What Are Response Times? Overshoot? Cumulative Deviation? - Performance Charts Explained

Support us on Patreon: / hardwareunboxed
Join us on Floatplane: www.floatplane.com/channel/Ha...
Buy relevant products from Amazon, Newegg and others below:
Alienware AW3423DW - geni.us/ymsAf
Gigabyte M27Q X - geni.us/0EOLg
HP X27q - geni.us/gQjEF
How we test response times: • What Are Response Time...
Testing performed using Portrait CalMAN Ultimate: www.portrait.com/
00:00 - Welcome Back to Monitors Unboxed
00:43 - What is a Response Time?
12:56 - What is Overshoot?
16:10 - What is Inverse Ghosting Rate?
18:32 - What is Total Response Time?
21:20 - What is Refresh Rate Compliance?
23:41 - What is Cumulative Deviation?
31:02 - Final Bits and Pieces
What Are Response Times? Overshoot? Cumulative Deviation? - Performance Charts Explained
Disclaimer: Any pricing information shown or mentioned in this video was accurate at the time of video production, and may have since changed
Disclosure: As an Amazon Associate we earn from qualifying purchases. We may also earn a commission on some sales made through other store links
FOLLOW US IN THESE PLACES FOR UPDATES
Twitter - / hardwareunboxed
Facebook - / hardwareunboxed
Instagram - / hardwareunboxed
Music By: / lakeyinspired

[@evila9076]

Please Test a crt. And some early lcds. Would be pretty interesting to see how far we've come (or haven't)

[@StefandeJong1]

Also a phone's oled screen like the Samsung Galaxy S22 Ultra. Just to see how good these screens are compared to oled and qd oled on TV's and monitors. These small screens are performing very highly as far as I know because they're easier to produce

[@polly_2526]

Crts work like some kind of utra high black frame insertion, so you can get perfect motion clarity, there is no point to test the response time of crts

[@newbietricki239]

I think CRTs are still the fastest response time regardless of refresh rate next is OLED. I still miss my old CRT monitor and tv sadly my parents threw it way for their First LCD tv. I could have used it for some old-school gaming

[@groundzero_-lm4md]

@@newbietricki239 OLED might actually be faster end to end since it removes the need for a digital to analog converter.

[@newbietricki239]

@@groundzero_-lm4md crt is direct analog

[@TheVillainOfTheYear]

Tim, your content is a rare gem in the tech KZbin scene. Excellent science, done professionally.

[@aspermypreviousemail5907]

Voiceover Tim, Monitor Steve, Moustache Tim... glad to see the team growing!

[@jakacresnar5855]

as someone who teaches kids for a living, you are an amazing teacher. Everything is super clear, you don't speak *too* quickly, you mark everything, your boardwork is exemplary and nothing is left out or glossed over. Bravo on all accounts!

[@benjaminoechsli1941]

First Monitor Steve, now Voiceover Tim? The HUB crew is growing exponentially! ;P Jokes aside, this video is _exactly_ the sort of content I want to see on Monitors Unboxed. Keep it up, HUB crew!

[@Alvin853]

Any particular reason the overshoot is given in percent and not absolute numbers? I don't have any hands on experience, but just thinking about it a 50% overshoot on a small transition will be far less noticable than a 50% overshoot on a large transition. As seen in the video, using percentage overshoot will make the transitions close to the diagonal seem pretty bad (mostly yellow-orange colors) on paper when the absolute overshoot isn't any bigger than on the transitions further from the diagonal.

[@zactron1997]

Understandable, and I reckon this is something subjective you could debate. I think a percentage makes more sense because in these kinds of system response analytics, you expect performance to be inversely proportionally to the step size. If that is not the case, then you may have a monitor which "flashes" before all chabges for example. Niether is right or wrong, what matters is to know what is being measured.

