Reduce Input Delay (Mouse & Keyboard Buffer Size)

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.
Post Reply
slaver01
Posts: 89
Joined: 21 Sep 2020, 01:48

Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by slaver01 » 25 Jan 2021, 05:36

I came across this video this morning. do you think there is a grain of truth to reduce the input lag of peripherals?

https://www.youtube.com/watch?v=FkVQm-hyL28

Best regards!

andrelip
Posts: 160
Joined: 21 Mar 2014, 17:50

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by andrelip » 26 Jan 2021, 00:33

The most significant inconsistency people are feeling is caused by the C-State messing things up. Windows don’t respect the Bios flag for some models. Also the PowerPlans are not fully optimised to use “C1 only”, not even the Ultimate and the Bitsum.

How to check:

You can check the amount of time in every step with cmd -> perfmon and then adding the counters for C1, c2, c3 and Idle. The C1 and Idle counters should be the same (for C1 max) or both zero (in case of real c0). Anything different is bad. The frequency graph in MouseTest is a good benchmark since it’s very sensitive to latency. Just check the variance. It should be much more precise in C0 or “C1 only” an more spread in deeper states. Variance is caused mostly by the cpu delaying the Interrupt Processing (making the counter lower than frequency) and then processing it in batches (making higher than the frequency of the mouse).

How to fix

Use Power Saver Explorer and choose one of the following:

A: Processor -> Disable Idle and chave to disable idle. This will force C0, the faster state but will generate MUCH more heat.

Disclaimer: It will not work well for systems with HT enabled. HT needs the idle thread to work efficiently and with idle disabled the Windows keep the C0 by doing an endless loop. With HT off it should be the fastest possible, but check your temps!

B: Under processor put both PROMOTE and DEMOTE state threshold to 100. It is trick because in one they use the Busy counter and in the other the Idle. It will force C1 State Max, in perfmon it should be equal to the idle graph as a idle is in C1. Probability the best, since Disable Idle They have diminished returns, and C1 is Already too powerful. It works well with HT.

After check MouseTester again. You don't need to reboot, so you can easily do A/B testing.
Last edited by andrelip on 26 Jan 2021, 10:53, edited 1 time in total.

Ragerlis
Posts: 32
Joined: 06 Sep 2020, 09:40

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by Ragerlis » 26 Jan 2021, 01:55

I remember from 2010's mousebuffersize was a problem with many error inputs etc, but after a few windows versions they just fixed and set the default to zero, so doing that you are actually increasing the buffer after all.

But im not sure and i dont have enough knowledge about it, but looking in my registry, the key doesn't even exists(latest win10 version) reforcing the default is just zero.

Ferr0
Posts: 31
Joined: 26 Jan 2021, 10:40

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by Ferr0 » 26 Jan 2021, 10:47

thizito wrote:
26 Jan 2021, 03:02
Sorry if i was too direct. But there is tons of things that can reduce or change input delay
Its very dificult to a single dword to do magic. magic no exists

Messing with OS like u just said and let me add , messing with BIOS and even hidden bios settings is like 99999% more effective and noticiable

We could make a post about Ram OC, tWR tRTP, grub settings, and people in the end will have better results.
Things that for me is way more noticiable, list:

-> Changing tWR tRTP (messing with this, to ME mess with game feeling alot)
-> Power downs in bios
-> Self refresh
-> Using PMT (gets lower performance and way smoother, depends the choice of the game.. old games are perfect)
-> Using an old operational system with way less updates and very minimalistic approach
-> Changing MSI(for the good and for the bad.. enabling can sometimes be worse, must always test this)
-> Disabling drivers i have a bunch of a list with idk 68 drivers for w10 that i just disable, most of them security or useless

few examples that comes to mind, that u could A vs B test easy

edit: Also using DisableWriteCombining and messing with Async in nvidia inspector is big changes both in performance and feeling, but atm im using the lastest nvidia drivers with default FRL, felt great.
Isn't setting frame rate limiter mode to disabled the best choice for input delay?
Ragerlis wrote:
26 Jan 2021, 01:55
I remember from 2010's mousebuffersize was a problem with many error inputs etc, but after a few windows versions they just fixed and set the default to zero, so doing that you are actually increasing the buffer after all.

But im not sure and i dont have enough knowledge about it, but looking in my registry, the key doesn't even exists(latest win10 version) reforcing the default is just zero.
I've had my mouse buffer size set at 0 and I've also had it set at slightly higher values like 1, 8, and 16. There is a drastic difference between setting it to 0 and setting it to 1 or a slightly higher value. At too low a value you get things like clicks not registering but at 0 none of that happens. I think 0 just sets the buffer size to its default, which is probably really high like 100. I do think setting it to dec 16 is DECREASING the buffer size from default. I've set it to 12 before as well and had no issues with dropped clicks but the cursor movement did seem a little less smooth, however it did feel snappier.

