Discrepancies with input lag measurement

P

philm2004

n00b
Joined
Jan 30, 2013
Messages
15
OK im trying to better understand input lag and response times when dealing with displays.

From my understanding:
Input lag is the time it takes for your graphics card to send a new image to the monitor and the time it takes for the monitor to get that image ready to display on the screen.

Response time is the time taken for the display to actually change to a new image once it has been told to do so.

Overall display lag is the input lag + the response time.

What gets confusing however is when certain reviews seem to contain conflicting information.

For the Dell UltraSharp U2713HM

TFT central measures the average input lag for this monitor as 22.2 ms with a maximum of 28ms.

From this i would assume to get the overall display lag you would add on top of that figure the quoted response time of GtG 8ms.

However they go on to say "This is the overall lag of the image compared with a CRT, taking into account signal processing delay and pixel response times. "

I am confused by this statement as i thought the signal processing delay and input lag were the same thing, can anybody explain to me why this isnt the case?
 
it's just the different terminology being used thats all. The term "input lag" has been around for years and years and was always used to explain the difference between the output of an LCD screen compared with a CRT screen. the older methods for measuring this difference in the appearance of both screens involved simple stopwatch methods which have since been proved to be somewhat inaccurate and can vary a lot. they were always trying to capture the difference in appearance between the two screens so in that sense, the original definition for the term "input lag" accounted for any signal processing + response time which must be considered.

It's only more recently that some sites have started to talk about the lag in two different forms when new more advanced processes and testing methods have come into play. some now refer to the "display lag" as the overall difference in images (ie what used to be simply called "input lag") and this is made up of their new definition for input lag which is the signal processing and then the response time is applied on top of that

The TFTCentral reviews talk about input lag in the old sense of the word, presumably because it is more widely used and commonly understood as the term. This is being based on the overall "display lag" though accounting for signal processing and response time. they are using SMTT which is a method measuring the difference ultimately in images between a TFT and CRT and so it's that overall display lag which is being measured

the only site ive seen who talk about input lag in a different way is Prad.de, who talk about input lag as the signal processing and then apply a response time on top of that to give the overall display lag.

so if you compare the Prad and TFTCentral reviews for the U2713Hm you will see they do match up in so much as TFTCentral has an average display lag (still referred to as input lag for traditionalists) of 22.2ms and Prad.de have an overall display lag of 22.6ms.

the signal processing is the element which gives the feel of lag in movements and in games, but the overall difference in displayed image is also important to consider. when comparing the two measurements you need to compare TFTCentral with the overall Display Lag of Prad. its just that Prad then have a further breakdown within that to tell you what part of that was signal processing
 
Thanks for the reply cleared quite a few things up.

So im assuming then that when tft central talk about input lag which we now know is really overall display lag. Do they just used the quoted GtG response times?

On prad.de for the Dell UltraSharp U2713HM on top of the signal processing lag they add 7.1 ms which is half of an averaged response time.

If tftcentral use the quoted response time i can see how their measurement would be a bit more inaccurate because prad.de independently measures the GtG response time as 10.8 ms.
 
No they don't make any kind of assumptions based on response time. The SMTT method is entirely based on what is displayed on the TFT compared with the CRT and so automatically accounts for the response time element anyway. So that will always be the overall display lag in total

You could in theory make an estimation as to the signal processing element of that overall lag if you knew accurately what the response time part was, and work backwards from there. Taking the quoted spec might be a little out though as they aren't always reliable so you'd only get an estimation doing it that way.
 
ok so for the dell UltraSharp U2713HM

The figures for the overall display lag between prad.de and tft central are almost identical as you said 22.2ms and 22.6 ms respectively.

However another review site called anandtech measured the input lag of the dell as 18.5 ms ( i dont know how they came to that figure) , they measureed the response time as 11ms and so they inferred that the Overall what they call Processing lag was 29.5 ms.
I'm guessing again that what they call proccesing lag is really overall display lag.
So ive found a variation of 7.3ms in the different reviews of overall display lag.