[@monitorsunboxed]

What you perceive on the screen is often dictated by the entire context of the screen, so for example in a mostly dark scene you'll find it easier to perceive small changes in values compared to if the scene had a larger range of values. So while at times it will be easier to see a certain percentage overshoot for larger transitions, that's not always the case, there are all sorts of weird quirks with human vision that can affect that perception - remember, your eyes are adaptive. Just imagine a scene that's mostly dark and shadowy, most content in the RGB 0 to 50 range. If you had 0 to 30 transitions overshooting by 100% to RGB 60, that would be very noticeable in that context. The overshoot would be exceeding the brightest parts of the scene. But in a higher contrast scene you may not notice say RGB 200 overshooting to RGB 230, the same 30 RGB value difference (but just 15% overshoot overall) I guess when it comes to these measurements it depends on what assumptions you want to make and each method has its pros and cons. Absolute values works well when a scene has high contrast, but underplays the effect of overshoot in low contrast scenes (in my opinion). Relative values has the opposite effect. We use relative values for all measurements (eg tolerances as well) which keeps it consistent across the board

[@tarsius]

@@monitorsunboxed That's why you use gamma corrected response measurements, isn't it? So the same RGB value differential would ideally represent the same light perception response from a human observer regardless of which part (dark or light) of light output range the transition is taking place. Again, ideally. But using percentages in this case is definitely wrong in its deceptiveness (from the color scheme perspective as well - green to red).

[@monitorsunboxed]

@@tarsius Gamma correction is mostly about solving the problem of where you set the start and end points for your response time measurement. The total response time is what it is, gamma correction doesn't change that. However when incorporating tolerances, using linear light output instead of gamma corrected light output creates a mismatch where the measurement start point can be way off the actual start point, but the measurement end point can be very close to the actual end point (perceptually). With gamma correction we get a much more even balance between the start and end points and how they compare to the actual start/end However even with gamma correction we're still using tolerances that are relative to the size of the transition, for similar reasons to what was described above - and also that with fixed tolerances eventually given a small enough transition you'll measure a response time of 0 (or even negative). Gamma correction doesn't solve the issue where dark transitions are more noticeable in the context of a dark scene. For example you could argue that the transition time of small transitions is less relevant than large transitions because they are transitioning over a smaller absolute RGB value range and therefore less noticeable. That's similar to saying that the overshoot percentage of a small transition is less noticeable because the actual absolute value of the overshoot is lower. Gamma correction doesn't correct for either of these perception issues, but it does make what you are measuring more consistent and fair in other ways

[@tarsius]

@@monitorsunboxed Your response сontains some quite contradictory (for me) statements. I’m not an expert in this domain, so you may be right and I may be entirely wrong here. But I’ll try to provide the reasoning behind my previous comment and be as brief as possible in that. So bear with me if you can and ask if you’ll lose my discourse at some point. Gamma correction which is used to encode and decode luminance is based on a power law between input and output values. This law was used here based on the assumption that human perception of luminance follows an approximate power function. This way, gamma correction of input values (voltage which is almost linear to the light output in LCD, AFAIK) produces output ‘RGB’ values. And equal steps (differential) in these values roughly correspond to subjectively equal steps in human perceived luminance. That’s why ‘absolute numbers’ (mentioned by @Alvin853) or ‘RGB’ values (in other words) as differentials (from-to) should represent the same subjectively perceived changes in luminance in different parts of the luminance range (90 to 105 and 195 to 210, for instance).That’s why measuring over/undershoot and tolerances (allowances) in percentages is deceptive (therefore wrong from this perspective). P.S. There is a ton of questions even in the fundamental reasons behind the usage of gamma correction based on power law here. For example, the difference in perceived brightness depends on the color of the objects observed, ambient lighting and other conditions (even the mood of an observer, actually) . So, different models of perception (psychophysics) based on power laws are all rough approximations at best (with lots of factors and their relationship lost in such approximation). But that is another topic for discussion.

[@Acid_Burn9]

Was waiting for this video for a while. Thanks, Voiceover Tim!

