36 replies to this topic

[koadic]

Hi everyone, I have come across something that may help some of you get a little more performance and reduce stutter. First things first, this will only help machines running Windows 7, as this allows you to disable an 'optimization' added in Windows 7 (LINK). What this registry tweak will do is expose the ability to allow you to disable core parking in your power settings, nothing else. It doesn't change any settings (you have to do that on your own), it just allows you to see the settings for you to change them. Core parking is what the OS uses to disable cores when the load is low allowing for better power savings and dynamically unparking them when needed. Google core parking if you want to read some of the discussions about it. It has been a subject of some debate on whether or not it helps, but I sent the registry entry to OldSkoolGamer to try out and it has made a noticeable difference on his Windows 7 cab, but then again results may vary. It has also been argued that this will only help systems with more than 2 cores.

I will link a zip file with a .reg file you can double click to apply below, but here is what needs to be changed:

Open RegEdit, navigate to HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Control\Power\PowerSettings\54533251-82be-4824-96c1-47b60b740d00\0cc5b647-c1df-4637-891a-dec35c318583. From there, you should see an entry named Attributes with a value of 1... change it to 0 to 'unhide' this setting from your power settings (again, will attach a .reg file below that will do this for you). Again, this change alone only makes the setting visible, nothing else.

After making this change, go to the Control Panel to Edit Power Plan, and click on Change advanced power settings, scroll down to the processor power management, and expand the one titled 'Processor performance core parking min cores' and make sure each is set at 100%, as this sets the minimum amount of cores you want to remain unparked (in effect turning off core parking). If you want to try this on a laptop, you can also leave the On battery setting alone so it only effects when it is plugged in. Or, you can create a completely custom power profile and leave the default ones alone.

Here is a zip file with the .reg file inside:
 UnPark.zip   396bytes   84 downloads

Hopefully this will help those having some issues even more gain a little more performance out of their machine.

Edited by koadic, 15 October 2012 - 03:31 AM.

[oldskoolgamer]

Just my two cents on this - I tested this reg entry on my desktop PC and my 3 screen VP cab system. It actually cleared up alot of ball stutter on multiple tables on both systems. Very very happy with the results....can actually even play Monopoly on my 3 screen cab without any ball stutter now! Thank you Koadic for this reg man!!

[htamas]

Sounds like this is only applicable to Win7/Vista and not for WinXP, correct?

[oldskoolgamer]

Sounds like this is only applicable to Win7/Vista and not for WinXP, correct?

Correct.

[rob046]

This doesn't seem to help me at all, or if it does its extremely minimal. Win 7 64 bit. Though I usually don't have an issue with performance. I've got a i7-2600k @ 4.3ghz, I believe. I believe that is 8 cores, 4 physical? I'm only on a single screen desktop setup where I run 1920x1200 or 1600x1200.

Also I was a little worried when I tried this because I wondered if it would mess with my overclock or cpu stepping. I like where I have things because I have a pretty solid overclock despite having all my power/heat saving features still enabled. I'm curious if anybody else on a desktop setup is getting any improvement. If so, maybe I'll try this again on some more tables.
But as of now it looks like it might only help more for those running 2 or 3 screens.

[FDSystems]

Is it reversible ?

[rob046]

Is it reversible ?

It is a setting that will appear in your power management, so you can just set it back to where it was. No need to reverse the registry hack itself, though you could if you wanted to.

[koadic]

Yes this adjustment is completely reversible, and depending on your setup you may or may not see any improvement (such as rob046). With a correctly set up system including os tweaks and bios settings, this may not be needed at all. A quick way to determine if this might help is to open the Resource Monitor, go to the CPU tab, and see if any of the cpus on the right by the graphs are listed as parked when the machine is idle.

Example:

[koadic]

Just bumping this as I am curious if any of the 23 downloads resulted in any sort of improvement or if everyone who has tried it (other than OSG) had similar results to rob046.

[The Loafer]

I downloaded it but have yet to try it,busy working on installing UVP for lots of tables. But it's next on my agenda and since I use a i7 quad core, but at its default 2.85 ghz, I suspect and hope I will see a difference. I will report back for sure.

[jimmyfingers]

Hey Koadic, I tried it on both my development and mini-cab systems and did not notice any change. However, my mini-cab is very tuned and I have spent many many hours tweaking and setting the affinity of all services, the VP executable(s), UVP, and my FE (3darcade). So, no, I haven't seen any real change In the few relatively quick tests I've done with the couple tables I often use to test and normally refer to the FPS for my quantitative analysis, followed by some qualitative visual inspection for fluidity or stuttering moments.

Like you say, on highly tuned systems it may not do much. My systems by the way are a single processor 2 core only for my development and a 4 core i5 for my mini-cab which is over-clocked quite high (2.6 to 3.6) - of course both are Windows 7.

Thanks for any suggestion and potential performance improvement you can find or offer - it is appreciated. I still think your alpha 0 and 1 finding was a game changer upon which my image based routine would not have been developed had you not found and reported on your findings to start.

[Arcade4]

I have an AMD 955 quad core running on Windows 7.
I will try and see what happens on my system soon.

[koadic]

Hey Jimmy, I think I read that if you have the affinity set for programs to use separate cores, that overrides Windows 7's core parking, so since you have separate programs forced to run on separate cores, disabling core parking has essentially already been done, and this little tweak won't really have any effect.

[ArcadiusMaximus]

Wow nice will definitely give this a shot and will report back. Hopefully it will clear up my UVP stuttering issues. FINGAS CRIZOSSED !

[The Loafer]

I completely forgot about trying this, will do so tonight and report back.

[TheMcD]

I think this might have improved something. Then again, I also think that my cat intentionally vomited right in front of my door, so who knows if it actually did something. I mean, I don't even know if it's my CPU or my GPU bottlenecking VP.

It certainly didn't hurt though, and I suppose it might help where my CPU is definitely the problem, so there's that. Cheers.

[Flying Dutchman]

Tried it and it did not improve significantly => I have only one Table that has a slight impact on performance 'Monopoly'.
I did see a small improvement running the Table without HP .. with HP again the same as before ..

Thanks!

Cheers,
FD.

[ArcadiusMaximus]

Hmm I have a question. Is this setting smart enough to move demanding programs to its own core ie VP, UVP, and Hyperpin seperate or do you still need to use set affinity or does it even matter? I never had any luck with set affinity, just curious. Should one or the other be used not both in conjunction? Thanks.

[koadic]

If you happen to use set affinity, then any programs set to run on specific cores will automatically unpark them, which may have been the real reason it helped (pure conjecture), but this doesn't do anything different in setting programs to use different cores.

[The Loafer]

Ok, took me longer to test, sorry, but I finally did tonight, IMHO, I am happy I did too as it improved some of the minimal stutter I experience and was a massive improvement on the b2s version of wrath of Olympus. Now the only way to be sure is to experiment more with this by turning cores back to their default setting, but no time tonight, maybe when I get back from visiting my mom.

Koadic, question for you: in your text above where you describe how to change the setting of the cores you stated "and make sure each is set at 100%". I have a quad core i7, but I only have 1 setting to change (from 10% to 100%). Should I be seeing different cores listed with the ability to provide a different setting for each?

Thanks again. Someday I will attach setting each main program to its own core but this does the job for now. Appreciated

To all: aside from t2 chrome, any other tables you experiment we test with?

Edited by The Loafer, 06 November 2012 - 04:08 AM.