A Look At The Linux vs. Windows Power Use For A Ryzen 7 + Radeon RX Vega Desktop

[cynical]

Most interesting from the Windows vs. Linux metrics was Windows 10 having a significantly lower peak power use at 198 Watts compared to 214~226 Watts on the four Linux distributions tested.

While the idle stat looks promising, I'm not sure the peak stats say much on their own since we are not taking into account the amount of work done per energy spent. Is the peak power lower because Windows is more efficient, or does it maybe alter the cpu frequency in some way? It would be interesting to see some of your recent benchmarks (where you showed Linux was generally faster if I remember right) and compare the performance difference to the difference in power usage in relative terms.

[starshipeleven]

This is a great example of why 450 watt power supplies are usually more than enough. Or 550 Watt, 650 watt. The 800-1000 watt power supplies are ridiculous with current tech.

They were supposed to run SLI/Crossfire (multiple cards in the same system) setups, even older tech didn't usually require more than 600w for a CPU + GPU + a couple HDDs.

Now they are mostly used in home servers with dozens of drives, as SLI/Crossfire basically imploded years ago.

[edwaleni]

Modern PC's when "off" but still draw power could be doing a range of functions.

Some are providing power to the USB ports for charging devices.

Some maintain power to the ME to be able to turn on the PC for a WOL event.

HP started putting LED's on the outside of the consumer PC's because people were sticking their hands inside the PC to change RAM before the capacitors had finished discharging. It notified them that power was still present.

Many workstations and servers, when plugged in will power cycle and have the ME do a low level diag on the system, then shut off. The device can't be turned on unless someone pushes a button after that.

With Linux tending to extract more from hardware historically, I would expect it (in general) to pull more power when it works, and less power when it isn't. Because Windows is not as demanding, or doesn't push the hardware as hard when demand warrants, will use less peak, and is not as good as going into low power states (since it isn't good coming out of them periodically).

So the tests I see here, are totally in line with that perception.

[Holograph]

I've been saying for years that most Linux distros catering to old computers for no good reason is environmentally unfriendly. Code should run efficiently just like everything else should run efficiently. Old furnaces are replaced with more efficient ones. And you wouldn't run an air conditioner while your windows are open. New cars are generally more fuel efficient. etc. Linux is common enough that having more optimized software on more deployments of it would have a tangible benefit I think. Of course the software optimization crusade should apply to other operating systems as well, but Windows doesn't have multiple distros and thus by design must be somewhat more generic in hardware support.

Too bad Gentoo isn't power efficient with each user compiling their own stuff (though at least there are reasons for it as USE flags would create way too many different combinations to distribute in binary form). Clear Linux is looking good. Not nearly as much in the repository but it seems to keep getting better.