[@solidreactor]

Would be nice to add icons next to each section like the "response time", "Total Response Time", "Inverse ghosting" and "CD". I am thinking of icons like the graphs you nicely drew :) One normal "s-curve" (response time), one S-curve that overshoots (total response time) e.t.c. you get the idea. Great video, thanks for sharing your amazing methodology

[@zactron1997]

Excellent explanation Tim, I learnt this stuff during my undergraduate from the perspective of minimizing system response errors. The only suggestion I'd make is to consider replacing the excel tables with actual heat maps that interpolate between values. Using something like Matplotlib in Python you could probably do that for free using that Excel data directly. For most people the exact numbers don't really matter on those heat maps, and just make them a bit harder to read. Excellent summary statistics though. The only one you could be missing is a measure of the derivative of the response curve. This would give you a good measure of how "flashy" the response is. For example, a low overshoot and a low response time will yield a mostly low CD value, but if it is highly unstable and fluctuates as it reaches the steady state value, it could look like it's strobing or flashing. But measuring that is probably a challenge given your discrete sampling, and if most responses only have a single overshoot peak, it won't matter much.

[@stpirate89]

The derivative wrt time?

[@zactron1997]

@@stpirate89 yeah that's the metric, but it's infamously unstable to measure on discrete data due to the division WRT time step. The other metrics you could use is something like the highest frequency above a certain magnitude in the Fourier decomposition of the dataset, but that also has its own drawbacks.

[@LunarLaker]

I don't get the heatmap idea, sounds like work and potential for introducing error to get something that might be easier to interpret. Are there any trends in the data not already visible? can it extrapolate overshoot for transitions in the diagonal? would these transitions have the right rule for the gradient (linear, exponential, hyperbolic etc)?

[@stpirate89]

@@zactron1997 ok I've had a chance to watch the whole video now. I didn't know if you meant the derivative of the "heat maps" hence why I asked. I also agree with Nice Lake, I don't see the need to interpolate the "heat maps", they're not even really heat aps, just tables with colour added, and I think they're fine as they are. Whilst I don't know what the time derivative is going to tell us about the monitors, assuming that the actual response curves don't deviate too much from the examples given, then finding the derivate it relatively straight forward, either as an average of the instantaneous derivates calculated from neighbouring differences, or as a difference from the cutoffs explained in the video. As for the FFT, I'm not sure what this is going to tell you either. For a perfect square wave we'd expect to see peaks at the fundamental frequency and all its harmonics, however this is only going to be half of one cycle, so we'd basically see nothing (a quick python test shows this, even if you have a gradual change of the signal levels). It's certainly not telling us the same thing as the time derivative of the response curve - a quick comparison of units show us that.

[@Your_Paramour]

I disagree that the chart should be switched to a heatmap for a couple of reasons. Firstly, presenting it this way shows you exactly what it is being measured and what is not being measured. Secondly, removing the raw value and just showing the colour means we as viewers are reliant on Tim's interpretation of what is acceptable, where as currently we don't have to. And thirdly, the only way to make the heatplots consistent across every monitor is to is to infer some relationship between rgb difference and response time. This relationship may not be consistent across different monitors at all, and it may not even be consistent on the same monitor across different refresh rates, rgb difference etc.

[@HazzyDevil]

Really makes you appreciate the work that goes into these reviews. Awesome stuff! I’ll never be buying a monitor without HUB and/or RTINGS going through testing it first.

[@brucethen]

Very interesting and informative, as an electronic engineer by trade, I knew some of the terms like overshoot and rise / fall time, but I now know more about how cumulative deviation is calculated and what it means

[@astreakaito5625]

Excellent! Now I can't wait for the video about black frame insertion and how incredibly important it is for most games (and how to use it correctly)

[@BUDA20]

I know all these already, but it will be useful to have a short video "showing" the best you can every visual "symptom" like dark level smearing, overshot, etc, obviously make a warning at the beginning that you can see only what the camera and the person monitor can display, but, I think a slow motion, photos and some real-time videos can visually illustrate a lot, and make it easy for a lot of us to share to others when we are explaining

[@gaurd3]

Nice. An explanation of the “I’ll take your word for it chart” 🤣

[@akawaiipotato3602]

Very in-depth and informative video yet easy to understand. Great job once again!

[@MPaulCezanne]

Fantastic video. I’m looking forward to future deep dives on: synch technologies, I/O standards, display calibration, color spaces, overdrive settings, panel tech, etc. Thanks voiceover Tim!

[@jeonghyunkim1579]

Now I understand your huge efforts for a single slide presentation. Thank you so much.

