Another look at HPET High Precision Event Timer

Hi from tomorrow I will start testing if msi drivers make a difference I am curious too.
By the way is there a way to see before installing if a driver is msi mode compatible ?

 

Only by searching inf-files for "MSISupported" values. If it is set to "1" (or "0") then MSI mode is "forced" on (or off). But if there is no "MSISupported" it does not mean driver is compatible or not - it can be both.

 

Ok Thank you.

 

https://forums.guru3d.com/threads/a-bit-detailed-info-on-intel-time-stamp-counter-tsc.433977/

 

Yes... I am here, once again. To say something about this. I recently did several benchmarks, clear of mind, finally free of latency groups, finally free of outside influences. To see for myself if my previous assumptions on this issue were true.

My conclusion was...

drum roll....


...

disabledynamictick reduces performance.
useplatformclock reduces performance, even with hpet enabled in bios.
useplatformtick makes no difference, hpet on or off.

Thus, Martigen's post is absolutely the correct answer, and none of these values should even be set to 'off'. Simply delete them all. There are very rare cases where hpet on and the platformclock might be relevant.

HPET off or on in bios is also irrelevant. I have left it off for peace of mind.

I tossed my benchmark results because they were insultingly inconclusive.

Felt like I needed closure so there ya go. Want to add my 0.00001 bitcoin into the mix of anyone who stumbles on this confusing as frack thread.

 

newest windows 10 installation.

game : valorant - training map / avarage fps 260

after i set useplatformclock no -avg fps => 350.

no. windows does not detect "the best one for you". everyone should try and find the best one for their hardware. it just takes a couple mins to do so anyway.

 

Ancient hardware?

PS
Also Windows detects not best for "you" but best for itself. Only reason to not select TSC as an implementation for QueryPerformanceCounter is a TSC vulnerable to C-states or P-states. If Windows failed to detect so called Invariant TSC it selects other counter/timer to avoid problems with "floating" TSC.

PPS But I completely agree with your last sentences "everyone should try and find the best one for their hardware. it just takes a couple mins to do so anyway."

 

HI, ¿how i can check the values i have for bcdedit /deletevalue disabledynamictick, bcdedit /deletevalue useplatformclock, bcdedit /deletevalue useplatformtickin in w10?

 

It is not clear what exactly are you asking - how to check whether values were deleted (from BCD) or how to check what has changed in system after the deletion.

 

Hi, thanks for the answer, i would like to to know wich values i have before doing changes.

 

BCD values "disabledynamictick", "useplatformclock", "useplatformtick" can be in 3 states:
- absent, not set (default state);
- "yes";
- "no".

To check them you have to execute command "bcdedit" in elevated command prompt. If you do not see a value it is absent (is not set).

 

Thanks again for answer me, one last question, ¿where or how i can compare perfomance?

 

Games, benchmarks.

 

Thanks.

 

A bit off topic but my Ryzen 2500U based notebook is sluggish (especially FireFox) after waking from suspend. I think this started with 20H1 (I also have the 20H2 ennoblement package but I think it started before that). I wonder if you 'tweakers' might have some ideas about that.

 

Why don`t create dedicated thread?

 

hello, i have managed to see different results with 0.5 and 0.4991 ms. Mouse polling rate is more consistent when timer resolution is set to 0.5. As you can see, witn 0.5 tr dots lie mostly between 1020 and 980; while with 0.4991 tr dots lie between 1040 and 980 units of mouse polling rate

 

And? This is such tiny difference.

 

Well, while i am sure i feel the difference in 1 game, i am not sure that this is not a placebo. But atleast there is proof that timer resolution tweaks are not totally useless.