However for another monitor say the Samsung S27A850D the figures do not correspond that well with what tft central are saying input lag (overall display lag) - 28ms
and prad.de saying total lag as - 19.3 ms. Anandtech does not have any figures for this monitor.

Thats 8.7ms difference, is this just the difference then between using an oscilloscope and using the SMTT program.
 
Not sure what method anandtech are using to measure lag or response time but there are in depth articles and studies on both TFT central and Prad about input lag and testing methods. It's probably a method being used with more error in it at Anandtech. Prad also don't have any explanation as to the method being used anywhere and it's not the same as that tested in their in depth article as those studies were conducted by a third party. So that's a bit of an unknown too and I guess we are just supposed to trust in the word "oscilloscope" :)

Theres many different techniques being used so I'd try and take a view from them all and somewhere in and amongst that you're likely to arrive at the truth. A combination of overall display lag results from a reliable testing technique (SMTT and not old stopwatch methods) and oscilloscope based tests should hopefully give you a decent view. There's no "standard" for input lag testing and many methods aren't explained in full / at all either so it's a bit hard to know. Ultimately I suppose also it's always going to be down to the individual as to whether it affects them so really it can only be a guideline
 
OK thanks for your responses very helpful.

On TFT central they reviewed a monitor called the HP ZR2740w

They came to a figure of 3.6ms for the input lag ( overall display lag) , but im confused as to how this could be the case if the quoted GtG response time is a whopping 12ms. How can the overall display lag be almost a quarter of the quoted response time?

Neither prad or anandtech have done reviews of this monitor so i cant do any comparisons.

I've never used SMTT before or the stopwatch method since i don't have an accurate camera or an old crt monitor to compare against. (wish i hadn't thrown it out now)

But from my understanding of SMTT surely a minimum value of overall display lag must be the response time.
 
yes, but its entirely possible that the response time for the specific colour transition used in SMTT is lower and G2G figures really do vary across the range. Its probably an indication of next to no signal processing lag (which would make sense given there's no scaler or OSD even) and the rest of the lag is down to response time.

I don't think you need to account for the whole response time either, just enough of it to display the next figure in the counter, so as the pixel begins to light up, it only needs to have gone as far as making it partially visible, not necessarily its full transition / full brightness. so in this way i guess it keeps the display lag down a little bit closer to the signal processing lag. i believe Prad actually take half of the average response time as well when they add it on to their signal processing lag. they don't specifically mention this in the U2713HM review but if you look at other older reviews they do. They probably take half of the average response time as a middle ground approximation for the same reasons as you don't need the full response time added.
 
Infact it turns out that anandtech did actually do a review of the HP ZR2740w.

They get figures of 2.6 ms input lag and a 11.95 ms pixel response time, which they go on to say gives an overall processing lag of 14.55 ms.

So along with the figure of 3.6 ms overall display lag from tft central. It seems there must be some methodology difference considering they are both apparently running the SMTT program.

As quoted from ananadtech for the HP ZR2740w review.
"This is our second review using the SMTT program to test lag,
but this lets us separate the input lag from the pixel response time"

So some how using the SMTT program they have seperated out input lag and pixel response time.

However it seems that tft central do not seem to do this.

So as far as i can tell the reasons for this difference is because possibly tft central don't take the full response time just a very small amount of this and that anandtech seem to include the Total pixel response time. Does that seem correct?

TFT review here : http://www.tftcentral.co.uk/reviews/hp_zr2740w_v2.htm

Anandtech review here: http://www.anandtech.com/show/5636/hp-zr2740w-high-resolution-ips-that-doesnt-break-the-bank/5

I have never used the program , but would love to test it out if i had the right equipment but alas.
 
there's no way to separate the two out using SMTT, it's a simple (well, pretty advanced actually in how it works) stopwatch method so it's all based on what is shown on the CRT and what is shown on the TFT and comparing them. if you look at the screenshots and explanations in the TFT central article you can see more about it. so there's no way to test that separately using SMTT. I've used it (and the old v1 as well) and can tell you thats not possible

perhaps Anandtech had taken the SMTT results to be an interpretation of the signal processing lag and then added on some response time as well (wrongly really as SMTT has an element of the response time built in as already explained)? As far as i know though Anandtech dont have any special method for measuring response time either?! they dont use an oscillscope setup for that so not sure how reliable those response time stats are
 
Ok thanks for your replies again.

