Is it ever worth disabling Windows P-states? (same as running "High Performance" always)

Edex

Distinguished
Nov 8, 2008
20
0
18,510
Hi!

There seems to be no benchmarks and no discussion about this topic. Is it worth running windows on "High Performance" (no P-STATES = always 100%) or "Balanced"?

1) Is it worth turning it on for everyday activities? Zipping, Internet browsing, general noticeable system responsiveness?
2) Is it worth turning it on only at high performance scenarios (working with software that uses a lot of CPU)?
3) Or just leave it on "Balanced"?

(I do have six cores and at most points at least on of them is in the highest p-state, I wonder whether Windows would start to run a suddenly demanding application in an already full-power core and then transfer once another one powers on... probably not).

My reasoning behind asking this is Skylake SpeedShift. I'm still on Haswell, and theoretically this would negate any difference whatsoever. But it's possible SpeedShift is more for mobile and for desktops it's not noticeable with powerful rigs anyways.
 
Solution
Yes, shitty programing does that too but you see, I have few programs keeping tab on performance on third screen and can instantly see what's going on with all cores. Now, FX processors have core boost (kind of turbo) that boosts only one core when SW demands it, not all of them but at constant rate they all jump at almost same time which produces better fluidity.
May I also add that Windows 10 tend to equalize core to core load a bit better than W7 does.
Benchmarks also tend to keep load at same level so they may not be good at showing stuttering and core responsiveness.
I keep all performance on full, 24/7 on overclocked system. That's on desktop computer of course, laptops are different matter all together. Power savings are small as most power is used during heavy work and than turbo and high performance kick in anyway.
 
I OC my FX 6350 to 4.8 GHz and it's quite noticeable performance increase. With proper cooling (Scythe Mugen3 and push pull fans) temps are no problem at all. Only at CPU usage close to 100% does it draw full power. There's less than 20w of more power consumption when it's used at full potential and that kinda offsets pickups while turbo kicks in. A lot of micro stuttering can be attributed to changing power states as it doesn't happen instantly.
 
Judign by the popoluarity of this post and the lack of info on this topic it's pretty easy to tell it doesnt impact anything. There are no benchmarks done. Your have anecdotal evidence. Stuttering is mostly caused by shitty programming and this point if you monitor resources most software is too slow even if nothing bottlenecks it.

For me its 34W instead of 24W. As I run all fans on minimum I got 3C higher liquid coolant temp (~39, max 41 - under daily work load). Someone should test this shit, we are just speculating.
 
Yes, shitty programing does that too but you see, I have few programs keeping tab on performance on third screen and can instantly see what's going on with all cores. Now, FX processors have core boost (kind of turbo) that boosts only one core when SW demands it, not all of them but at constant rate they all jump at almost same time which produces better fluidity.
May I also add that Windows 10 tend to equalize core to core load a bit better than W7 does.
Benchmarks also tend to keep load at same level so they may not be good at showing stuttering and core responsiveness.
 
Solution