Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

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.
User avatar
Discorz
VIP Member
Posts: 999
Joined: 06 Sep 2019, 02:39
Location: Europe, Croatia
Contact:

Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by Discorz » 28 Jan 2022, 16:33

Gamers want the lowest latency possible, but how much does this actually matter? Can you even feel the difference between 5 or 10 or 20 milliseconds of lag? Aperture Grille's Latency Split Test (LST) is a program designed to help you determine your personal sensitivity to lag.

LST is a UE4 "game" which tasks you with choosing between two randomly presented scenes: one with no lag added, and one with a user-configurable amount of latency. Each test is composed of 16 trials. If you can correctly choose the high-latency scene at least 13 out of the 16 trials, you can be confident that you can feel a real, perceivable difference at that latency setting. If you successfully complete a test, lower the added latency and try again. Once you can no longer score 13/16, that will be your sensitivity threshold.

What is your threshold? Share your results!

Download latest version of Aperture Grille's Latency Split Test

Setup:
  • Run in fullscreen or borderless mode. V-Sync is off by default.
  • Cap frame rate to exactly 1000 FPS (1 ms frame time).
  • If frame rate is under 1000 FPS, lower rendering resolution until 1000 FPS can be held continuously.
Test:
  • Use mousewheel up/down to increase or decrease the added latency.
  • Spacebar to begin test (or end test prematurely). Latency cannot be changed during a test.
  • Right click when mouse is stationary to swap between A and B.
  • Left click to select what you believe to be the high-latency scene.
Post-test:
  • Decrease the latency until you can no longer correctly guess at least 13 out of 16 trials.
Check "Latency Split Test Readme.txt" for more information.


If you have any suggestions to make this program better, or if you encounter any bugs, email Ashun.
Last edited by Discorz on 29 Jan 2022, 05:29, edited 1 time in total.
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup

justnvc
Posts: 4
Joined: 28 Jan 2022, 18:40

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by justnvc » 28 Jan 2022, 18:44

This is great! To make it better, I would personally suggest adding a pill shaped target, or thicker inverted color line behind the white line that the user controls, that changes direction at random intervals on the horizontal plane. There would be no interaction between both lines, but tracking these changes is part of what makes input lag perceivable at sub 10ms.

Currently I can perceive 10ms correctly every time on my 360Hz display, but once it dips to 5ms it becomes much harder. I would be very interested to see if this changes when I am following a target. Perhaps you can make the speed of this target variable, so that the user can change it?

Good stuff!

User avatar
Stalast
Posts: 6
Joined: 29 Jun 2015, 13:34
Contact:

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by Stalast » 28 Jan 2022, 19:51

Here are my results in my sleepy, ready-for-bed state of alertness. I might give this another shot another day in the morning when I'm more awake.

Using a Logitech G203 and BenQ XL2411Z overclocked to 192 Hz

Image

Some feedback:
  • Both the A and B scenes felt pretty laggy despite running in fullscreen mode at 1000 fps. The test felt like I had to identify which scene is laggier as opposed to which scene has lag. Perhaps this is as intended and if so, that's fine.
  • I think that latency sensitivity may vary depending on the user's level of immersion in the game world. For me, I found it difficult to stay focused enough to care about moving a white line on my screen. I had to force myself to track the line. In a first person game world, I think that my level of focus and connection to the game world is stronger and more sensitive. Just something to think about.
  • I'm using a non freesync/g-sync monitor and the tearing is very distracting in this test.
The tearing:

Image

PiCKiN
Posts: 22
Joined: 03 Sep 2020, 00:27

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by PiCKiN » 29 Jan 2022, 04:13

Image

Here are my results with a BenQ XL2430T.

Couldn't pass 10ms test or below. Yikes!

I wonder if there would be a difference if I put a higher DPI on my mouse.

User avatar
Discorz
VIP Member
Posts: 999
Joined: 06 Sep 2019, 02:39
Location: Europe, Croatia
Contact:

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by Discorz » 29 Jan 2022, 06:48

Program only adds latency on top of your base system or program latency, so it shouldn't matter.

There are reasons why v-sync is off by default and why it needs to run at 1000fps (check Readme for more info). I guess you could test it at lower framerate with adaptive sync active but than you can't change additional lag in 1 fps increments. One way around is buying 1001Hz monitor with adaptive sync hh.

Adding 3D FPS also shouldn't matter, but it would be nice to see. With v-sync off new frames can pop anywhere on the screen, top-bottom. That is why full vertical line is used and not just mouse cursor. 2D is easier to run and those 1000 fps are important to maintain in these types of programs.

Since we are sharing images, here are my results. Anything below 9-10 ms becomes imperceptible. Even 10 is hard. Don't mind the uncompleted 8 ms test.

Image
With M32Q and G305

EDIT:
just got 14/16 for 8 ms, but holy c**p that took time, I was lucky too.
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup

MatrixQW
Posts: 279
Joined: 07 Jan 2019, 10:01

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by MatrixQW » 29 Jan 2022, 20:28

It was pretty clear to me with 10ms.
With 5ms I just need to focus, memorize how it feels and when I switch between scenes I can feel the difference between my hand movement and the delay on the vertical line.
A few times I could tell right away that the first scene was the lagged one and without switching to the next scene I took my chances and got it right.
What I do is swing the mouse left and right, not too wide and not too short, with a smooth and medium speed.
Tomorrow I wil try with 3 and 2ms.

I have an Asus VG258QR 165Hz TN and Logitech G400S.