I personally think now that anandtech's testing for input lag is wrong.

I havent done an exhaustive comparison but on the few monitors that i have checked out tftcentral and prad.de seem to be pretty close on the lag times, however anandtech's seem to be way off on. An example:

Asus pb278q

tftcentral - average overall display lag of 16.6ms with a maximum of 17ms

A user on another forum ( i believe he may use this one aswell) tested the screen with
SMTT 2.5 and stated an overall display lag of 17ms.
http://wecravegamestoo.com/forums/gadgetry-electronics-discussion/11962-asus-pb278q-review.html

Anandtech quoted 16.7ms of input lag with a measured pixel response time of 11.5 ms to give an overall delay lag as 28.2 ms.

Unfortunately prad.de has not reviewed this monitor.

Do you think its likely they are wrongly adding the pixel response time onto the values which they get off SMTT, because if they are that seems very bad of them to do that.
 
yes that other review of the PB278Q is by our very own NCX and like you say, is pretty much identical to the results from TFTCentral. Both using SMTT 2.5 tool. It looks to me like Anandtech have used SMTT to obtain the lag result of 16.7ms which is then pretty much the same as TFTCentral and NCX's review. but then theyve oddly added on a response time element from an undisclosed source / method
 
It would be nice to hear NCX's view on this since he has actually done the testing.

Do you know of any other websites where they test for input lag? as i can only find the 3 that i have previously mentioned.

I actually own the Asus pb278q and as i play pvp arena in wow and also a bit of BF3 . i am actually interested to know the overall display lag of this monitor, i was playing yesterday and to be honest i couldnt really tell any input lag although sometimes i thought i might have been able to cant be sure though. thats why i'm interested.

i know monitors are not the only source of lag their are others. my old monitor was a tn panel so maybe im judging to harshly but one of the reasons i bought this monitor was becasue of the relatively low values of display lag quoted on tft central. Then i read the review from anandtech and i thought maybe i can tell the input lag afterall .

Sometimes i think my mind is playing tricks on me.
 
if it's not causing you issues day to day i wouldnt worry about it too much. expect NCX will be along to comment on his tests of the PB278Q specifically. i know he's a firm believer of "input lag term = signal processing" :)
 
I've just seen an update on tft central's website dated for 31 january on the Dell U2713H, prad.de hasn't reviewed this monitor and i haven't seen any other reviews that state the overall display lag.

They some how seem to have got their hands on an oscilloscope, and measured the pixel response time. They say and i quote "for the pixels to reach the required brightness" so im guessing this isnt the full response time just the time it takes for your eyes to register the image.

They measure this pixel response time as 4.5 ms , which they then subtract from the overall display lag to get a value for the signal processing lag which is 20.8 ms.

So am i right in saying then that an oscilloscope can only help you detect pixel response times? and can in no way measure the signal processing lag. They only way of doing this must be using a crt comparison method such as using SMTT?
 
It's possible to use it to measure signal processing (see prad full article) but as far as I know that required kit costing >100,000 euros and some very complex methods. That's not being used any more at prad as was a third party investigation. They apparently still measure the signal processing somehow but that's not explained anywhere at all (and they don't seem to respond to emails about it either!)
 
Prad.de is the only site that produces reputable numbers on input lag. If TFTCentral or Anand says one number and Prad says something else, trust Prad. They've established their reputation here with unrivaled depth.

Prad's input lag numbers are pulled from an oscilloscope setup, completely independent of a PC. IIRC it's the time it takes for a black to white transition to reach 10% luminance from the time the signal was sent. The times nearly always fall on or just past discrete frame times, usually indicating 0 or 1 frame of buffering going on in the monitor.