User avatar
kurtextrem
Posts: 41
Joined: 05 Mar 2017, 03:35
Location: Munich, Germany

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by kurtextrem » 01 Feb 2021, 04:56

andrelip wrote:
26 Jan 2021, 00:33
B: Under processor put both PROMOTE and DEMOTE state threshold to 100. It is trick because in one they use the Busy counter and in the other the Idle. It will force C1 State Max, in perfmon it should be equal to the idle graph as a idle is in C1. Probability the best, since Disable Idle They have diminished returns, and C1 is Already too powerful. It works well with HT.

After check MouseTester again. You don't need to reboot, so you can easily do A/B testing.
Can you please tell me which setting it is based on the order of the settings? I think promote is first setting next to the CPU power management (in PowerSettingsExplorer) and demote the 5th?
Or maybe give us a screenshot?
Acer XF250Q, R6 competitive player

Ferr0
Posts: 31
Joined: 26 Jan 2021, 10:40

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by Ferr0 » 01 Feb 2021, 07:16

andrelip wrote:
26 Jan 2021, 00:33
The most significant inconsistency people are feeling is caused by the C-State messing things up. Windows don’t respect the Bios flag for some models. Also the PowerPlans are not fully optimised to use “C1 only”, not even the Ultimate and the Bitsum.

How to check:

You can check the amount of time in every step with cmd -> perfmon and then adding the counters for C1, c2, c3 and Idle. The C1 and Idle counters should be the same (for C1 max) or both zero (in case of real c0). Anything different is bad. The frequency graph in MouseTest is a good benchmark since it’s very sensitive to latency. Just check the variance. It should be much more precise in C0 or “C1 only” an more spread in deeper states. Variance is caused mostly by the cpu delaying the Interrupt Processing (making the counter lower than frequency) and then processing it in batches (making higher than the frequency of the mouse).

How to fix

Use Power Saver Explorer and choose one of the following:

A: Processor -> Disable Idle and chave to disable idle. This will force C0, the faster state but will generate MUCH more heat.

Disclaimer: It will not work well for systems with HT enabled. HT needs the idle thread to work efficiently and with idle disabled the Windows keep the C0 by doing an endless loop. With HT off it should be the fastest possible, but check your temps!

B: Under processor put both PROMOTE and DEMOTE state threshold to 100. It is trick because in one they use the Busy counter and in the other the Idle. It will force C1 State Max, in perfmon it should be equal to the idle graph as a idle is in C1. Probability the best, since Disable Idle They have diminished returns, and C1 is Already too powerful. It works well with HT.

After check MouseTester again. You don't need to reboot, so you can easily do A/B testing.
Doing this does add input lag though

User avatar
BTRY B 529th FA BN
Posts: 523
Joined: 18 Dec 2013, 13:28

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by BTRY B 529th FA BN » 02 Feb 2021, 13:41

Ferr0 wrote:
01 Feb 2021, 07:16
andrelip wrote:
26 Jan 2021, 00:33
The most significant inconsistency people are feeling is caused by the C-State messing things up. Windows don’t respect the Bios flag for some models. Also the PowerPlans are not fully optimised to use “C1 only”, not even the Ultimate and the Bitsum.

How to check:

You can check the amount of time in every step with cmd -> perfmon and then adding the counters for C1, c2, c3 and Idle. The C1 and Idle counters should be the same (for C1 max) or both zero (in case of real c0). Anything different is bad. The frequency graph in MouseTest is a good benchmark since it’s very sensitive to latency. Just check the variance. It should be much more precise in C0 or “C1 only” an more spread in deeper states. Variance is caused mostly by the cpu delaying the Interrupt Processing (making the counter lower than frequency) and then processing it in batches (making higher than the frequency of the mouse).

How to fix

Use Power Saver Explorer and choose one of the following:

A: Processor -> Disable Idle and chave to disable idle. This will force C0, the faster state but will generate MUCH more heat.

Disclaimer: It will not work well for systems with HT enabled. HT needs the idle thread to work efficiently and with idle disabled the Windows keep the C0 by doing an endless loop. With HT off it should be the fastest possible, but check your temps!

B: Under processor put both PROMOTE and DEMOTE state threshold to 100. It is trick because in one they use the Busy counter and in the other the Idle. It will force C1 State Max, in perfmon it should be equal to the idle graph as a idle is in C1. Probability the best, since Disable Idle They have diminished returns, and C1 is Already too powerful. It works well with HT.

After check MouseTester again. You don't need to reboot, so you can easily do A/B testing.
Doing this does add input lag though
What creates input lag between those two choices?

For me, memory latency increases when only using Disable idle. That's with my 5800X. However it was the opposite with my 3800X.

