Disable Core Parking is a bad tweak!?

[BenYeeHua]

http://www.tweakhound.com/2013/07/15/bad-tweak-core-parking/

Basically if you are using Processor Power Management in Windows 7 & 8 at any setting under 100% then you are using core parking.
If you don’t want to use core parking:
via Control Panel > Power Options > Pick your Power scheme(s) > Change plan settings > Processor power management > set Minimum processor state to 100% > press Apply and OK buttons.
or
From Microsoft White Paper â€" Performance Tuning Guidelines for Windows Server 2012
To turn off core parking, set the Processor Performance Core Parking Minimum Cores parameter to 100 percent by using the following commands:
Powercfg -setacvalueindex scheme_current sub_processor CPMINCORES 100
Powercfg -setactive scheme_current

lol
He don't know which setting for the power profile to disable it!
But he know which Powercfg commands to disable it!

http://bitsum.com/about_cpu_core_parking.php

So be careful if you want to trust him, as he don't do research correctly and test for it.

[Jeremy Collake]

That's his opinion . He's likely just focused on expanding the scope of his site.

We've well documented the positive effects of disabling CPU parking under at least some workloads.

I would disagree that it is a bad tweak, though it is true that users should be aware of the consequences of disabling it. Since it's a power profile setting, the decision to disable it for at least the High Performance power profile seems fairly solid. This won't affect mobile devices and laptops, which normally use a different power profile when on battery power.

Also, power savings aren't completely lost with CPU parking disabled, as frequency scaling still occurs.

[edkiefer]

He lists it as bad tweak IHO .
Which I can understand for general user with laptops, not advanced/power users .

Not sure whats up with his picks with power management, I think its just bad copy/paste on system with no support for HT .
As those are defaults for high performance and there is no core parking option . Which is how mine is with i5-3570k (non HT) .

[BenYeeHua]

Ya, I also wonder why windows don't disable core-parking for High Performance.
As a laptop user, I also disable it, on AC, so when I am using battery, it still works, without switching the Power Plans.
---
And ya, I also found out some game has become more better(for example, AC3, which is CPU bound), I wonder did it cause by some windows 8 update?
Because the CPU usage has been increased to 100%

[andy123]

I really like your way of expressing the opinion and sharing the information.

[BenYeeHua]

You are welcome.
And most of the time I will not posting something like this if I am not very sure I am right.

[chris635]

I think this could be a bad tweak depending on the circumstances. I had my cores un parked on fx 8350 clocked at 4.9 ghz. While trying to check internet speed and certain games my kids play (anything using flash online to work, mainly flash games and tools to check internet speed online), my system would freeze up. It took several days to figure it out. If I dropped my overclock down to below 4.4 ghz it would be stable with core parking dis-abled. Over 4.4 and my system would freeze. Once I re-enabled the core parking feature everything worked great even with my high overclock. Food for thought with high overclocks and core parking dis-abled.


By the way Process lasso works great with my overclocks

Chris

[edkiefer]

that sounds like you were not totally stable with all cores enabled, probably voltage drops more with all cores .
Either raise your LLC up or just add few steps of Vcore .

[chris635]

I tested for a week straight with insanely high voltages and on my current clock I pass 12hrs + on prime 95, occt, aida64, and intel burn test (avx)on high mem settings. Now I do understand it could be the fact the I am also running 2400mhz ram, but I doubt it based off my testing. It runs fine on everything but piddly flash games through the browser (some not all). Once I re-enable core parking, the problem goes away for high overclocks.

[chris635]

Upon checking my settings in the bios I noticed I neglected to turn off C6 State. Thats what was causing my freezes with certain flash games with core parking dis-abled. Once I turned off C6 state and unparked the cores everything ran fine. So it looks like depending on how your system is set up, unparking the cores can cause some problems with high overclocks. Something I didn't catch earlier. This can be a bad tweak if your system isn't set up right for it, as mine clearly was until I pulled my head out of my butt and fixed it.

[BenYeeHua]

Upon checking my settings in the bios I noticed I neglected to turn off C6 State. Thats what was causing my freezes with certain flash games with core parking dis-abled. Once I turned off C6 state and unparked the cores everything ran fine. So it looks like depending on how your system is set up, unparking the cores can cause some problems with high overclocks. Something I didn't catch earlier. This can be a bad tweak if your system isn't set up right for it, as mine clearly was until I pulled my head out of my butt and fixed it.

Yes, I hear some people is having freeze after they disable it, and I also lack of research on it, as I don't has the similar system.

