SONOFF Smart Plug Energy Monitoring: Total PC Power Consumption only between 32-50 watts?? Is this thing reliable?

soup

Smack-Fu Master, in training
54
I bought a SONOFF Smart Plug that has Energy Monitoring to feed my curiosity about how much power my two PCs take and how much adding a NVIDIA T400 GPU changes it.

Total PC Power Consumption w/ i7-4790S and NVIDIA T400 only between 32-50 watts. Right now typing this message in the web browser and it's staying under 40 watts. Note that's pretty much the only thing running, a few tabs in the browser and no other windows running or doing anything.

My other PC with 2 SSDs and 1 5400 rpm HD and just integrated graphics (i5-10400) has similiar power draw but I imagine the HD and other SSD are sleeping most of the time as they are not accessed / not in use.

Is this smart plug reliable? I feel like the power draw should be more than twice that. I thought that Intel CPUs that are rated at 65 watts were notorious for sucking up twice that amount of power at the drop of a dime.

Granted when I play Valhiem on the i7-4790S at 1020p with medium to high settings it hovers around 85 watts, but thats the total system draw, with SSD, and the GPU around 95% utilization.
 
Last edited:

soup

Smack-Fu Master, in training
54
The other strange thing is the sensor on my Nvidia T400 with HWinfo is detecting that the card is at 18-19 watts while I'm typing this message. Which only leaves about 21 watts for the rest of the system??? Somethings not right, perhaps the T400 power sensor.

40 watts for total system draw I feel like thats nothing and better to worry about buying a better refrigerator or install new windows in my house for the summertime air conditioner to save on electricity.
 

continuum

Ars Legatus Legionis
94,897
Moderator
Sounds in the ballpark. I won't say "right" but definitely sounds in the ballpark.

I have extensively documented power consumption of quite a few systems over the years and those numbers you give-- 32 to 50W AC from the wall outlet sound right for a typical Haswell-era box with a low-end GPU at idle. A modern Comet Lake box should be about half that at idle.

I really should dig up all those years of Post-Its and put them in a spreadsheet...
 

malor

Ars Legatus Legionis
16,093
You're using a CPU that has a TDP of 65W, with a GPU that has a TDP of 30W. As you say, Intel's TDP figures are kinda fictitious, so I'd expect your max theoretical system draw to be somewhat under 150W, including your motherboard, memory, and drives, all maxed out at once, which basically never happens.

My guess would have been that Valheim would take about a hundred watts, but 85W is in the ballpark. The weak GPU may be throttling the CPU somewhat, keeping total power consumption down. And idle power around 40W sounds normal-ish for a machine of that vintage. Newer Intel chips idle cooler, but run much, much hotter under load.

I think your meter is probably giving you fairly accurate results.
 

teubbist

Ars Scholae Palatinae
823
It's a Haswell generation CPU, PL2 was 1.25 of TDP. And an S SKU as well. The silly PL2 excursions are a more recent thing with Intel CPU's and mostly apply to K-series, where Intel is willing to turn a blind eye to motherboard "default" silliness.

So yes, the readings are likely accurate.

But you can verify the accuracy with something that produces a known load, like a soldering iron, halogen or incandescent bulb and/or a ceramic heater(the motor in the fan is small enough relative to the heater load to be largely ignorable).
 

steelghost

Ars Praefectus
4,975
Subscriptor++
I use a few items of Sonoff-branded "smart" gear, reflashed with Tasmota firmware, for a variety of power and / or temperature measurement jobs around my house. They're not the last word in lab-grade accuracy for sure, but for the price they are fine; based on the rough and ready calibration / validation I did on a couple, probably 95% accurate. Take that anecdata for what it's worth (aka, yes, the readings you're getting seem sane enough as everyone else has said, I'm coming at it more from the "can you trust your measuring stick" end of things).
 
