Windows Server has less CPU lag [Lower DPC Latency]

Everything about latency. Tips, testing methods, mouse lag, display lag, game engine lag, network lag, whole input lag chain, VSYNC OFF vs VSYNC ON, and more! Input Lag Articles on Blur Busters.
DPRTMELR
Posts: 165
Joined: 12 Apr 2022, 13:42

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by DPRTMELR » 28 Jul 2022, 11:50

olofsson_tom wrote:
28 Jul 2022, 10:37

Yes, this did indeed improve everything a little bit overall. Thanks! If anyone has any more tips you are more than welcome to share :)
I really don't think there's much else to do on server (all the telemetries/unnecessary services already being off) other then maybe try different version drivers when you have time. I haven't even disabled hpet yet because I am too lazy to google where msvcr100.dll comes from and install it. Like this thing doesn't even need you to undo meltdown/spectre software patches lol
EnablePreemption/CoalescingTimerInterval
First one seems like its more related to vm/rdp heavy environment and second one is some timer that i don't understand so its scary D: any noticeable effects?
olofsson_tom wrote:
28 Jul 2022, 10:42
1000WATT wrote:
28 Jul 2022, 08:52

Idk maybe. Feels like with the way his channel is heading is that he wants easy to use pain-free guides that fit everyone and maybe messing with timers could help some and hurt some others. Besides, he seems to share all his tweaks for free so you dont need to buy his services, unless you want to spare time or if you just really dont know much about anything besides launching your game.
S E L L O U T lmao
I mean he's gotta get some return for spending all those times tweaking windows (and tweaking out in IRL probably), good for him. (according to you guys I don't have him subscribed, no clue what his positions are) At least he is maintaining he is not focusing on such minor details, instead of saying they don't make any differences.
Last edited by DPRTMELR on 28 Jul 2022, 12:08, edited 1 time in total.
Most adults need 7-8 hours of sleep each night. - US FDA

olofsson_tom
Posts: 26
Joined: 11 Sep 2021, 18:55

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by olofsson_tom » 28 Jul 2022, 12:07

DPRTMELR wrote:
28 Jul 2022, 11:50
olofsson_tom wrote:
28 Jul 2022, 10:37

Yes, this did indeed improve everything a little bit overall. Thanks! If anyone has any more tips you are more than welcome to share :)
I really don't think there's much else to do on server (all the telemetries/unnecessary services already being off) other then maybe try different version drivers when you have time. I haven't even disabled hpet yet because I am too lazy to google where msvcr100.dll comes from and install it. Like this thing doesn't even need you to undo meltdown/spectre software patches lol
EnablePreemption/CoalescingTimerInterval
First one seems like its more related to vm/rdp heavy environment and second one is some timer that i don't understand so its scary D: any noticeable effects?
Yeah, seems really good out of the box. Can recommend. Only thing that was annoying to do was install the ethernet adapter driver, but the guide in this thread worked to fix it.
Not sure, havent really tested between every single tweak. But i got the tweak tips from Adamx on youtube, and his recommendations seems to work good in my experience. He also made .reg files to disable defender, quite handy since it gets activated back sometimes with updates.

DPRTMELR
Posts: 165
Joined: 12 Apr 2022, 13:42

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by DPRTMELR » 28 Jul 2022, 12:25

olofsson_tom wrote:
28 Jul 2022, 12:07
So what is preemption then? In a nutshell, it’s the ability to interrupt an active task (context switch) on a processor and replace it with another task, with the further ability to later resume where you left off. Historically this is a concept that’s more important for CPUs than GPUs, especially back in the age of single core CPUs, as preemption was part of how single core CPUs managed to multitask in a responsive manner. GPUs, for their part, have supported context switching and basic preemption as well for quite some time, however until the last few years it has not been a priority, as GPUs are meant to maximize throughput in part by rarely switching tasks.
Timer coalescing is a computer system energy-saving technique that reduces central processing unit (CPU) power consumption by reducing the precision of software timers to allow the synchronization of process wake-ups, minimizing the number of times the CPU is forced to perform the relatively power-costly operation of entering and exiting idle states.
I am only gonna try premption since it will be immediately noticeable if something goes wrong, but not mess with timer coalescing and just trust the bios/windows power settings for now. you also reminded me to disable gpu hardware scheduling.
Most adults need 7-8 hours of sleep each night. - US FDA

olofsson_tom
Posts: 26
Joined: 11 Sep 2021, 18:55

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by olofsson_tom » 28 Jul 2022, 13:00

