryzen 7700X, 32GB RAM, 1.6TB NVME SSD, X710-DA4, X550-T1) consumes around 45W idle. From memory, as they're not plugged into a UPS with a power readout or a split power cable (for a clamp ammeter) at the moment. Of course, the CPU alone on the Ryzen can draw another 150W or so, which will ramp up the fans as well, and the main fan can draw up to 65W. The i3-9300t maxes out at around 100W (system total) - the 9300t is really a 65W device with turbo enabled.

Er, let's just say you wouldn't save much money with either of those.

I'm a bit surprised your Skylake is so high. Of course, my measurements and/or memory could be off.

[Linwood]

I'm a bit surprised your Skylake is so high. Of course, my measurements and/or memory could be off.

I'm using three different UPS' for the measurements (due to where they PC's were), so it's also possible their readouts are not the best.