Deleting A Few Lines Of Code Can Yield "Significant" Power Savings On Some Linux Systems

[andyprough]

That's why Linux needs to increase its market share to get more users for more bug reports and more developers to try and fix stuff.

Linux is a kernel, not an OS. The Linux kernel has massive market share.

[PapadakosPanagiotis]

Sometimes I am just wondering about the reviewing processes of the kernel code.

[zdzichu]

sounds like it would not make a difference on laptops?
I think my PC has 4 pci express slots only one is used

Looking at PCIe slots tells you basically nothing. Check

lspci

(look for Bridge). Or even better

lspci -t

to see how it is all connected. Laptops certainly have no fewer bridges than desktops - ExpressCard, ThunderBolt, NVMe and USB4 are all exposing PCIe.

[gfunk]

lspci -t

to see how it is all connected. Laptops certainly have no fewer bridges than desktops - ExpressCard, ThunderBolt, NVMe and USB4 are all exposing PCIe.

on my Dankpad, 15 things came up..

I have my NIC disabled which apparently saves power I guess this would do something similar?

[Adarion]

Just wondering: Maybe there was a reason to do so in the past? What if some PCI2PCIe bridges were misbehaving? I remember some problem with an ASMedia (1083 before rev. 3??) chip for example. Maybe they didn't want to switch it on by default, but later things got forgotten and it was never switched on (+ blacklisting or "work-arounding" for the faulty chips)?

[ms178]

For something being this "simple", this has been overlooked for way too long. How many power on earth could have been saved by implementing it right from the start?

[Raka555]

Just wondering: Maybe there was a reason to do so in the past? What if some PCI2PCIe bridges were misbehaving? I remember some problem with an ASMedia (1083 before rev. 3??) chip for example. Maybe they didn't want to switch it on by default, but later things got forgotten and it was never switched on (+ blacklisting or "work-arounding" for the faulty chips)?

I was also thinking that they put it there to fix some problem. Hopefully that "problem" does not exist anymore.

[Azrael5]

Ahahahhaha this further proves that linux developers have boycotted linux operating systems and end-users as well, for years!!! the best collaborators of micro$$$oft!

[lowlands]

For something being this "simple", this has been overlooked for way too long. How many power on earth could have been saved by implementing it right from the start?

Not much if not insignificant when put in perspective: how much power could have been saved by *not* mining bitcoin?

[peterdk]

I tested the patch on Ubuntu 20.04, with a Ryzen 3700 server that is continously power usage monitored. It did not seem to have any effect. Or not more >1W. I use powertop to auto-tune all optimizations. It might have more effect on Intel systems, since they have much more powerstates that are dependent on ASPM.