DPRTMELR wrote:
28 Jul 2022, 12:25
olofsson_tom wrote:
28 Jul 2022, 12:07
So what is preemption then? In a nutshell, it’s the ability to interrupt an active task (context switch) on a processor and replace it with another task, with the further ability to later resume where you left off. Historically this is a concept that’s more important for CPUs than GPUs, especially back in the age of single core CPUs, as preemption was part of how single core CPUs managed to multitask in a responsive manner. GPUs, for their part, have supported context switching and basic preemption as well for quite some time, however until the last few years it has not been a priority, as GPUs are meant to maximize throughput in part by rarely switching tasks.
Timer coalescing is a computer system energy-saving technique that reduces central processing unit (CPU) power consumption by reducing the precision of software timers to allow the synchronization of process wake-ups, minimizing the number of times the CPU is forced to perform the relatively power-costly operation of entering and exiting idle states.
I am only gonna try premption since it will be immediately noticeable if something goes wrong, but not mess with timer coalescing and just trust the bios/windows power settings for now. you also reminded me to disable gpu hardware scheduling.
You mean HAGS? isnt the reccomendation to use it unless you are streaming?

DPRTMELR
Posts: 165
Joined: 12 Apr 2022, 13:42

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by DPRTMELR » 28 Jul 2022, 13:17

olofsson_tom wrote:
28 Jul 2022, 13:00

You mean HAGS? isnt the reccomendation to use it unless you are streaming?
I know it's better disabled with 2 gpus but now with only 1 gpu I am not too sure lol.
Benchmarks from back when this was first introducted (2020?) shows they are within margin of error of each other in terms of raw fps, probably not too much difference.
Most adults need 7-8 hours of sleep each night. - US FDA

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by Chief Blur Buster » 28 Jul 2022, 22:01

1000WATT wrote:
27 Jul 2022, 13:20
vrr+mbr (benq)144hz - vrr+ulmb (nvidia)165hz
Important variable: Existence of VRR
If gametime is in perfect sync with photontime, there's no jitter, even if frametime varies, because of the magic of VRR synchronously jittering the photontime with the gametime, which is how VRR is magically jitter-free when it works.

Now that's different from flicker behaviors. VRR strobing is very hard to be engineered to be comfortable, which is why things like ELMB-Sync automatically goes into a non-strobe mode at various moments.

Picture it this way, 10 microseconds can make strobed VRR flicker. Strobing at 1ms MPRT, versus 1.01ms MPRT is a 1% brightness difference, which can be human-visible -- Since it's a 1% difference in the number of photons per refresh cycle hitting the human eye balls.

It's one of the examples within The Amazing Human Visible Feats of the Millisecond
1000WATT wrote:
27 Jul 2022, 13:20
This was an excellent real-time indicator of how unstable the system was. It took about a week before I gave up.
Strobed VRR on an unstable strobed VRR system (e.g. early implementations of ULMB+VRR) is an excellent indicator of system instability since this is where microseconds become human visible, as explained above.

The virtue of concentrating strobe backlight pulses in a ultrashort time period per refresh cycle (e.g. 1ms or less) means any percentage change to 1ms now is amplified to be human visible, by virtue as a percentage change in brightness!

So it's a jawdrop how 10 microseconds (0.01ms) becomes human visible in a strobed VRR defect!
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

DPRTMELR
Posts: 165
Joined: 12 Apr 2022, 13:42

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by DPRTMELR » 29 Jul 2022, 23:23

So it turns out I had my power options set to balanced, even though I swear it was set to bitsum maximum oops D:

Still no actual tweaks (enablepreemption is disabled but not sure if it does anything and it's gpu related), I just figured something is weird with such low averages but random high spikes.