Based on what you has report at here, it is because core-parking will force most of the core/module having more time on C6/C3 state.
But, after you disable the core-parking, and your has overclock+having a bad LLC, while it(windows AMD CPU driver) switch more often between C6, C3 and C0.

So, when most of the module switch from C6 to C0, which is very low voltage to Vcore that you set, and also the highest frequency which is 4.9 ghz, if you disabled the C&Q.
And even worst, the LLC is not kick in time, and causing this.

url=http://forums.anandtech.com/showthread.php?t=2318675#5]http://forums.anandtech.com/showthread.php?t=2318675#5[/url]]The problem with LLC, or any voltage tweaking for that matter, is that the voltage changes faster over time than any software app is going capture and report...but the CPU itself will crash the moment the voltage drops below a critical threshold value.

Voltage readouts you are getting from your BIOS or from CPUz may seem like they are updating in realtime but they are only sampling the actual voltage in steps over time.

CPUz is even worse because it doesn't even show you the observed voltage, it quantizes the voltage in 0.008V increments, and to make matters even more worse it rounds down the number when reporting to you.

So if the BIOS is reporting "1.287V" for Vcore then CPUz will take that number and report to you that the voltage is "1.280V"...a full 0.007V lower than what it actually is.

So let's say your CPU really does need 1.286V and not a drop lower, but CPUz is telling you it is running along just fine with 1.280V. (which is not true, because it would crash at 1.280V) So you are led to think your chip is fine at 1.280V.

Now you start up prime95, and so long as LLC keeps the Vcore at a value between 1.280V and 1.288V then CPUz is going to keep telling you the Vcore is 1.280V.

Well lets imagine the actual voltage is 1.287V, just a smidge above the 1.286V minimum threshold. CPUz is still saying it is 1.280V but the CPU is stable nevertheless. Now a momentarily change in the Prime95 load causes a transient voltage to drop that Vcore from 1.287V to 1.285V...boom, CPU is now unstable.

But CPUz is going to keep on reporting to you that the voltage is still 1.280V.

In short, CPUz sucks and in ways that fool people into thinking their system, or parts of it like LLC, are not acting rationally when infact they are. It is CPUz that is problem as it is generating crap data for you.

url=http://forums.anandtech.com/showthread.php?t=2318675#16]http://forums.anandtech.com/showthread.php?t=2318675#16[/url]]You don't need perfect programs, you just have to know the limitations of the ones you are using.

Take the statement of yours above which I quoted from your OP. The problem you are observing is not that the "average" voltage is high enough or too low, rather the problem is that you are blind to the transient voltage swings that are occurring during loading and unloading (which can happen on the timescale of micro-to-milliseconds).

The "maximum negative undershoot" value during idle and during loading is what causes your processor to become unstable, not the average voltage over the long-run at idle or under load.

PIC that I can't quote at here.

^ those voltage oscillations are called "transients" and your BIOS and software-based voltage monitoring software are blind to them because they happen very very quickly (milliseconds) but when they happen they can still cause your processor to be borked.

url=http://hardforum.com/showpost.php?p=1036997079&postcount=329]http://hardforum.com/showpost.php?p=1036997079&postcount=329[/url]]LLC on REGULAR and Offset @ +0.140 = 1.368v under load (1.088 on idle) PRIME STABLE
LLC on MEDIUM and Offset @ +0.095 = 1.368v under load (1.040 on idle) PRIME STABLE
LLC on HIGH and Offset @ + 0.060 = 1.368v under load (1.000 on idle) PRIME STABLE
LLC on ULTRA HIGH and Offset @ + 0.025 = 1.368v under load (0.982 on idle) - BSOD
LLC on EXTREME and Offset @ - 0.020 = 1.368v under load (INSTANT DEATH)

I guess you are using LLC if I am right, and because LLC is based on the load of the CPU, and having a delay before it get kick in.
So when the vdroop is too high, because of the core-parking is enabled, it is lucky that the other module is still in c6/c3 states, and not causing more vdroop to the core/module that is using.(it still may causing some damage for a long time.)

But, it become a issues when core-parking is disabled.
When it is just having a little load, only some module will be switching between c6/c3, c1e and c0 at difference time, which might not a issues.
When it got a lot of load, the OS will wake up all the core at the same time, from very low voltage to vcore with 4.9 ghz, and bang.
The LLC is late to this party, it is lack of vcore/voltage to let all core/module continue to works at that Frequency, and it is freeze.