[@Sock-qv9wr]

Soo, the other day I discovered there once existed a Thinkpad with an integrated color calibration tool. We need something like that these days for our laptop monitors

[@samgoff5289]

You can just buy one that works on any screen

[@zaid19961]

I cant stress enough, how insanly amazing this video is, i hace been watching you for years, i never fully understood these numbers accurately, now i hqve a very good idea about how monitor response time is measured and read on those charts. Keep it up steeeeeve

[@shaharcarmi1783]

Waiting for this so long ! excellent explanation, You make it very easy to understand and now i can finally understand at such a deep level the data in these calculations. incredible :)

[@domm6812]

That is fantastic Tim! Genuinely impressive. Thank you for going through it in such detail.

[@two-keytech7578]

I have always wondered how you did all this. Thanks for the great walkthrough made it easy to understand.

[@aahlala]

This is the level of detail you love to see. Thank you!

[@grievesy83]

That was both fascinating and extremely useful information. Love it.

[@Blafard666]

Thank you for this, I am finaly ready to fully understand your monitor testing videos !

[@grospoulpe951]

very informative, thanks! :) Now, I also see the amount of work you put on testing a new display; again, thanks for that!

[@moevor]

Excellent video Tim and HUB, thanks for making this public!

[@Tony-Tech]

Nice work Tim! Well done. I learned a lot. Thanks for explaining everything.

[@Gc7788m]

I've always wanted to go more in detail about this

[@omarfouda1994]

Now back to watching all the reviews😂😂 This is exactly what I expected from this channel, appreciate the hard work

[@jeremyf1901]

Thank you so much for creating this video! I never fully understood and couldn’t find an understandable explanation.

[@ddpwe5269]

Great video Tim! It definitely has filled in the blanks when watching your monitor reviews =)

[@jjames267]

Seriously thanks for making this video!! I could make sense of good and bad from charts before but now I know much more about what I'm looking at. Now to go rewatch your reviews for the monitors I've been looking at!

[@TheCgOrion]

This is a fantastic video, and I'm sure it has helped many people understand your reviews more clearly now.

[@rayndooma5624]

Appreciate the insanely detailed video Tim! Keep up the awesome work.

[@LiraeNoir]

Thanks, I think it was comprehensive, clear, and certainly very useful for those who didn't know it. Good job! I certainly had forgotten that cumulative deviation was the area difference, calculus style. I'll wait for the color chart explanation, that I know very little about and want to know more.

[@AndyU96]

Thanks Tim, this was a long time needed one

[@jtmcgee]

Thank you for this explainer. I kind of understood what it depicted but now I have a better grasp. Really appreciate the work you put in to keep consumers informed.

[@davidbetancourt4028]

Wow. Just wow. How did I miss this? Why did I not watch this when it must've originally popped in my recommended months ago? This is ... just a lot to take in. With the way my brain works, I'm going to have to watch this at least 2-3 times more for it all to sink in. This is an amazing video. Thank you for publishing it.

[@shariarrahman7562]

Love this. Would love another video that expands on this when it comes to a monitor that has proper multi zone HDR and a monitor that has proper backlight strobing. I remember watching a few of your videos on such monitors exhibiting some weird characteristics and it would be nice to understand why that is the case and what that means for the end user.

[@HeadBassVTEC]

great job explaining these Tim, was easy to understand 👍 would be interesting to see how the monitor testing look like to fully understand how much effort you put into these reviews

[@avikdas9649]

Thank you... I specifically asked for it in the last video...❤

[@ayavankerkhove8394]

Very well explained. I enjoyed all of it. I learned a lot!

[@DelawareBrooks]

Fantastic info. I was under the impression you were measuring sections of the screen independently so that say bottom left was slower than top right. Glad to be corrected XD

[@almostmatt1tas]

Cheers Tim, very informative and you make it very easy to understand.

[@St1ckl3r]

Impressive video, Hopefully you will make more because you are one of the best teacher I ever seen.

[@Stalast.]

Excellent teaching, Tim! I'm feeling very educated after that one.

[@johpur]

Thank you for this one, it was very clear and understandable 👍

[@rc2276]

You and your team are amazing. You put in so much time and effort.

