ParkControl (aka CPUPark or CPUParkControl) - Feedback

[Jeremy Collake]

Please provide any feedback for the alpha release of this little freeware tool. Control of CPU parking will be integrated in version 6, so any feedback is appreciated. I know some concerns already, which I'll enumerate so you don't have to .

1. There is no detection of whether the hardware and OS support CPU Parking to start with
2. The interface is not very polished

http://bitsum.com/about_cpu_core_parking.php

[gman68w]

How does this differ from similar functions found in the BIOS setup menus in 990FX boards like the ASUS Sabertooth (HPC Mode, Core C6 State)?

[Jeremy Collake]

How does this differ from similar functions found in the BIOS setup menus in 990FX boards like the ASUS Sabertooth (HPC Mode, Core C6 State)?

ParkControl

1. Allows for change of parking behavior based on the active Power Profile.
2. Allows for change of parking behavior in real-time, no reboot required.

Of course, if you disable CPU parking in the BIOS, it can NOT be enabled by this tool.

[DeadHead]

Way back, I disabled core parking through registry change. Using ParkControl to try and enable it again does not seem to work, I get no 'P' on every second core like it used to read. Any hints on why it won't work? Intel i7 920 processor. After applying changes and clicking ok, then go back to check, the settings are not saved for me.

Edit: Will have to check if I did something in Bios that I forgot about, will post again later.

[Jeremy Collake]

That's why ParkControl is preferable to registry changes . I wanted to release the tool just for that reason. That *said*, it should work fine to revert those changes you made, so long as your BIOS settings are OK. Do also be aware of what power profile you are in, of course. It is also possible if you changed some registry setting that shouldn't have been touched, perhaps it somehow got messed up.

ParkControl does *not* make direct registry edits, just FWIW.

So, I dunno. I am able to turn the on/off fine here, but I have not done manual edits to the registry in the past... though the *same* values do end up being changed by the OS when the ParkControl changes are applied.

[DeadHead]

I examined Bios, but nothing related to this had been changed as far as I could tell. I'll have to search the web in order to see if the change I made originally can be reverted, not that it matter much, most out of curiosity. It was so long since I made this change now. Would be interesting to see the values of an untouched Windows 7 system, unless they're different depending on installed processor.

[Jeremy Collake]

I just updated ParkControl, but it was minor stuff, so doubt it is applicable to your situation.

[Jeremy Collake]

As for values on a fresh Windows 7 install, that is what I created RegMerge for (well, one of many uses). If you install it in a virtual machine, you can Export the registry key(s) in RegEdit, then compare them to your PC. Many OEMs would have customized these somewhat in their creation of custom power profiles, but otherwise they should be the same. If the CPU doesn't support parking, it just wouldn't be used.

That UUID they tell everyone to search for is simply the identifier for CPU parking in the power profiles. That is why it is replicated X times.

[DeadHead]

Found the changes I had made and reverted them. Now using core parking settings in v 6 alpha works just fine, so I enabled core parking for the energy saving profile, and disabled core parking for high performance profile - nice!

[Jeremy Collake]

Great .

I will have to consider cases where people hacked up their registry maybe, though it is a cumbersome thing to do - find all the 'recommendations' and make sure they are fixed properly. If you DELETED those values then that may explain this, come to think of it ...

[DeadHead]

Would it be possible to enable even more parked cores on an i7 with this tool, or is this a limitation of windows/the cpu? For energy saving mode, it would be interesting to try and leave just a single running core, and see what that would do to the heat of the processor. I tried setting a value of 25%, thinking that would do it, but so far it don't seem to work?

[Jeremy Collake]

I actually have hidden options not yet shown in v6 integrating parking stuff into energy saver, more on that later. I didn't want to reveal it, I keep a *few* secrets . HOWEVER, I don't think it would decrease heat all that much, but perhaps a little. If your cores are IDLE they aren't emitting much heat anyway. Core Parking was originally more about power utilization than heat dissipation.

Generally Windows uses maximum core parking by default, so you can only decrease how many cores it is allowed to park. That said, there is a separate setting I have not spoke much about, because I don't know much about it. I discovered it while mucking around, and it is possible that using it could further increase the aggressiveness of parking (not how many are allowed to be parked, but how often). That said, it is completely undocumented, so I am still investigating if it is even implemented and has any use at all to the user.