I think it only happen with medium load, which switch between C6 and C0 very often, and the LLC is not design very well for C6, which can reduce the voltage.
------
I might be wrong on this case, as I am lack of information about the computer, and also knowledge, but I think it is LLC and/or vdroop only if it is overclocked.

[chris635]

I think you may be right.

     "But, it become a issues when core-parking is disabled.
When it is just having a little load, only some module will be switching between c6/c3, c1e and c0 at difference time, which might not a issues.
When it got a lot of load, the OS will wake up all the core at the same time, from very low voltage to vcore with 4.9 ghz, and bang.
The LLC is late to this party, it is lack of vcore/voltage to let all core/module continue to works at that Frequency, and it is freeze.

I think it only happen with medium load, which switch between C6 and C0 very often, and the LLC is not design very well for C6, which can reduce the voltage."

Good explanantion

[BenYeeHua]

Yup, I guess the LLC just don't know how much it should add when the voltage it switch between C6 voltage(which might be 0V too, but I guess this is not possible)and vcore too many times, it can be the LLC adding too much voltage, adding too less voltage, or it might kicking too late like when the core/module is C6 states, which is worst...

Except we has a special tools that connect to the motherboard VRM etc, and record with a very high interval, we will never know the answer.

[edkiefer]

I am not that familiar with AMD stuff , but as has been posted when you OC so much the range of voltage and current gets extended .
This makes it harder for voltage to keep on track with load .
only thing you can do is either raise LLC, raise Vcore (offset mode) or go manual mode, static voltage . Also you can disable the lower voltage states as you have done/mentioned .

[BenYeeHua]

I think it should be lower the LLC, and increase the offset, but it is not as useful as disable C6, which can just prevent the LLC become worst.
Raise LLC is not a good choose, as LLC is not respond enough to keep a better voltage, as the info shown about.

[edkiefer]

from what I see it very dependent on the MB platform .
67, 77, 87 they all behave differently, OC wise . Even MB manufacture have different LLC implements .

I don't think there right, wrong, you just need to test what works best according to CPU, MB etc your running .

[chris635]

I agree with testing to see what works. I am using the Asus Crosshair V Formula Z motherboard, top notch overclocker. My vcore is 1.48v with LLC set to ultra high, it will jump to 1.512v for 4.9 ghz. With this kind of set up, having core parking dis-abled but having C6 state enabled just don't play well togerther. Choose one, either have core parking dis-abled with C6 state dis-abled or have them both enabled for my set up.

[BenYeeHua]

Just a question, did Core-parking get automatic disabled by OS after you disable C6?
I think Core-parking works for C3 state enabled too, but too bad I am using laptop...

And yes, if they really having a good solution for LLC implements(I am not very know much about LLC), like having 5Ghz etc. to control the voltage, I think they will making it as ads, but I guess most Motherboard only saying like VRM!!! only.

[edkiefer]

many people like to set higher LLC and be able to run a lower offset or even manual voltage . Many think that because the voltage spikes and then drops (Vdroop) that, that is bad .
IMO when your trying to go real high on OC and if you do use higher LLC modes, your relieving to much on LLC . IMO if you can use a lower LLC and just raise offset/manual voltage you be better off in many ways .
Its less of a jump and also the higher LLC spike way higher than you think (using software to monitor ) , but again when going for highest, you do whatever works .

[chris635]

BeenYeeHua,  Dis-abling C6 state DID NOT dis-able core parking.


edkiefer, I use offset mode for vcore. AS far as LLC goes, I had to set it to ultra high (I have one more setting..Extreme..to much LLC with this one). High is just below ultra high and with this option I get just enough droop to be unstable in certain situations. I cant disable LLC on this board and setting it on AUTO just means it will use the extreme setting. I would have liked to keep a constant voltage range, but with LLC on this board you will either overshoot or undershoot voltage depending on LLC setting.

[edkiefer]

right, sounds like you have Asus MB , auto = first setting in LLC .
What I menat is Intel made LLC to give more voltage for intial load, then back off after that, that is normal .

Now of course if you OC and depending on MB voltage regulation , you might get more droop as you clock higher , my point was basically try high but with more offset , as base Vcore voltage .
I know many, many get good results with Ultra high with Asus latest MB's (At least on Intel side) .

[chris635]

yes it is a asus motherboard and i see what your saying. Try higher vcore with a lower LLC, will get some vdroop though. I am using the crosshair v formula z board. With my current settings it only applies LLC when it is needed just like on the intel side of house.

