What will a fine PC's latencymon result look like?

[Hyote]

Just gotta make sure to get a low sleep delta, I forgot this previously.

[kyube]

Just gotta make sure to get a low sleep delta, I forgot this previously.

You've ran a program at idle for 30s, to showcase what exactly?
A random number whose value you don't understand so poor souls think you know something they don't?

LatencyMon is not a good tool to test/benchmark/evaluate ANY windows driver performance. It never was, it never will be.

ETW data & WPA (Windows Performance Analyzer) scripts such as xperf or it's alternative, easy-to-digest, front-ends (XTW) are the only proper way to benchmark windows drivers the official way.

Refer to the thread in my signature on what methodology I've employed for my testing of USB controllers & their different drivers for 8kHz mice.

[Hyote]

Just gotta make sure to get a low sleep delta, I forgot this previously.

You've ran a program at idle for 30s, to showcase what exactly?
A random number whose value you don't understand so poor souls think you know something they don't?

LatencyMon is not a good tool to test/benchmark/evaluate ANY windows driver performance. It never was, it never will be.

ETW data & WPA (Windows Performance Analyzer) scripts such as xperf or it's alternative, easy-to-digest, front-ends (XTW) are the only proper way to benchmark windows drivers the official way.

Refer to the thread in my signature on what methodology I've employed for my testing of USB controllers & their different drivers for 8kHz mice.

I am aware that the program is not to be used for actual benchmarking and I'm not trying to flex on others by showing these numbers, it is a continuation of my previous answer where I wrongly stated that kernel timer latency will always be in the hundreds because of the timer resolution.
And to be honest regarding Windows optimization I really do know more than most people because I am not afraid to explore the depths of ideas and not just call it a day when something doesn't seem sensible. That being said I don't like actual bullshit either and want to learn whenever I can but the approach from actual tech guys denying everything because of how they know things are right, is just as bad. I watched a lot of people who were the cornerstones of the tweaking community become attached to the idea of how things should work and throw away everything they found and made everyone else chant the magic words placebo.
I am never going to understand the approach of "sensibility" when you can achieve a lot by sacrificing unnecessary things that people find too hardcore or risky because of how minimal the gains are in their imagination.
Also I ran the tool on idle that is true, but if I open basic programs and move my mouse as fast as I can there is not much difference. But again I just wanted to explain one thing and this was a good way to do so.

[jax0n99]

I mostly used Calypto latency guide for the results(big impact), but I used a lot of different tweaks I learned in the past 5 years from varius sources(some of them were: EverythingTech, Fr33thy, World of PC).

I'll make a list of apps i used to optimize my pc(some may be placebo, but it works in the end*):

winaero tweaker(big impact), iobit uninstaller, ccleaner(first 3 uninstalled after use), quickcpu(run once when pc starts), cpu core parking, parkcontrol, process lasso(i think this is the best app out there), mem reduct, rtss, msi afterburner, islc, timer resolution(last 7 are running in the background at all times), morepowertool, DDU, moreclocktool(those last 3 for gpu tweaks), interrupt affinity policy tool(very important when balancing cpu usage on the cores(at affinities, select everything except core 0 on usb, gpu, network and sound)), msi util v2, autoruns(disable useless background tasks).

It takes me about an hour to do all the tweaks(got used to it after having to do it so often).

I can get most computers/laptops to run in under 2.5us average interrupt to process latency and under 1us average interrupt to dpc latency using the above apps.

I am willing to learn about more tweaks to get my pc to have an even lower interrupt latency.

Hi, I registered on this forum to message you, but it won't let me DM you because my account is new.
Would you be willing to fix the Latency on my system if I paid you?


I am an older guy and not so techy.

[jc008titan]

Just gotta make sure to get a low sleep delta, I forgot this previously.

You've ran a program at idle for 30s, to showcase what exactly?
A random number whose value you don't understand so poor souls think you know something they don't?

LatencyMon is not a good tool to test/benchmark/evaluate ANY windows driver performance. It never was, it never will be.

ETW data & WPA (Windows Performance Analyzer) scripts such as xperf or it's alternative, easy-to-digest, front-ends (XTW) are the only proper way to benchmark windows drivers the official way.

Refer to the thread in my signature on what methodology I've employed for my testing of USB controllers & their different drivers for 8kHz mice.

The results people are looking for only show really terrible latency issues. |
What you actually want to look at is the Average measured interrupt to DPC latency (µs) from Stats page. That's the only thing that you can actually lower by having better drivers and better overall windows optimisation, and that result actually translates to how your computer feels and acts like.

I get about 0.465us as an average right now, and I made sure my optimised stock windows 10 is still able to run very intrusive anticheats like FaceIT or Vanguard properly, unlike people who overly optimise their windows in a way that completely breaks any compatibility with them. It is also insanely hard to find people who have both heavily optimised computers and also play with invasive kernel anticheats.

In my opinion, if you don't want kernel anticheats on your computer and you want low latency, just switch to linux as a much easier option.

I personally use latencymon as a spyware and malware checker. Having less than 50 Processes running in background and 41 services allows me to track each one in task manager and services any time i notice some big delays or weird windows behaviour.

One other thing that helps latency is balancing cpu usage between all cores. Straining one single core so much that it throttles subsequently makes the other cores also underperform.

Power plans and disabling idle options are the most impactful to latency, prioritise looking for those even before disabling random useless services and startup apps.

As for periferices you wanna prioritise, you gotta go with higher refresh rate monitors(144hz+) and a better mouse, as those are the most impactful on latency of all. I've upgraded from Deathadder v3 to op1 8k v2 and got a 30ms overall improvement for my reaction time.

[daviddave1]

I can get close results on windows 11, but here I am using a debloated version of Windows 10.

I have a 12700K with an all-core OC at 5.2ghz (disabled ecores). Constant clock, which is important for DCP latency. I manually configured my drivers to certain cores to get my results this low. I also changed my timer resolution in windows to .500ms.

Why not show the part that shows the average latency like asked before? Or share what motherboard and ram you you use? What cooling? What kind of debloated version of win 10? Homecooked?
What debloat made the biggest difference?

People would kill for your stats if its near average latency that is. Me personally I think your photo is photoshopped or has some querk to it. Like unplayable or something. Correct me if i am wrong please.