WinFix: i7 owners Win7 by default is crippling our CPU's

[garikfox]

i7 owners this is a must read below

http://www.overclock.net/t/1250666/winfix-i7-owners-win7-by-default-is-crippling-our-cpus

 

[ShintaiDK]

I would be careful. MS would have fixed ti if there was something to coem after. So the problem is, whats the downside.

 

[Phynaz]

I call BS.

Have anything to backup your claims?

 

[Denithor]

Would be curious to see if there are measurable differences with this enabled/disabled.

Looks like when enabled it prevents W7 from properly utilizing HT cores. Wonder if only on i7 or if also on i3? Because I run W7 and when encoding on my i3 (DVD Shrink) all four cores peg at 100%. Nothing else I do really pushes my cpu...

 

[N4g4rok]

I'm willing to bet that utilization of the threads remains a specific application issue. Windows might see some benefit from it, but applications running on 4 threads aren't going to suddenly start running on 8.

 

[exar333]

Seems a little like a 'tin foil' solution to me. I highly doubt something this impactful would have been in Win7 this whole time...

 

[garikfox]

Would be curious to see if there are measurable differences with this enabled/disabled.

Looks like when enabled it prevents W7 from properly utilizing HT cores. Wonder if only on i7 or if also on i3? Because I run W7 and when encoding on my i3 (DVD Shrink) all four cores peg at 100%. Nothing else I do really pushes my cpu...

If the Intel CPU has HyperThreading then I think this is great information

 

[Topweasel]

The downside and what this ignores is that HT is there not only to increase performance but smooth out system use. Core parking was developed specifically because of HT. Back in the day systems would run worse when HT was enabled because applications would treat a HT "core" as a real core and schedule to it like it was a real core and it would cause the two threads to compete heavily with the real cores affection.

Disabling means that poorly written applications can once again bang heads against itself (though the Win 7 scheduler should keep this to a minimum) and that your machine will now handle all eight cores as though they are real, meaning that even if its not budding heads with each other intensive apps or intensive multitasking will cause your machines responsiveness to drop.

Which is why the update that started this whole thing was meant for BD systems. Windows was treating BD like the extra cores per module were HT cores.

 

[Don Karnage]

My Cinebench scores went up .02 with core parking disabled. Should go up more for my 3770K

 

[garikfox]

I'm willing to bet that utilization of the threads remains a specific application issue. Windows might see some benefit from it, but applications running on 4 threads aren't going to suddenly start running on 8.

Yes it does, example with parking enabled when I compress a file with WinRAR it only uses 4 threads, When parking is disabled then WinRAR will use all 8 threads.

It's also made a noticeable impact with the game BF3 also.

 

[TheRyuu]

For example: With a i7 2700K when cruising the net with Firefox/Safari etc. or compressing a file with WinRAR youll notice that only 4 threads are being used not 8. The core parking feature of Windows 7 is crippling our performance.

I wasn't aware that browsing the web was the new standard for looking at how well an application will make use of all my cpu cores. It sort of seems like the same thing as saying "well, winrar really sucks at being multi-threaded since it can't even fully utilize one core" (because it's i/o limited).

Why do we even care about what the cpu utilization looks like. I was under the impression that what matters is the real elapsed time of an application when you want to measure the performance of something not the way the pretty graphs look in task manager.

 

[Don Karnage]

I wasn't aware that browsing the web was the new standard for looking at how well an application will make use of all my cpu cores. It sort of seems like the same thing as saying "well, winrar really sucks at being multi-threaded since it can't even fully utilize one core" (because it's i/o limited).

Why do we even care about what the cpu utilization looks like. I was under the impression that what matters is the real elapsed time of an application when you want to measure the performance of something not the way the pretty graphs look in task manager.

What are you talking about? Web surfing needs MOAR CORES

 

[garikfox]

Lol

 

[TheRyuu]

I'm not saying there may not be a benefit of doing what it says to do, I'm just saying that the method he's using to prove it isn't very good. Actually I suppose I mean the method he's using the prove there's actually an issue in the first place isn't very good (or both).

