Seasons of Jason
Over the next few weeks, I'm going to TRY to accurately measure my power consumption doing daily work tasks on my big tower PC, the Steam Deck, and the HP Dev One laptop. (Using a Kill-A-Watt P3).
ALSO? Gaming power consumption comparisons between my desktop PC & the Xbox Series X, both at 4K resolution.
I expect it will be an eye-opening experience and I'll never turn on my tower again.
This will undoubtedly branch off into all kinds of software testing too. YES, I'll record & share data...
Seasons of Jason
@NO_PULSE_XBL same here. It should be enlightening!
NO PULSE
@killyourfm ooof that makes me nervous to even think about. Especially when I’ll leave my PC on and go downstairs and do something else for half the day. 😭💀
Seasons of Jason
@trezzer FASCINATING. Standby as in idle, or standby as in sleep?
Trezzer (aka Helvedeshunden)
@killyourfm I'll not spoil the details for you, but I actually found that the standby power usage when not doing anything at all was far worse than the couple of hours I actively used my gaming rigs.
Seasons of Jason
@darth @trezzer No no, I don't mind spoilers. It's good to compare data!
Darth, are you using a Kill-A-Watt? Measuring power at the wall?
Linux Renaissance
@killyourfm @trezzer I just checked his claim on my Windows 10 PC. It uses about 80W on idle, literally doing nothing. Sorry for the spoilers hehe, but this is a bit too much for doing nothing. I'll have to check the Linux numbers later.
Seasons of Jason
@darth @trezzer Would you be so kind as to start a doc somewhere or shoot me an email with some of your findings?
It's for a future L4E Episode ;-)
Linux Renaissance
@killyourfm @trezzer wall power, yes.
Seasons of Jason
I ran some power consumption tests on #Windows 11, measuring from the wall.
My system: Ryzen 9 3900X, Radeon RX 6800 XT, 32GB RAM.
1% CPU utilization (Idle): 67W
5% CPU + 10% GPU (Streaming with Plex): 130W
Dirt 5 Benchmark (4K, High Preset, Uncapped FPS): 431W
Dirt 5 Benchmark (4K, High Preset, 60FPS Cap): 355W
Dirt 5 Benchmark (1440p, High Preset, 60FPS) Cap) : 288W
The power savings gained JUST from capping FPS to 60 is more than 2x the total power of a Steam Deck under heavy load.
Seasons of Jason
@trezzer @darth Just so I have some background, do you have access to Windows + Linux on your desktop and laptop?
Having Windows isn't really that important, but I'm absolutely doing comparative testing ;-)
Trezzer (aka Helvedeshunden)
@darth @killyourfm My AMD/Nvidia laptop uses around 40W while idle/light use. My Linux desktop slightly less. My M1 Mac... single digits.
Seasons of Jason
@vsaw This is JUST the PC! (I can only measure one input on the Kill-a-Watt)
Valentin Sawadski
@killyourfm Thats a lot! My 2018 MacBook under full load uses the same amount of power like your idle system.
And to clarify does this include the screen(s) and peripherals or just compute?
Valentin Sawadski
@killyourfm If you are interested in measuring your entire setup, You could get an extension cord plug all devices in and measure all of them at once with the Kill-a-watt.
Seasons of Jason
@vsaw Huh. that's a fairly simple solution I hadn't thought about!
Valentin Sawadski
My guess is though, the screens won’t make a big difference probably 20-50W. At least my “old” 1080p screen was using this much.
Seasons of Jason
@vsaw True, true. And what I'm really trying to capture is the power consumption differences between operating systems, form factors, and eventually proprietary vs open-source software.
Seasons of Jason
I'll repeat the same tests with #Fedora 38 #Linux tomorrow.
What I'm beginning is a journey to capture the power consumption differences between operating systems, form factors, and eventually, proprietary vs open-source software.
Please watch this space.
Seasons of Jason
@trezzer @darth that’s far from sad, but I know what you’re saying 😉
Trezzer (aka Helvedeshunden)
@killyourfm @darth My Linux machine is a dedicated box, sadly.
Seasons of Jason
@lminiero So, in 2019 I published this: https://www.forbes.com/sites/jasonevangelho/2019/05/06/benchmarked-the-most-power-efficient-ubuntu-19-04-flavor-will-surprise-you/
The takeaway was that GNOME had a very similar power envelope compared to Xfce (surprising, yes) and KDE was indeed pretty power hungry.
MATE was the most power efficient at the time. I am VERY much looking forward to revisiting these tests with better tools and a few years of optimizations in the rearview mirror.
Lorenzo Miniero
@killyourfm Fedora Linux with which desktop environment, though? Gnome and KDE are known resource hogs compared to XFCE and others for instance. Or is this part of the software analysis that will follow?
Seasons of Jason
@lminiero To answer your question directly: the flagship, which would be GNOME.
Seasons of Jason
@PublicLewdness @lminiero I really loved what @wimpy told me back then about improving MATE's efficiency:
"MATE is power efficient due to an aggregation of small margins. The margins between the competitors are so small there is no one thing you can improve to leap frog your opponents. So, you identify lots of small margins of improvement. Singularly they don't move the needle. Collectively, you win gold."
PublicLewdness
Seasons of Jason
Starting #Fedora 38 power consumption tests, and decided to see how much extra power gets consumed by simply switching refresh rate from 60Hz to 120Hz (as this demands more work from the #Radeon card).