[chris635]

After more testing with benchmarks such as cinebench, super pi, real bench, unigine valley this is what I have seen. On my overclocks, my system did slightly better with core parking enabled, when my system needed the extra cores for the threads it used it. Now for everyday use, I could not tell the difference between core parking enabled or disabled, except my problem with C6 state which as we know now I got that fixed. So I decided to run AMD's cool'n quite, C1E and C6 states with my overclocks with core parking enabled. Ran my benchmarks again and got the same results as stated above with core parking enabled. Less energy required and less heat generated. Of course someone else may get different results depending on hardware and system setup. In short for my system (once I got the C6 and core parking issue worked out), dis-abling core parking showed negative results on benchmarks (I did have the AMD hotfixes installed for cpu scheduling), could not feel or tell the difference on every day use either way. I can only attribute the results to my high overclocks, CPU at 4.92 ghz, CPU/NB at 2600mhz, HT Link at 2600 mhz and my ram at 2400 mhz. I think maybe with lower overclocks, dis-abling core parking could be more of an advantage.

[BenYeeHua]

Just a question, what windows?

It look like you are using Windows 7 with AMD patch, which is still bad.

Based on my collection.
Windows 7 with AMD patch - Core-parking will park the core inside the module
Windows 8+ - Core-parking will park the module, but not core inside the module, with also 20-40% performance increase that MS and AMD said by switching to Windows 8

So yup, if it is Windows 7 with patch, it might seeing more performance than disable Core-parking, as the overhead of core-parking is taking less CPU time than the wait for shared resource inside the module.

I wonder what it will happen with newer system that support AMD Module more better than Windows 7.

[chris635]

I tested with the patch and with out. The only difference I saw with it was when playing crysis 3. Every now and again (with out the patch) there would be a slight stutter. You have to really pay attention to see it, as I was, (If you were not looking for it you could miss it, it was that small), to check cpu performance since crysis 3 was more cpu bound. On cinebench however, it did worse with the patch. In some areas it helps, in others it doesn't.

[BenYeeHua]

So how about Windows 8?
While you are testing without the patch, I guess you has core-parking enabled?

And yup, as cinebench is using ICC(means it love Intel more than AMD, and disable advanced feature for non-Intel CPU like AVX etc), so it might not a good choose to using for benchmark.

[chris635]

I don't have windows 8. I will upgrade at next windows release.

Yes..core parking enabled without patch.

[BenYeeHua]

hmm...
I wonder what will happen if you disabled it without the patch, I guess it can be the stutter reduced, as Crysis 3 love multi-thread, even some of the threads is low CPU load.
---
A little off-topic, but this is how the newer AMD steamroller works.
http://www.realworldtech.com/steamroller-clocking/

[chris635]

It actually gets slightly worse. I'll just leave core parking enabled. My system runs smoother. I did run my system at stock speeds, and the patch with core parking disabled worked better than with out the patch and cores parked. So I can say for my system on my over clocks with the windows patch and core parking disabled, it doesn't run as good. Somebody might have different results depending on how their system is set up.

[BenYeeHua]

Roger, I guess maybe core-parking helps for I/O bound or the issues that get fixed by Steamroller etc.

So did the Resource Monitor and the Park Control showing it as park?
I still not sure did core-parking works with at least C3 or not.

[chris635]

yeah it showed the cores parked.

[BenYeeHua]

Thank for that info.

So I guess core-parking affecting the performance of Intel, but somehow don't affect AMD bulldozer, and it is still a unknown factor for Windows 8+, which park the whole module for AMD again.
---
Just a little bit off-topic, did you waiting for Windows 8.1 update 2 coming out, which might change the full screen Metro Start Menu back to the classic size Metro Start Menu?

Or just hold on until Windows 9 coming out?

[chris635]

I am waiting on windows 9.

[BenYeeHua]

Roger, I hope Windows 9 will not having too much thing to hate, as Windows 8+ did changing many good thing to the kernel.

[Jeremy Collake]

To my surprise, I discovered that Surface 4 Pros come with a special Windows 10 Edition that has *one* power plan (Balanced) and disallows creation of other power plans. Apparently they had some issues with anything but Balanced and the heat dissipation that disabling core parking might bring. Now, this isn't an issue for desktops or most laptops, but seems to be for at least this rendition of the Surface 4 Pro.

[Corsino]

I  waiting on windows 9.

What about Windows 10?

[Jeremy Collake]

BenYeeHua posted that in early 2014 before Windows 10 was announced.

[edkiefer]

Yes, How time flies