Ferr0
Posts: 31
Joined: 26 Jan 2021, 10:40

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by Ferr0 » 02 Feb 2021, 22:17

BTRY B 529th FA BN wrote:
02 Feb 2021, 13:41
Ferr0 wrote:
01 Feb 2021, 07:16
andrelip wrote:
26 Jan 2021, 00:33
The most significant inconsistency people are feeling is caused by the C-State messing things up. Windows don’t respect the Bios flag for some models. Also the PowerPlans are not fully optimised to use “C1 only”, not even the Ultimate and the Bitsum.

How to check:

You can check the amount of time in every step with cmd -> perfmon and then adding the counters for C1, c2, c3 and Idle. The C1 and Idle counters should be the same (for C1 max) or both zero (in case of real c0). Anything different is bad. The frequency graph in MouseTest is a good benchmark since it’s very sensitive to latency. Just check the variance. It should be much more precise in C0 or “C1 only” an more spread in deeper states. Variance is caused mostly by the cpu delaying the Interrupt Processing (making the counter lower than frequency) and then processing it in batches (making higher than the frequency of the mouse).

How to fix

Use Power Saver Explorer and choose one of the following:

A: Processor -> Disable Idle and chave to disable idle. This will force C0, the faster state but will generate MUCH more heat.

Disclaimer: It will not work well for systems with HT enabled. HT needs the idle thread to work efficiently and with idle disabled the Windows keep the C0 by doing an endless loop. With HT off it should be the fastest possible, but check your temps!

B: Under processor put both PROMOTE and DEMOTE state threshold to 100. It is trick because in one they use the Busy counter and in the other the Idle. It will force C1 State Max, in perfmon it should be equal to the idle graph as a idle is in C1. Probability the best, since Disable Idle They have diminished returns, and C1 is Already too powerful. It works well with HT.

After check MouseTester again. You don't need to reboot, so you can easily do A/B testing.
Doing this does add input lag though
What creates input lag between those two choices?

For me, memory latency increases when only using Disable idle. That's with my 5800X. However it was the opposite with my 3800X.
So my personal anecdote is that when I first used that tweak I did feel like it was worse then idle enabled. However, now that we have some latency monitoring tools someone has actually tested it. This is the only actual data point we have regarding this tweak as far as I'm aware. https://youtu.be/sVK96m9I_U8

Ferr0
Posts: 31
Joined: 26 Jan 2021, 10:40

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by Ferr0 » 03 Feb 2021, 00:34

thizito wrote:
03 Feb 2021, 00:11
THIS IS NOT LATENCY MEASUREMENT
NOT LATENCY


N O T

L A T E N CY


UNDERSTAND THE FX!@%$ TEST.

the tests are as usefull as fr33thy as human being.
`Click to photon doesn't CONTINUOUSLY measure latency, it it only capable of measuring the latency of a certain event (eg, click.) This is a problem because latency is not stable, latency spikes exists. Which may go COMPLETELY unnoticed by this method of testing. You can slave alot of trials (like 100+) to minimize this, but its still entirely possible that all latency spikes would dodge the capture.`
By MaxTendency on this forum, just grabbed a quick search.`

To make a plus, fr33thy is terrible at tech. he burn people hardware and his own motherboards more then two times because he copied a XOCer timing settings... He is the guy who recommends Hardware Accel in programs to be off cause have 0 understanding of the setting, this is the level.

Plus v2. I would bet my pc that his shit is unstable linpack + poorly configurated and with alot RNG
Its entirely possible that latency spikes go ignored by a test with even 100 samples but using a hypothetical to dismiss data that does exist is a big misstep imo.

No offense but fr33thy has about as much credibility as anyone else in the pc tweaking community. There are a lot of missteps as people try to find the right answer. Arguing that fr33thy is terrible at tech is completely irrelevant and misses the point entirely. I really don't care if someone has said something wrong in the past, the point is he did click to photon tests in each game after showing the exact changes mentioned in the post above and showed that the latency slightly increased. On top of that a lot of the other benchmark categories got worse as well.

To me, this makes sense with what I've noticed but obviously it could be entirely different for someone else depending on their setup. For now, I won't be using idle disabled.

Brainlet
Posts: 100
Joined: 30 May 2020, 12:39
Contact:

Re: Reduce Input Delay (Mouse & Keyboard Buffer Size)

Post by Brainlet » 03 Feb 2021, 01:29

DO NOT disable idle when using Hyperthreading. His test is meaningless since he only tested it with HT on... In general I'd take his "testing" with as much scepticism as e.g. r0ach's list. Both make claims without proof. No, writing down some numbers in a spreadsheet with 0 context or how they've been measured is not "proof". Just because he has a tool it doesn't mean his data or testing methodology is correct nor do his settings and general setup match everyone else's. As usual, test things on YOUR OWN system and be aware of dependencies like these.
Starting point for beginners: PC Optimization Hub

Post Reply