[@duckysyn]

Thank you for breaking this down it's nice to see what the numbers actually are :3

[@MistyKathrine]

I would like to see you guys do more reviews on OLED smart tvs. I use LG C1 for my main PC monitor which I bought largely because of your review. In your charts the C1 was green over pretty much the entire chart, and I absolutely love it. Most smart tv reviews don't really focus on it's PC and gaming uses, and with smart TVs become more and more popular as monitors, I would definitely like to see more reviews that really focus on the PC use and gaming aspects of them.

[@abheekgulati8551]

This was excellent, thanks so much Tim!

[@casualplayerschannel1432]

yes tim i am interested! thank you for your videos

[@berndkemmereit8252]

awsome, I was wondering what these actualy mean when watching the Monitor reviews, thanks for the education

[@Jeffrey_Wong]

Thanks for making this kind of content even if it doesn't rake in views. I found this explanation really helpful.

[@ReaperHackz]

im loving these videos way more info im so glad you made this channel.

[@josipreponj1928]

Great video as always Tim. Next up similar video on how input lag is measured? I feel like we need more information on this.

[@Sw3d1shGuy]

Great video! I would love to see what the monitor looks like in use compared to the charts. The overshoot, slow response times etc.

[@hkalisvaart]

Very nice reference video for your reviews.

[@Sharleee]

Thank You a lot Tim for that video. I've decided to buy my next monitor based on Your recomendation since I started watching HU chanel despite the fact i didn't understand some of the things You were talking about. This helps me better understand and enjoy this content.

[@sootynukkels8366]

Loving this content and channel. Much love from Indiana, USA.

[@trickyrat483]

Brilliant stuff. Thanks, Tim.

[@Komirowsky]

Grate methodology. I wondered about why you test only grayscale not color ones, but you explained it.

[@vator_rs]

Yes! Thanks for explaining this!

[@lowkey_commenter]

Thank you for uploading this.

[@rebuiltHK47]

Thanks for this. It was very informative.

[@cj_zak1681]

solid work, brilliant stuff 🙂

[@LEGnewTube]

Super helpful. Thank you so much!

[@dylvmo]

Thank you very much. You explain very clearly.

[@ahmednofal3865]

I wish to see how these theoretical numbers should look like in the real world especially the overshoot

[@HitMarkersAreFun]

02:41 argh!!!! Tim, you had me rubbing that spot on my monitor until a few seconds later when you started moving the mouse doh

[@OKHermit]

best video ever on monitors . well explained

[@raytunes64]

I would love to see a more in-depth about HDR in PC games, how almost all of the screens are 400nits instead of the usual 1000 and how is HDR on Windows besides gaming. Also, how does Auto HDR on Windows 10 and 11 actually work? And is it a good thing to turn it on? And also, which is better in a monitor with 400 nits, HDR on or better SDR?

[@Kazyek]

From that explanation, "regular" response time feels pretty pointless and "total response time" feels much more useful, so it would be nice to also have stats on the refresh compliance from total response time (which we can't calculate ourselves because we don't have the heatmap, only the averages)

[@rodrigoferreiramaciel4815]

not actually pointless because even though the actual time from the transition is bigger, the change in values isn't so much, so when we look at the regular response time it represents most of the difference between the rpg values. For example, going from 0 to 255 we may take 10ms but u can actually go from 10 to 240 in 5 ms and that excessive value is not as perceptible as it doesnt actually represent a big change in the colors, even tho it does take its time to happen. Kinda like thinking that transitioning from 0 to 10 is much more perceptible than from 245 to 255. it has to do with how much of the change in value actually represents a change in the color that we are seeing That's what I think makes sense at least, but I'm not any professional om the topic by any means

[@bardacuda82]

This video was great. I'll definitely be paying much more attention to the CD heatmap in your awesome monitor reviews. I ignored it before because I didn't understand it, but now I see it's the most important of the 3 in a lot of ways. Keep up the great work both of you, behind-desk Tim and voice-over Tim! Do you have any recommendations like the HP 27" in the 1440p 144-165Hz range for people that live in Aussieland or Canadia? I can't seem to find that one here for a reasonable price, but would be nice to know what's a good monitor in the $300-$400 CAD range.