[Jeremy Collake]

And don't forget Frequency Scaling, as you can modify in the Power Options of Windows for specific power profiles makes the *biggest* difference in heat dissipation while idle. Setting it to allow the frequency to be scaled down to its minimum value will make a huge difference, if it is not already set as such.

[gman68w]

Is there some kind of 'Submit System Info' feature attached?

[Jeremy Collake]

Is there some kind of 'Submit System Info' feature attached?

Attached to what? I don't get the question, maybe I haven't woke up yet, I'm sorry. Please elaborate.

[gman68w]

Attached to ParkControl. I mean, does it collect info in the background or is there a menu option somewhere?

[Jeremy Collake]

Attached to ParkControl. I mean, does it collect info in the background or is there a menu option somewhere?

No, there is neither. It doesn't do anything with regards to actually checking the system to even see if CPU Core Parking is supported by the hardware at this time. That is something to be added, though it was written with the assumption that most would be using it to disable CPU Core Parking, not try to enable it on systems that don't support it. I suppose I should add some detection as to whether this is a supported feature, though that may be easier said than done. Indeed, this feature was kinda slipped in so silently that I'm not sure what generation of processors first started supporting it, though there is surely some way to determine if it is supported. Maybe checking the processor feature set.

[edkiefer]

According to this link it started with Nehalem/Opteron CPU’s.

http://blogs.technet.com/b/klince/archive/2010/09/14/tweaking-cpu-core-parking.aspx

For Intel that's core i3, i5 and i7 cpu's

[Jeremy Collake]

Thanks Ed . I'm not sure what mechanism I'll use, but I will definitely add this at some point - it seems an obvious necessity in hind sight. Whether or not this was what the original poster was even talking about, I don't know ;p.

[DeadHead]

Just noticed something quite odd. When my computer enters energy saving mode, I have enabled core parking in that power profile. When core parking gets enabled, it parks cores 1, 3, 5 & 6, leaving core 0, 2, 4 and 7 running!

[Jeremy Collake]

Just noticed something quite odd. When my computer enters energy saving mode, I have enabled core parking in that power profile. When core parking gets enabled, it parks cores 1, 3, 5 & 6, leaving core 0, 2, 4 and 7 running!

Sounds typical to be honest. The CPU parking tries to go for every other CPU, and as for that 'last' one, which is off pattern, probably due to activity on it.

[DeadHead]

Ok, I have noticed that there's no real consistency with how windows parks my cores at all, lol! I have seen other patterns in a random mix. Editing the registry as I did might have caused this, since as far as I can remember, it used to be only the hyperthreaded cores that got parked? Feels rather odd!

[Jeremy Collake]

Ok, I have noticed that there's no real consistency with how windows parks my cores at all, lol! I have seen other patterns in a random mix. Editing the registry as I did might have caused this, since as far as I can remember, it used to be only the hyperthreaded cores that got parked? Feels rather odd!

It is by design. You see, the scheduler assigns new or context swapping threads pseudo-randomly to 'free' cores. Anyway, so the load(s) are likely to be on a random core. That means the scheduler bases its parking decision on what cores are active, which is not fully random, but practically random.

And it is NOT only HyperThreaded cores that get parked, BUT they have a MUCH HIGHER likelihood of being parked. So, the slightest change to your system could change this change in behavior.

[Jeremy Collake]

Thought this chart might be of interest to people.

[BenYeeHua]

Thought this chart might be of interest to people. I found it at http://www.overclock.net/t/1180334/xtremehardware-it-core-parking-on-windows-7/50 . The original source appears to be http://xtremehardware.it

Yup, as Google record that image.
http://www.xtremesystems.org/forums/showthread.php?277129-Core-parking-on-Windows-Seven/page4

Correction: They are the same person  Just that person post on there first.

But I don't know where he get and collect that data.

[gman68w]

Thanks Ed . I'm not sure what mechanism I'll use, but I will definitely add this at some point - it seems an obvious necessity in hind sight. Whether or not this was what the original poster was even talking about, I don't know ;p.

I was thinking of the collection of system info to compile some kind of list of supported (and non-supported) CPUs.

[Jeremy Collake]

I was thinking of the collection of system info to compile some kind of list of supported (and non-supported) CPUs.

Oh, that should be available somewhere, or achievable in an easier way than collecting user stats. As for runtime detection, I discovered, I think, that I will be experimenting with soon.