Interesting Linus video about mouse input lag

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
do0om
Posts: 34
Joined: 15 Jan 2018, 13:14

Interesting Linus video about mouse input lag

Post by do0om » 04 Jul 2019, 21:38

Linus compared mouse latencies using slow motion camera to verify Logitech claim that wireless is even faster than wired.
I think he makes some mistakes like when he talks about 1000 hz polling rate that does not matter on 240hz monitor.

Results like these are interesting.
I wish he could test more mouses like this.

https://www.youtube.com/watch?v=orhb7Njj3h8

Morkai
Posts: 10
Joined: 23 Jan 2014, 15:15

Re: Interesting Linus video about mouse input lag

Post by Morkai » 05 Jul 2019, 10:17

do0om wrote:Linus compared mouse latencies using slow motion camera to verify Logitech claim that wireless is even faster than wired.
I think he makes some mistakes like when he talks about 1000 hz polling rate that does not matter on 240hz monitor.

Results like these are interesting.
I wish he could test more mouses like this.

https://www.youtube.com/watch?v=orhb7Njj3h8
Yeah, the 500Hz polling mouse is on average 1ms slower, assuming everything else is identical. But if you switch one single component to a slower frequency, its fastest measured value will still be just as fast as the others with faster frequency. (if he tried a 1Hz mouse, one sample out of 1000 would still show a perfect latency just as low as a 1000Hz mouse, etc. Its important to use the average or median in these types of tests. using the best value basically removes the mouse from the equation rendering the test of the lower refresh mouse pointless).

Using the lowest value is still interesting for the input->rendering->output pipeline, but not for the mouse.

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

Re: Interesting Linus video about mouse input lag

Post by Chief Blur Buster » 05 Jul 2019, 10:25

Good to see these things addressed more often! It's not just latency but also visible stutters.

1ms vs 2ms can produce some noticeable visible differences in microstutter amplitudes for high Hz, for variable refresh rates, and for blur reduction.

1ms translates to 1 pixel per 1000 pixels/sec, so at 4000 pixels per second mouseflick, that's a 4 pixel microstutter for a 1ms poll misalignment to real-world time.

For extreme situations (0.5ms MPRT), even 1ms precision is not enough -- we really need 2000Hz+ mice for 0.5ms MPRT displays. (In this case, I'm talking about strobed operation -- MPRT, not GtG -- as sub-refresh MPRT(100%) is only possible with strobe-backlight-based motion blur reduction)

Many things that mice do can help compensate (e.g. internal 6000Hz sensor readout, to the point where 500Hz poll does not feel quite as granular) but not all of them. How the poll frequency and DPI translates to increased microstutters can be quite a complex topic, but the point is that 1ms (and even 0.5ms) microstutters can be human visible on ultralow-MPRT displays.
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!

ad8e
Posts: 68
Joined: 18 Sep 2018, 00:29

Re: Interesting Linus video about mouse input lag

Post by ad8e » 05 Jul 2019, 17:20

There's a way to show the effect of 1 ms delay in a way that viewers can decide if its impact matters. I have a Tearline Jedi-alike that Chief Blur Buster also has access to. When displaying both the hardware cursor and an internal cursor, the internal cursor can outrace the hardware cursor by cheating, if you tear the screen in the middle. This is because the hardware cursor position is fixed at the beginning of scanout. When the internal cursor is fixed at a later tearline, it displays a more up-to-date position below the tearline and a delayed position above the tearline. Dragging the tearline up and down, the internal cursor changes latency by +-16 ms. At 4 ms, it's very obvious, 2 ms is guessable at 100% with a few mouse swipes, and 1 ms is hard to guess accurately in a blind test.

As for blind tests without the hardware cursor to compare to, viewtopic.php?f=10&t=1134 has some community results, although no comments on the accuracy of that program. I used my own setup: 16 ms can be guessed at 100% rate without any effort, but I didn't test anything below that.

Mouse latency is already well tested, but everyone usually tests clicks. It's good that Linus tested movement instead, which adds to the body of knowledge. I hope he tests keyboard latency too while he still has his nice high speed camera, since there's little testing of the low-latency keyboards (Logitech G Pro Keyboard/G513 Linear, Corsair K65 Rapidfire, Wooting One, Cherry MX 6.0).