This thing be cray. god damn. i should really go install 0.5 timer

-edit
i deleted enablepreemption 0 registry. if ibm knows about it, if microsoft knows about it, and nvidia knows about it.. on top of the fact that I was relatively satisfied without it and I can't tell if this is doing anything anyway. better be safe and not have to deal with formatting later because I forgot it was this change D:
Untitled.jpg
Untitled.jpg (138.96 KiB) Viewed 3875 times
Untitled.png
Untitled.png (75.57 KiB) Viewed 3859 times
Most adults need 7-8 hours of sleep each night. - US FDA

DPRTMELR
Posts: 165
Joined: 12 Apr 2022, 13:42

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by DPRTMELR » 30 Jul 2022, 04:05

You can just use the results and compare them relative to each other. ie move the start line up for everyone, because latencymon sucks. Like the actual musicians really use this as a measurment(not sure how hollywood people does it though) also im 99% sure you are thinking of dpc latency checker when you say the owners admitted inaccuracy after windows 8
Untitled.png
Untitled.png (20.5 KiB) Viewed 3785 times
idunno microsoft thinks my hardwares are sweet
Most adults need 7-8 hours of sleep each night. - US FDA

1000WATT
Posts: 391
Joined: 22 Jul 2018, 05:44

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by 1000WATT » 30 Jul 2022, 08:10

thizito wrote: Its pathetic that u guys and even chief
are fooled by latencymon(the owner itself admited that it cant measure shit correctly)
and not only , thats not latency, not even close
All screenshots of latencymon in this topic do not say anything. But this program shows critical problems in the system.
Sometimes small problems can be solved.
Using windows performance analyzer or other programs, parsing graphs for a beginner who came to the forum is an extremely difficult task.
thizito wrote: focus on hardware, stop focusing on stupid useless windows,
i debugged and reversed windows heavily and there is no fix.
Yes, you are right. You can send everyone who has a motherboard with the loud names "super mega ultra gaming oc" and a price of less than 600 bucks to the store to buy a new one.
thizito wrote: and you all lack of scientific method
lack of science at all . is alarming sad and frustrating
the whole blurbuster forum has 0 useful tests
This forum is not focused on solving system latency problems.

(Strange messages from a guy with a 2014 registration.)
I often do not clearly state my thoughts. google translate is far from perfect. And in addition to the translator, I myself am mistaken. Do not take me seriously.

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: Windows Server has less CPU lag [Lower DPC Latency]

Post by Chief Blur Buster » 24 Nov 2022, 05:38

Just only reading this now.
thizito wrote:
30 Jul 2022, 02:29
[beep]
Patience, my padawan.
Maybe, one day, you will learn why this forum exists.

If you wanted to get to Popular Science level, you go to www.blurbusters.com/area51 and NOT this forum.

If you wanted to get to Scholar Paper level, you go to www.blurbusters.com/research-papers and NOT this forum.

Not all posts here are always accurate but many hobbyists users here make a faithful-attempt results. Sometimes latencymon results are definitely junk, but not 100% universally always.

The forum discuss >99% casual/anecdotery/useful/semiuseful/junk to get golden <1% peer-preview. Many of our free test inventions sometimes come from forum discourse.

This section is not the Laboratory section.
If you are a researcher, you post in Area51 instead.
If you are a user or hobbyist, and just want casual discussion, you post here.

Some things that happened to displays in the last 2 years came from a lot of forum discussion 8 years ago, that consequently later cascaded to research papers, eventually vetted by researchers.

Here, this is only the butterfly of the entire Chaos Theory. But sometimes the good things only happen when the millionth butterfly happens.

Indeed, latencymon doesn't measure perfectly, but it does measure usefully enough to be a troubleshooting tool for some cases.
Even mere things like www.testufo.com/animation-time-graph sometimes accidentally discovered inefficient RGB-animating system tray utilities, even though that tool wasn't designed to detect performance problems caused by system tray software -- but ended up useful for that purpose. So a test that is not perfectly/nor designed for specific purposes, may still produce some useful data.

