Another look at HPET High Precision Event Timer

Thanks for that, it's working now. WinTimer puts it at 14.32Mhz.

I just ran a Metro 2033 benchmark using it's tool. Seems to be the exact same performance. I did benchmarks for a couple of my friends 2-3 days ago, so I have the results for accurate comparisons.

Minimum and average FPS are within the margin of error. I'll play BF3 MP and post back.

What realistic difference should I be looking for? I don't care about synthetic benchmark improvements.

 

^
It should be Smoother with HPET - on, i didnt have any fps increase with ON or OFF though.. Maybe it depends from mobo manufacturers and how they implement this HPET in bios.

Some have better results with it on (gigabyte), some worse (asus). At least that's what I saw.

 

I had the exact opposite. I have a MSI GD65 Z68.

This is what I can say so far.

I played BF3 3 (2 with HPET) 64 player maps, all B2K as they seem to be more prone to stuttering.

With HPET on in BIOS and Windows, at both maps there was huge stuttering as soon as I spawned. It went away in abhout 5 seconds, as if the processor was idling and then ramped up to meet the demand (just an example of how it felt, don't know if that is what happened)

After it was "smooth" there was still stuttering but tolerable. I used a DPC checker in Windows desktop, the time ranged from 80-150ms when idling.

Then I turned HPET off from BIOS and Windows. Jumped into a B2K server and there was no spawn stuttering. In fact, as I played more and more, it was easily to tell that stuttering was reduced. Finished the map smooth and then checked latency in Windows. 10-90ms.

It should be noted that the FPS was more or less the same, but with HPET off it was definitely smoother for me. I'll play around some more and post obvious differences if I can find any. But I'm pretty sure I'm not going back as it was originally (HPET BIOS on, but Windows off).

 

strange but I did:

bcdedit /set disabledynamictick yes

then

bcdedit /set useplatformclock true

checked it with WinTimerTester_1.1 and the clock was 14.31 and my guild wars 2 performance in windows 8 64 bit went from the sluggish 30-40 fps to 70-80 fps.

 

It turns out my OC wasn't stable. I got it stable @4Ghz and updated my audio drivers and it settled down but my DPC latency still jumped high.
I couldn't figure out why. So I updated to 306.23, ran it again and then I looked at the stats. It only ever jumped that high once or twice.

Spoiler

http://i.minus.com/iblq00GT2taUeZ.png
http://i.minus.com/iboa19gydGRA8Y.png
http://i.minus.com/i8NitWIko1mQv.png
http://i.minus.com/iYTDpGJW2edRN.png
http://i.minus.com/icyRE5QOdGyre.png
http://i.minus.com/ijHT5p1a3oTen.png
http://i.minus.com/ibm1Bq6JjuUT3X.png

Then it hit me. "Wait a minute, i'm running this benchmark at 1280x720 full screen...what if when it's adjusting resolution(Monitor res is 1600x900) is when it happens?"

So I started SFIV, Alt-tabbed out, started latency mon. Alt-tabbed back in and low and behold it spiked! I thought "Ok, let's run this 720p windowed and see what LatencyMon reports this time.."

Wahlah!

Spoiler

Nothing spiked other than the Hard Fault.
Once the bench finished and I took screen grab with Afterburner of the game. I took the screen of LatencyMon, immediately after I took the shot one of the other meters spiked(I can't remember which one).

I thought it was just because of the screen grab, so I tried to replicate it by screen grabbing again. Nothing. I ran the benchmark with a fresh start on LM again, did the same. And it didn't happen again.



Should I still try turning off Core Parking? What are the pros/cons of it?
I read something along the lines that it could cause problems with certain programs that don't play along with Hyper Threading if it's turned off.

My CPU is HT'd

 

So now it says your system is capable of handling real time audio too?

Interesting findings about the rest, i always ran SF4 in fullscreen, but the funny thing is sometimes I saw deviations up to 6000 points (one time ~ 21000 then if i changed few settings up to 26500 at 1280x1024, no AA, for cpu testing), i kinda pinpointed it down to advanced memory timings/cpu OC and in driver threaded optimization - ON or OFF (depends on driver). I never tried in windowed screen though..


About Coreparking off, it won't hurt anything, just make a separate power plan profile and activate core parking on it, you can always switch back to default high performance profile..

What it does is unparks parked cores (HT too) and eliminates any delay when OS tries to resume them.

 

Yes, try it. At least that cured my BF3 frame stuttering issue immediately (that or disabling HT in the BIOS, which was no longterm solution for me). BF3 runs very smooth now even though the DPC latencies are not that low on my system as on yours (got HPET enabled in BIOS and Windows).

 

Core Parking. Good or bad?:
http://www.overclock.net/t/1240512/c...ng-good-or-bad

Disable CPU Core Parking Utility:
http://www.coderbag.com/Programming-...arking-Utility

 

I have better gaming experience and less stuttering with Windows 8 and HPet only enabled.
But i cannot try disabled Hpet From bios cause i've it forced on my Asus P5kc.

Remember that win8 first releases had it enabled by default, but it crashed some devices so they decided to disable it as in win7!

 

I disabled HPET in BIOS and Windows now after about 2 weeks of using. And i can't rly feel or notice any difference apart from 3D Mark 11 being 200 points higher with it Off.

 

My LatencyMon Tests with HPET On and Off.

Spoiler

A lot less pagefaults with HPET On.

The lowerst latency with HPET Off never came below 800, while with HPET On it goes as low at 45.

 

okay thats odd i tried it again early this morning today by enabling in on windows command "bcdedit /set useplatformclock true" and i forgot last time i didnt have hpet enabled in the bios now i have both enabled bios and windows and i can feel windows run snappier and games seems it running smoother no stuttering at all but i have no idea how much frames i gained on games im keeping it on for now.

 

did you also use the windows command "bcdedit /set useplatformclock true" ?

 

Yep, else it was still off when checking with the Timer tool.

 

Can have a go at that, i can select 32-Bit or 64-Bit. I ran the test on 64-Bit.