You can test beyond the framerate of the monitor with a different methodology that uses the computer clock; I think it's described in the keyboard latency thread. Even a 1 Hz monitor can produce accurate results down to 0.1ms, but the camera still must be high speed.

Morkai
Posts: 10
Joined: 23 Jan 2014, 15:15

Re: Interesting Linus video about mouse input lag

Post by Morkai » 08 Jul 2019, 08:44

ad8e wrote: Mouse latency is already well tested, but everyone usually tests clicks. It's good that Linus tested movement instead, which adds to the body of knowledge. I hope he tests keyboard latency too while he still has his nice high speed camera, since there's little testing of the low-latency keyboards (Logitech G Pro Keyboard/G513 Linear, Corsair K65 Rapidfire, Wooting One, Cherry MX 6.0).
This guy has tested quite a few keyboards:
https://danluu.com/keyboard-latency/
http://danluu.com/input-lag/

When i read it it i tested my keyboard too (novatouch tkl topre) and it did indeed seem slower than my mouse (about 30ms, but it wasn't a perfect test. I tried a reaction clicker with mouse vs keyboard many series - but key travel time is much greater on keyboard so that could account for most of the delay. Anyway, this site is what triggered me to make the full chain input/rendering/output test in the other thread, I'm going to expand on it with keyboard.
I am going to re-test it properly in 960fps but i want to find a ps/2 adaptor first to try usb/ps2 at the same go.

There are probably a LOT of people playing with 30-50ms delay on keyboard without noticing it. (and/or possible similar delay just to press a long-travel mechanical button).
It's probably because we think initiating movement has acceleration and the brain just processes the delay in acceleration as part of the games movement. (while a similar delay in panning with the mouse would feel like the whole world is syrup because we can feel the eye-hand coordination).

ad8e
Posts: 68
Joined: 18 Sep 2018, 00:29

Re: Interesting Linus video about mouse input lag

Post by ad8e » 08 Jul 2019, 09:19

For lower end and common keyboards, there's a variety of published tests, but I didn't find your Novatouch in any of them. Generally, those keyboards all suck, so your Novatouch will probably not perform well either.

What you say about people just being used to the delay is probably right. My Thinkpad keyboard, which was considered the holy grail of laptop keyboards, has even worse problems than a simple 50 ms delay, but it took a long while before I became suspicious enough to test it properly. I think for the average user, keyboard latency is worse than mouse latency because it has the potential of hiding ordering problems. A 50 ms keyboard delay is not so bad for typing, but sending keys out of order is.

Morkai
Posts: 10
Joined: 23 Jan 2014, 15:15

Re: Interesting Linus video about mouse input lag

Post by Morkai » 08 Jul 2019, 10:39

ad8e wrote:For lower end and common keyboards, there's a variety of published tests, but I didn't find your Novatouch in any of them. Generally, those keyboards all suck, so your Novatouch will probably not perform well either.
Well the novatouch tkl, like all topre. was high end when new so should hopefully have electronic as good as any (and interestingly, since its the only topre keyboard that accepts standard keycaps, price has more or less doubled since new and seems to commonly sell for hundreds of dollars). Ill get back with results soon.

Do you have a list somewhere of keyboards that perform better than the ones Luu have tested?

ad8e
Posts: 68
Joined: 18 Sep 2018, 00:29

Re: Interesting Linus video about mouse input lag

Post by ad8e » 08 Jul 2019, 11:17

I stated that wrong. What I mean is that I consider all mechanical keyboards low end, including the most expensive and custom ones, unless the manufacturer makes claims about latency and performance. This is because good latency is sufficiently unusual that a keyboard with engineering effort towards this goal would brag about it. Absence of such bragging implies poor qualities.

My current notes on keyboards are at ---- [edit, it's obsolete now, check viewtopic.php?f=10&t=5540]. Those are music oriented, and make four keyboard recommendations, sorted from best to worst. A Wooting One is being shipped to me, so if you wait a week I'll have better info on that keyboard, and its ranking might improve to first. The Apple Magic Keyboard's latency looks good, but I seem to have written it off for reasons I can't recall now.