I really should dig up all those years of Post-Its and put them in a spreadsheet...
If you believe you have enough data, that would be insanely helpful to a lot of people, I think. A Github page linked to places like homelab forums and subreddits would be an awesome resource for people just starting out or looking to build lower power systems, or with used parts. Especially those in places where electricity isn't as cheap as in Canada and the US.
 

soup

Smack-Fu Master, in training
54
It's a Haswell generation CPU, PL2 was 1.25 of TDP. And an S SKU as well. The silly PL2 excursions are a more recent thing with Intel CPU's and mostly apply to K-series, where Intel is willing to turn a blind eye to motherboard "default" silliness.

So yes, the readings are likely accurate.

But you can verify the accuracy with something that produces a known load, like a soldering iron, halogen or incandescent bulb and/or a ceramic heater(the motor in the fan is small enough relative to the heater load to be largely ignorable).
Great idea! I tested that with a small incandescent bulb rated at 40 watts on the package, the SONOFF smart plug detected ~41.6 watts.
 

soup

Smack-Fu Master, in training
54
You're using a CPU that has a TDP of 65W, with a GPU that has a TDP of 30W. As you say, Intel's TDP figures are kinda fictitious, so I'd expect your max theoretical system draw to be somewhat under 150W, including your motherboard, memory, and drives, all maxed out at once, which basically never happens.

My guess would have been that Valheim would take about a hundred watts, but 85W is in the ballpark. The weak GPU may be throttling the CPU somewhat, keeping total power consumption down. And idle power around 40W sounds normal-ish for a machine of that vintage. Newer Intel chips idle cooler, but run much, much hotter under load.

I think your meter is probably giving you fairly accurate results.
Hrmmm..... It's quite possible I disabled intel turbo boost in the bios and/or under the win10 power settings, which is probably why it's only 85 watts under load. The case is crowded and very small and I am afraid of over heating it, I'll check later.
 

soup

Smack-Fu Master, in training
54
Hrmmm..... It's quite possible I disabled intel turbo boost in the bios and/or under the win10 power settings, which is probably why it's only 85 watts under load. The case is crowded and very small and I am afraid of over heating it, I'll check later.
Ug, I forgot, the BIOS was indeed set to "maximum non-turbo mode", with the attached pic for windows 10 power settings:

BoostMode.PNG

The gpu had no limitation set.

I will rerun this later with Intel Turbo Mode enabled if anyone is interested.
 

soup

Smack-Fu Master, in training
54
With Turbo Mode/Boost enabled, the i7-4790S and total system took anywhere between 85 - 105 watts on a run through Valheim, instead of hovering around 80 - 85 watts with Turbo Mode/Boost disabled. I'm guessing if all threads and GPU at ~100% it would peak even higher with Boost Mode enabled, I doubt Valheim is maxing out all available threads.
 
Coincidentally, i got a Sonoff S31 recently, was going to hook it up to the 3d printers to calculate power use, but i've been using it with the PC for the moment:

A 13600k (undervolted) on a B760 itx, a 2070, 4+1 fans, 2 DDR5 sticks, 2 NVME ssds all powered by a CX550-F and it idles/does basic browser tasks around ~50w from the sonoff app.

Breaking it down, have to multiply it by .8 to remove PSU losses (as a bronze 80+ PSU) leaving around ~40w, Afterburner metrics show the CPU idles at 3-5w and GPU at 11w, let's say combined 12w (12v/1A) for all 5 fans and the chipset TDP of 6w leaves ~9w for two SSDs and 2 DDR5 sticks which actually looks... correct? (kinda wish i could still get the GPU to go down a little more)

Also gaming usually consumes around ~300w from the outlet and most of it is the GPU at 190-210w, with the CPU consuming less than 70w most of the time....
 
Do you think it's accurate, though?
Likely just good enough. I also have a clamp-on ammeter for my DMM, but havent checked if they both measure the same (or at least close enough).

FWIW, i wouldnt be surprised if the newer CPUs are just faster at doing tasks so they dont use up too much power/go back down to idle quick, I still have my older PC beside it ATM (a 4790 non-k and a GTX 970) and (IIRC) the CPU never idles lower than ~20w (reading MSI afterburner) while internet browsing.