And check this:
viewtopic.php?f=10&t=5065&p=39205#p39203
Attachments
matrix-latency-test.jpg
matrix-latency-test.jpg (24.27 KiB) Viewed 15675 times

MatrixQW
Posts: 279
Joined: 07 Jan 2019, 10:01

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by MatrixQW » 30 Jan 2022, 20:18

I always felt I am very sensitive to latency and I was right.
I notice my state of mind determines how much I can feel.
With a first attempt I still managed to make it in the 1 and 2ms. It's harder but I still notice it and that's what matters.

To be able to see/feel it, the movement must be constant and medium/high speed but can't be too fast because if the line shows tear or blurs too much it gets disguised.
So to me, it matters the processing input lag and also the response time of a monitor.
That's why I'm not so keen yet in IPS monitors because the response time is outside the refresh cycle.
I admit I never tried one but that's what I can tell by reviews.
If there is a 240Hz IPS with a response time of 4,16ms I would like to try it.
Attachments
matrix-latency-test2.jpg
matrix-latency-test2.jpg (31 KiB) Viewed 15288 times

User avatar
Chief Blur Buster
Site Admin
Posts: 11653
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by Chief Blur Buster » 30 Jan 2022, 20:45

Relevant link of why sometimes certain milliseconds matter, and why other milliseconds do not matter:

The Amazing Human Visible Feats Of The Millisecond
- Authored by me

This is a very good test that pushes the extremes of esports performance differences with milliseconds — it also can vary whether or not you keep a fixed gaze or use eye-tracking. Sometimes the way your eyes behaves, can mean you easily pick up 5ms differences, while you’re unable to pick up 20ms differences.

It can also partially explain why some people are performing differently even in sub-5ms legaues, while others don’t notice even at 10-20ms leagues — what kind of cues the human is using, because things like even VSYNC OFF “tearing pattern differences” look different at 5ms vs 10ms at certain refresh rates, and can provide a competitive advantage. A latency stimuli, a motion stimuli, a motion-difference stimuli, etc — this test has multiple potential stimuli hidden in it that is a carbon copy of a lot of stimuli found in esports games that different esports athletes picks up on. Many researchers DO NOT understand this and end up putting insensitivitizing-biases into their tests.

It underscores the scientific importance of not dismissing milliseconds — because there are MANY cues than be picked up temporally, depending on what a person is doing with their eyes (stationary gaze on monitor versus eye tracking, as well as pre-training effects — aka the moving archery target effect)

I’m happy to consider commissioning additional scientific research that can reveal human performance differences in millisecond timescales, perhaps even using Ashun’s test — I’ve historically comissioned esports research in the past (e.g. www.blurbusters.com/human-reflex …) as Blur Busters’ prime directive is to mythbust all kinds of display temporal bleep (MPRT, GtG, lag, VRR, Hz, etc)
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!

User avatar
kyube
Posts: 129
Joined: 29 Jan 2018, 12:03

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by kyube » 31 Jan 2022, 08:56

MatrixQW wrote:
30 Jan 2022, 20:18
I always felt I am very sensitive to latency and I was right.
I notice my state of mind determines how much I can feel.
With a first attempt I still managed to make it in the 1 and 2ms. It's harder but I still notice it and that's what matters.

To be able to see/feel it, the movement must be constant and medium/high speed but can't be too fast because if the line shows tear or blurs too much it gets disguised.
So to me, it matters the processing input lag and also the response time of a monitor.
That's why I'm not so keen yet in IPS monitors because the response time is outside the refresh cycle.
I admit I never tried one but that's what I can tell by reviews.
If there is a 240Hz IPS with a response time of 4,16ms I would like to try it.
So, when are you dropping the best peripherals, hardware and software setup guide, latency guru? :D
I'm astonished there are people that perceive under 5ms of input lag!

teo
Posts: 33
Joined: 15 Apr 2021, 19:28

Re: Should you be worried about 1, 5, or 10 ms of input lag? Use Latency Split Test (LST) and find out!

Post by teo » 31 Jan 2022, 22:20

this was a fun exercise, thanks for sharing. I always considered myself fairly latency insensitive--just the other day I realized I played a run of Hades with my tv at 60hz rather than 120hz. that said, I've also tried playing 2d platformers when my tv was not on 'game mode' and found some early mario levels to be extremely challenging until I figured that out.

I started at 8ms and ended up with 13/16. toning it down to 5ms and I didn't beat chance, 7/16. I changed my strategy a few times (slower wiggles, faster wiggles, faster flicks) and eventually felt like I found the difference to be most noticeable at the onset of movement. anecdotally this seems to line up with my experience at different frame rates, where somewhere beyond ~180 I struggled to notice much improvement to the 'snappiness' of gameplay.

I'd love to see some esports professionals' results in something like this. we know that there's a latency penalty at low dpi, but plenty of pros still use 400-800. and then there's polling rate where I don't think anyone has decided the jump to 8k is worth using the viper, not to mention the tested movement latency discrepancy between one mouse model to another at 1000hz. I have to imagine most pros could differentiate the ~10ms that some would gain from that alone. although as chief stated, maybe noticing this difference in a lab test is inconsequential if their latency sensitivity is lower during gameplay.

let's pretend that 8ms is the lowest latency where I can hit 13/16. would we also expect that I would only be able to distinguish differences of 8ms or greater? that is, would I perceive a 20ms delay as slower than a 15ms, or would they feel the same? could I always pick out 30ms as better than 40ms, or 100ms as better than 110 or 120ms? it might be a cool feature to let the user assign a baseline latency penalty for both A/B in addition to the discrepancy.

Post Reply