For keyboards intended for typing, I'll describe how priorities are different from keyboards for music. 10KRO is not important at all. 4KRO is enough, or 2KRO with modifier key handling. Latency matters less, but ordering really matters. (Ordering = the keys arrive in the order you pressed them.) However, nobody tests ordering so we can only measure latency. I read that PS/2 keyboards have a 200 ms max rate, and that might imply a small delay problem when sending two or more keys, but it could be meaningless or unimportant. I haven't bothered to test this since PS/2 is dying out anyway.

For gaming, latency still matters a lot, and 10KRO is not important but 6KRO is.
Last edited by ad8e on 11 Jul 2019, 17:52, edited 2 times in total.

LatencyLag
Posts: 1
Joined: 17 Jan 2020, 22:43

Re: Interesting Linus video about mouse input lag

Post by LatencyLag » 17 Jan 2020, 22:49

do0om wrote:
04 Jul 2019, 21:38
Linus compared mouse latencies using slow motion camera to verify Logitech claim that wireless is even faster than wired.
I think he makes some mistakes like when he talks about 1000 hz polling rate that does not matter on 240hz monitor.

Results like these are interesting.
I wish he could test more mouses like this.

https://www.youtube.com/watch?v=orhb7Njj3h8
The best way to find out latency is to actually measure it yourself--don't believe any manufacturer's specs, they are notoriously optimistic and also may be degraded by your local conditions, both in RF interference (for wireless), the mouse surface, or even dust on the sensor and/or low battery. The best meter for this is the $95-ish LagMeter from Goose Enterprises (available on Amazon), which is an external hardware measurement instrument and doesn't rely on the computer itself--it therefore gives exact lag data down to 0.01 milliseconds on a digital readout on the instrument. Using a LagMeter, you'll see that lag for some mice and surfaces is more consistent than others. If, for example, the optical detector fails to pattern-detect on a particular frame of mousepad fabric, it will simply skip, and the result will be an extra 4-10 milliseconds of lag for that one test. Or, if a wireless packet is dropped, there will be a similar variance in your lag measured. With a LagMeter, you can test repeatedly, say 10 times in a minute, and get a feel for the consistency of latency, which may be equally or more important than the median value of the latency in effect on gaming performance, since by Murphy's Law, those dropped packets will happen exactly when you most don't want them to, e.g. at the moment an opponent suddenly leaps into your face.

deama
Posts: 368
Joined: 07 Aug 2019, 12:00

Re: Interesting Linus video about mouse input lag

Post by deama » 15 Jun 2020, 22:20

ad8e wrote:
08 Jul 2019, 11:17
I stated that wrong. What I mean is that I consider all mechanical keyboards low end, including the most expensive and custom ones, unless the manufacturer makes claims about latency and performance. This is because good latency is sufficiently unusual that a keyboard with engineering effort towards this goal would brag about it. Absence of such bragging implies poor qualities.

My current notes on keyboards are at ---- [edit, it's obsolete now, check viewtopic.php?f=10&t=5540]. Those are music oriented, and make four keyboard recommendations, sorted from best to worst. A Wooting One is being shipped to me, so if you wait a week I'll have better info on that keyboard, and its ranking might improve to first. The Apple Magic Keyboard's latency looks good, but I seem to have written it off for reasons I can't recall now.

For keyboards intended for typing, I'll describe how priorities are different from keyboards for music. 10KRO is not important at all. 4KRO is enough, or 2KRO with modifier key handling. Latency matters less, but ordering really matters. (Ordering = the keys arrive in the order you pressed them.) However, nobody tests ordering so we can only measure latency. I read that PS/2 keyboards have a 200 ms max rate, and that might imply a small delay problem when sending two or more keys, but it could be meaningless or unimportant. I haven't bothered to test this since PS/2 is dying out anyway.

For gaming, latency still matters a lot, and 10KRO is not important but 6KRO is.
Have you tested the Wooting One yet? Any results?

Post Reply