[@TestarossaF110]

Very good video, thank you very much!

[@mohamedsherif2964]

Thanks a lot for explanation and videos

[@SerKBer]

Man this is so good. Thank you so much. Next one about color performance :D

[@mind1296]

Great, informative video. You could add a heatmap line over the bottom tables so viewers can see easily how good/bad is a value is on a scale.

[@julfy_god]

excelent video, thanks for that content!

[@anomalous8775]

Amazing job guys, thanks for the detailed explanation. Any change of reviewing Huawei GT 34? Thank you in advance. ♥

[@Kevangelo14]

I just realized how much of a great math teacher you'd be if you were not on the monitor scene Tim!

[@MJS-lk2ej]

I mean does have an engineering degree (not that that's the bee all and end all of teaching, but it certainly doesn't hurt)

[@Aggnog]

Are the response times and other values generally consistent across the displays, or is there in some cases a difference between their physical location, e.g. the middle or sides/corners of the display? I suppose if there was it would be averaged out in the charts, but I'm just curious.

[@Neeb1337]

While I am a math major, so I should just translate these charts effortlessly, the explination was very useful. Thank you!! =)

[@JohnDuthie]

Very informative and helps my understanding of the graphs immensely. I wonder why fall times tend to be slower than rise? Do they run some negative voltage to bring down the LED? Is that even possible?

[@JuanGarcia-lh1gv]

Excellent video! Can I make make similar charts at home with a colorimeter? What kind of hardware and software do you use?

[@High_Alpha]

Monitor Stats 101, thanks Prof. Tim!

[@miranda6717]

Informative

[@johnnyc.9488]

Great content! A question on overshoot : How are you measuring overshoot on the 255 values? For example here, transition 204 to 255 and 230 to 255 gives you a non-zero value for overshoot. How would you measure an overshoot on a pure white value - if 255 your maximum?

[@ibbles]

Not Tim, but I can make a guess on what's going on. The sensor used is a light sensor and it doesn't care about RGB values, signal bit depths, and all that other monitor input stuff. It only cares about light levels. Photons. Trillions (I think) of them. The monitor produces a certain amount of light for a particular RGB input signal and the amount of light produced is measured by the sensor. Increase the brightness on the monitor and you will see an increase in the light produced for the same RGB values, even for 255. The purpose of monitor calibration is to tune the "machinery" in the monitor to produce the correct amount of output light for a particular RGB input value. So in short the 0.255 values only exist on the input side of the monitor, the output can really be in whatever nit/lumen/candela (not sure on the unit here) range.

[@oussamabouyarmane2166]

@@ibbles basically saying the sensor has a bigger detection range than the monitor's output range

[@ilpastor]

Thank you!

[@TestarossaF110]

Are you guys gonna do the LG 42 C2? Was waiting for it and expected it to release around now (almost exactly a year ago the C1 was uploaded).

[@mongotrip9999]

great content!

[@MuzdokOfficial]

Excellent!

[@ramie0n]

I don't know what to say, but you are really making a content of GEMS

[@igavinwood]

Brilliant Tim (pun intended). Nice breakdown of what we're seeing in your reviews. I have a question. Are colour, brightness or response times changed by the GPU, connection or cables used?

[@itsprudhvirazz]

Absolutely Awesome. can you also please add a similar videos for color related things? i don't really understand what over saturation , gamma etc mean.

[@whismerhillgaming]

Thanks, great video, even though I follow hardware unboxed since the transition to Cumulative Deviation, so I'm aware of a lot of these already some values still escaped me like average dark level or refresh compliance or inverse ghosting rate

[@Krazie-Ivan]

I finally (!) fully understand these charts & can make purchasing decisi... ...wait, nope, lost it. I'll still have to listen to your summaries & look at your comparison charts, lol, but maybe I'm a step closer to being less lost next time?

[@pabo6708]

Do you guys have an ETA of when the new MSI MEG 342c QDOLED is coming out? Great vids btw, very informative.

[@SaadKidwai]

Yes, thank you!