Blur Busters Forums

zudau wrote: ↑

18 Sep 2024, 17:10

Hello,

I am using Windows 10 and tried to do timer resolution tweak (for lower pc latency).

I made 2 benchmarks with Timer Resolution tweak (0.5ms) and default value + CS2 gameplay.

My results:

With this tweak method: su programa.JPG
Without tweak method: be programos.JPG

Because of this tweak lower my latency from 15ms to 1ms (15times lower).
Game feels more responsible , smooth, mouse feels like without “heavy featws”.

Can somebody explain me the results I get?
This is the placebo or real difference?
Any feedback on this topic?

Thanks!

Chief Blur Buster wrote: ↑

18 Sep 2024, 18:32

Timer event granularity is a big rabbit hole, and different motherboards and different OSes (Win10 vs Win11) use different defaults -- but higher timer resolution definitely improves latency in many games, as long as the higher timer resolution is not overloading the computer.

On one hand, high timer-event precision and/or frequencies can thrash a CPU core, slowing framerates down.
On the other hand, 15ms vs 1ms is definitely a human-feelable jitter & latency improvement depending on lag context (see Milliseconds Matters).

In stutter variances, 15ms is like almost four 240Hz refresh cycles, aka a major stutter.
In absolute latency, 15ms is like a 15 pixel lagbehind at 1000 pixels/sec
In latency variance, [0..15ms] is like a [0..15] pixel sudden off-misaim at 1000 pixels/sec
Etc.

Your sawtooth is a proven latency variance and a potential accompanying stutter varance, turning 240Hz into 60Hz-feel. 240Hz monitors can feel worse than 60Hz monitors when given a terrible 15ms timer granularity.

Many people don't need to do this tweak because many computer already ship with 1ms timer resolution by default. But not all do, and that can be a problem oneself needs to solve.

zudau wrote: ↑

19 Sep 2024, 02:09

Chief Blur Buster wrote: ↑

18 Sep 2024, 18:32

Timer event granularity is a big rabbit hole, and different motherboards and different OSes (Win10 vs Win11) use different defaults -- but higher timer resolution definitely improves latency in many games, as long as the higher timer resolution is not overloading the computer.

On one hand, high timer-event precision and/or frequencies can thrash a CPU core, slowing framerates down.
On the other hand, 15ms vs 1ms is definitely a human-feelable jitter & latency improvement depending on lag context (see Milliseconds Matters).

In stutter variances, 15ms is like almost four 240Hz refresh cycles, aka a major stutter.
In absolute latency, 15ms is like a 15 pixel lagbehind at 1000 pixels/sec
In latency variance, [0..15ms] is like a [0..15] pixel sudden off-misaim at 1000 pixels/sec
Etc.

Your sawtooth is a proven latency variance and a potential accompanying stutter varance, turning 240Hz into 60Hz-feel. 240Hz monitors can feel worse than 60Hz monitors when given a terrible 15ms timer granularity.

Many people don't need to do this tweak because many computer already ship with 1ms timer resolution by default. But not all do, and that can be a problem oneself needs to solve.

Thanks for your detailed answer! I fixed my ms at 0.5220ms

Actually, I used program called "MeasureSleep" to track results of ms:
Resolution: 0.5220ms Sleep (1) slept 15.1ms (delta 14.1)

When I turn on DPC Latency checked I got these results:
Resolution: 0.5220ms, Sleep (1) slept 1.04 delta 0.03

With DPC latency checked ON, I feel that this timer resolution tweak is working and responsiveness goes to another level, as my numbers say. Can it be? Do you know more about this program / topic?

Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\kernel]
"GlobalTimerResolutionRequests"=dword:00000001

Slender wrote: ↑

19 Sep 2024, 04:44

nice tool. i was wrong about 0.5033.
0.5034 is better (w10 / w11 same).
but 0.5033 shows 0-1us timer latency in latency moon (kernel test), 0.5034 shows 27-28us.

zudau wrote: ↑

18 Sep 2024, 17:10

Hello,

I am using Windows 10 and tried to do timer resolution tweak (for lower pc latency).

I made 2 benchmarks with Timer Resolution tweak (0.5ms) and default value + CS2 gameplay.

My results:

With this tweak method: su programa.JPG
Without tweak method: be programos.JPG

Because of this tweak lower my latency from 15ms to 1ms (15times lower).
Game feels more responsible , smooth, mouse feels like without “heavy featws”.

Can somebody explain me the results I get?
This is the placebo or real difference?
Any feedback on this topic?

Thanks!