[As a refresher, my this is a tower PC with an #AMD Ryzen 3900X and Radeon RX 6800 XT]
- 1% CPU utilization (idle) at 60Hz: 67W (this is the same result as Windows 11, btw)
- 1% CPU utlilization (idle) at 120Hz: 97W
So an extra 30W pulled from the wall just to double my desktop refresh rate.
Seasons of Jason
In the UK, at 8 hours of operation per day, that 120Hz refresh rate boost would only result in about £30 per year.
But that ecological footprint multiplied by millions of people? Well, that might truly suck. Especially when we add up all the other "quality of life" things we enjoy with modern desktop PCs.
(I don't know how to make these calculations, and maybe this is something you would excel at @baldpolnareffart?)
Bald Polnareff Art
@killyourfm I love what you're trying to do!
How are you collecting the data? Keep in touch, I'm interested at doing something with it
Seasons of Jason
@baldpolnareffart Right now I'm just throwing notes and results in Joplin. Not very organized yet. But I would like to make all this public, and have other people running similar tests.
Seasons of Jason
@vsaw @fershad
Some context around that super cool fact: I watched Joseph's presentation (from KDE) at LAS last weekend about their Blue Angel certification and software sustainability -- that is precisely what inspired me to start this testing!
(I'll also be seeking the same certification for Thunderbird)
Valentin Sawadski
@killyourfm Oh there is loads of fun to be had with this. A while back I found a research group looking into energy consumption of video streaming on Android. TLDR, 720p cuts it down in half compared to 1080p https://s2group.cs.vu.nl/2022-01-05-green-lab-experiment/
Valentin Sawadski
@killyourfm Here are some people and organizations you could check out if you want to learn more about energy efficient software
the software and sustainability research group https://s2group.cs.vu.nl/
https://www.thegreenwebfoundation.org/ which build the awesome CO2.js library
@fershad who works for the green web foundation and has a CO2 aware homepage
https://www.blauer-engel.de/en/productworld/resources-and-energy-efficient-software-products a German certificate. Super cool fact: the first software to get the certificate is #KDE #Okular
Seasons of Jason
@Zelfir Agreed! This is something iOS and Android are already doing, right?
Mago27 🇪🇺🇨🇵🇺🇦🐧
@killyourfm
We really need a variable refresh rate. There is no point to stay at 120Hz all the time.
We could use a lower refresh rate while typing and only bump it before playing animations.
Seasons of Jason
@s31bz @baldpolnareffart I will be measuring a LOT, and publishing all the data...somewhere :D
S31bz :linux:
@killyourfm @baldpolnareffart that sounds perfect for him. Will you be measuring differences between Fedora and Windows throughout the weeks? I could possibly point to this data as to why I'll migrate from my desktop to a Framework 16 by the beginning of next year, if the power savings help the choice along lol
Seasons of Jason
@baldpolnareffart I'd prefer looking to your for guidance on methodology. Let's take this to email!
Bald Polnareff Art
@killyourfm I think before collecting data from various different people, we should try to organise some sort of guideline to follow, just to make sure everyone who's involved is on the same page.
We could start collecting individual data and try to look for interesting patterns before asking the community to submit anything
Seasons of Jason
@mathieucomandon hardmon looks really useful! After all, CPU and GPU are going to drive the majority of power usage.
Does it depend on the machine having various system sensors available? (For background, I use to use Phoronix Test Suite for stuff like this, but yours gets straight to the point and would probably be a better option for me).
Mathieu Comandon
@killyourfm that's a really cool project! I started a small tool to gather this kind of data (https://github.com/lutris/hardmon) but it's for CPU and GPU only. What do you use to measure power consumption from the wall? I have some smart plug which might do the trick
Mathieu Comandon
@killyourfm would be worthy to note that your low (288W for Dirt 5 at 1440p) would pretty much be my high. I measured around 210W max from a 5700XT and 80W max from a Ryzen 5 2600. 4k gaming has a much heavier impact on power consumption than 120Hz gaming. I haven't tried 4k gaming, but with FSR and DLSS it may no longer be necessary to render at such high resolutions?
Seasons of Jason
@mathieucomandon And now you have my brain working overtime.
What would the power savings of DLSS & FSR look like when applied to millions of games? Damn....
Bald Polnareff Art
@be4foss @killyourfm That was gold, thank you!
Seasons of Jason
@baldpolnareffart @be4foss Indeed, that handbook is a great resource.
And I think the Expert Power Control 1202 might be JUST A BIT more useful than my Kill-A-Watt :D
KDE Eco
An established methodology is one of the benefits of the Blue Angel ecolabel award criteria. We document the process in the KDE Eco handbook:
See these publications for information about the methodology.
"Sustainable software products -- Towards assessment criteria for resource and energy efficiency" (Sec 4.1)
https://doi.org/10.1016/j.future.2018.02.044
And this German Environmental Agency report, pp. 22-27 summary in English:
https://www.umweltbundesamt.de/publikationen/entwicklung-anwendung-von-bewertungsgrundlagen-fuer
Seasons of Jason
*Furiously scribbles notes*
KDE Eco
You can follow Engineer Detlef Thoms, who does relevant back-of-the-envelope calculations in the "Sustainable Programming" course (see 04:20–06:10).
Calculate the energy savings, make certain assumptions about use (e.g., how often the savings apply), and multiply out. Given the global nature of software, the numbers scale up very quickly.
https://open.hpi.de/courses/cleanit2021/items/5DHsS3tJsXAqfUE4q4F82Z
GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.
All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.