Here, (for the most part) let discussion happen, without assumptions. Back in 1993, people trusted forum discussions in places like sci.* usenet newsgroups more than today. Trust in science has eroded. But even before deciding on what merits peer review or creating a paper, this is a watercooler room of discussion that we encourage.

30 years ago people didn't believe humans could tell apart 30fps vs 60fps. I micdropped it over 30 years ago with MOTION.EXE for MS-DOS.

Image

Even when there's, say for example, 1%-5% chance your assumption is wrong and is not placebo, I pounce on blanket statements like "1ms doesn't matter" (depends on the millisecond) or things like "It can't have any effect" or things like "It's definitely placebo" (maybe true for the OP or you, but might not be true for entire population). Some are important and some are not important. As refresh rates go up, motion blur goes down, resolutions get sharper (4K+), framerates go higher, displays get bigger, etc, ever-smaller motion imperfections become more and more visible, that was invisible at 640x480 30fps or was invisible at 1920x1080 120Hz etc.

Yes, quality of forum (signal to noise) has degraded since 2014, as all forums/twitter/facebook has, but if you scroll down to the Area51 ("Laboratory"), the posts there are vastly higher quality -- or better yet -- stick to www.blurbusters.com/area51 instead.

Sometimes a one-in-a-million posts (...actually our batting average has been better than that...), butterflys eventually to a real peer reviewed research paper, with a research team. There's many topics that will remain unresearched, but we let discussion happen, and sometimes a researcher pounces on it.

Here's an example of a real peer reviewed science paper triggered by one of my tweets:
viewtopic.php?f=7&t=9785

There are many examples of micro-situations. Let's randomly pick one real-world example. Even a delay of a few tens or hundreds of microseconds to a mouse poll can jitter an 10,000 pixels/sec mouselook turn or a mouse cursor to the next poll interval (e.g. random gap instead of the regular gapping), and if happens often enough, you have awful mousefeel when you feel like things are randomish even in solo offline gaming (aimtrainer, bots, etc). And, also, there is also computers that can only do 4000 Hz mousepolls (cant do 8000Hz) because that specific computer had really crappy USB implementation that also created very high DPC latencies. Run latencymon on the 2000Hz/4000Hz-only mouse system, and run latencymon on the 8000Hz-mouse-capable system, and sometimes (not always) you see really INTERESTING differences -- that triggers you to run other software or try to track things down. That's one REAL-WORLD situation... Latencymon was made for a different purpose (sound), but as mouse pollrates start enroaching audio ranges (8000Hz), even latencymon sometimes creates interesting data on systems that is unable to do 8000 Hz poll rates.
DPC latency can cause sound issues (see conference paper mentioning latencymon https://www.aes.org/e-lib/online/browse.cfm?elib=19592 ...) but nobody has yet tested how latencymon affects visuals, but in the era of 8000Hz mice (that has NOW a peer reviewed research paper), it's a legitimate stone to unturn.

People continually produce better software, and there can be better software than latencymon, but sometimes you have to use the blunt end of a screwdriver as a temporary hammer for a couple of nails, when no hammer is within 15 miles of you. Right Tool For The Right Job sometimes is limited to the tools you know about / have access to. Not everyone here is Chief Blur Buster.

Blur Busters has indirectly triggered many research papers, some that cite me and some that don't -- but over 25 do -- www.blurbusters.com/research-papers (Google Scholar). And you can always read www.blurbusters.com/area51 -- the centralized research portal.

Blur Busters push the bleeding edge of display-related temporals (GtG, MPRT, lag, PWM, blur, stroboscopics, step effects, Hz, fps, VRR, jitter, stutter, etc) -- as we're in the Temporal business....

But yes, sure, this topic may be 100% junk. Who knows? However, discussion remains open.

As a deaf person, this forum is my watercooler, rather than a lunchroom of of casual-chatty researchers.

This forum is the Butterfly Tank running the Chaos Theory that sometimes outputs good stuff. You've probably forgot to read about the good stuff, that originally came from posts/indie/forum/discussion.

Please browse www.blurbusters.com/area51 if you want researcher-quality stuff.
This is but a forum.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

Post Reply