 

[Don Karnage]

I'm not saying there may not be a benefit of doing what it says to do, I'm just saying that the method he's using to prove it isn't very good.

For processors with Hyperthreading its worth turning core parking off.

 

[Topweasel]

Yes it does, example with parking enabled when I compress a file with WinRAR it only uses 4 threads, When parking is disabled then WinRAR will use all 8 threads.

It's also made a noticeable impact with the game BF3 also.

I would watch out. BF3 for example is one game maybe the only game that has been proven to while not max out every core, at least utilize more then 4 cores and upwards as far as we can tell 8 (though no discernible difference in performance over 6). WinRar can use all the cores. But what you have done is with at least Winrar, is limiting your systems multitasking capabilities as windows then has to find a break in core usage, to make room for other processes.

If we want to go back into the dark ages of computer gaming where we kill every other task running in the background to game, then this is good for a little extra performance. But I am not sure its worth not being able to browse the web or play a game because I am running a task and felt like running this "fix".

 

[jimpatrick]

I wasn't aware that browsing the web was the new standard for looking at how well an application will make use of all my cpu cores. It sort of seems like the same thing as saying "well, winrar really sucks at being multi-threaded since it can't even fully utilize one core" (because it's i/o limited).

Why do we even care about what the cpu utilization looks like. I was under the impression that what matters is the real elapsed time of an application when you want to measure the performance of something not the way the pretty graphs look in task manager.

lol'ed at this.

 

[Stuka87]

Telling Windows to treat a hypertheaded core as a real core is going to do nothing but hurt performance. Because a hyperthreaded "core" is not a real core. You will ultimately run into some strange contention issues if you disable core parking. As the CPU cannot always push through multiple threads at once, so one (the hyper threaded thread) has to end up waiting. Where as if that thread had been run on a real CPU instead of the HT one, that issue would not have occurred.

 

[Don Karnage]

Telling Windows to treat a hypertheaded core as a real core is going to do nothing but hurt performance. Because a hyperthreaded "core" is not a real core. You will ultimately run into some strange contention issues if you disable core parking. As the CPU cannot always push through multiple threads at once, so one (the hyper threaded thread) has to end up waiting. Where as if that thread had been run on a real CPU instead of the HT one, that issue would not have occurred.

Never had an issue with my 3930K when i had core parking disabled

 

[TheRyuu]

Never had an issue with my 3930K when i had core parking disabled

I wasn't aware that the opinion and experience of one user could be blanketed upon the entire world of users with SMT cpu's such as:

For processors with Hyperthreading its worth turning core parking off.

From everything I've read so far that seems like suggestion based on nothing. The only things I really see things "feel" better or task manager happens to graph stuff in a more eye pleasing way so it seems a bit premature to draw that conclusion.

 

[SHAQ]

I'll try it after Tweakboy tests it out.

 

[Rvenger]

I'll try it after Tweakboy tests it out.

+1

 

[LOL_Wut_Axel]

LOL, no...

You guys realize the claims been made here are patently false, right? Core Parking, or core power gating, simply shuts down a core when it is inactive. If the program has x% of utilization, the core will be awaken instantly so the task can be completed. It doesn't reduce performance in any way, especially since in the situations when it's enabled (like when you're browsing the web or listening to music) CPU utilization is too low to take advantage of having core gating turned off.

 

[Blades]

Turning this off would be a good way for a stress test of a selected vcore, making sure that your comp doesn't crash under less than ideal situations - better put that core parking may mask insufficient vcore at certain clocks. That said, its better to leave this on - SMT scheduler is the Linux equivalent, correct?

 

[IntelUser2000]

On Windows 7, it priotizes using physical cores over logical cores because consumer applications tend not to be as well threaded.

On server Windows it turns whatever logical cores off.