Prad's setup also shows its versatility with pixel response time: they don't need to bother with collections of vague pixperan screen pictures when they can directly show graphs of the monitor's response over time. Overshoot, undershoot, fast or slow response, stepped response, and/or asymmetric response all show up on these graphs for comparison.

philm2004 said:
So am i right in saying then that an oscilloscope can only help you detect pixel response times? and can in no way measure the signal processing lag. They only way of doing this must be using a crt comparison method such as using SMTT?
Click to expand...
No, what you're looking for is information like this:

fhijx0D.jpg


Control signal in green, luminance response in red (Dell U3011). You can read both input lag (the delay before response, ~1 frame for this monitor) and the pixel response (the slope and shape of the rise).
 
Last edited:
Yes, good in theory except that the article linked was written by Thomas Thiemann who doesn't work for Prad and is a third party free lance writer who carried out all of those tests independently. He's also the person who made and designed SMTT v1 and v2. Those extensive tests were carried out with kit he borrowed worth >100,000 euros and was very complicated and difficult to achieve. There is no explanation of what prad now use to measure input lag at all and it's not detailed anywhere. They slso dont seem to respond to emails asking them, i (and others i know) have tried before. You can be pretty sure they don't use that same setup and really a lot of trust is being placed on assumptions and based on the word "oscilloscope" :)

I'm not saying it isn't necessarily accurate, maybe they have found a decent alternative method which is reliable. My point is just that we don't know what method is being used and yet it seems some people take it as absolute gospel
 
Nihilanth99 said:
Prad.de is the only site that produces reputable numbers on input lag. If TFTCentral or Anand says one number and Prad says something else, trust Prad. They've established their reputation here with unrivaled depth.
Click to expand...

Ok even if Prad.de does use the very detailed test method explained here:
http://www.prad.de/en/monitore/specials/inputlag/inputlag-part26.html

(Which in all the monitor reviews there is absolutely no mention of)

In actual fact tft centrals input lag/ display lag numbers are on the whole similar to prad.de's numbers a few examples:

Dell U2713HM

TFT central Average input lag/display lag - 22.2 ms

Prad.de display Lag - 22.6ms
Difference - 0.4 ms


Samsung S27A850D

TFT central Average input lag/display lag - 29 ms

Prad.de display Lag - 19.3 ms
Difference - 9.7ms


Iiyama Prolite G2773HS

TFT central Average input lag/display lag - 12 ms

Prad.de display Lag - 19.8 ms
Difference - 7.8 ms

BenQ XL2420T

TFT central Average input lag/display lag - 13 ms

Prad.de display Lag - 7.9 ms
Difference - 5.1 ms


So the biggest discrepency is with the samsung at 9.7 ms difference but the others aren't to far off.
This topic is really annoying me because i find it hard to believe that monitor manufacturers cant do the very accurate input lags themselves, the equipment is expensive yes, but not to a company like samsung, infact they probably do similar tests but just dont release the data.

This makes it worse with reviewers like anandtech seemingly testing in an incorrect manner, and prad.de not divulging how they get their results.

It seems like consumers are being hoodwinked into believing they have a "fast" monitor when they are looking at the listed response time, many people myself included look at the GtG pixel response times, and think well the smaller the better, however you may have a monitor with a very high signal processing time and therefore have a noticeable display lag.

So the only other option for consumers is to listen to anecdotal evidence from other users, i dont want hear'say i want hard cold numbers!

It makes it all the worse when many people would rate input/display lag very highly on their list of considerations of a new monitor , but have no accurate results to go from.

Just thought id add a link to this thread:
http://hardforum.com/showthread.php?t=1743302

People really care about these numbers but there is so much misinformation around the web , that people are going to be making incorrect assumptions. So far ive seen the same phenomenon called: input lag , dislpay lag , signal processing lag , total lag etc. These terms are being used interchangeably but they all mean different things...
 
Last edited:
The main issue for manufacturers is the trade off between achieving a low lag and providing a decent scaler with decent aspect ratio control most of the time. Hard to keep everyone happy but would be good to see more screens with options to bypass some of that to reduce input lag (like thru mode, game modes etc)
 
Drags said:
Yes, good in theory except that the article linked was written by Thomas Thiemann who doesn't work for Prad and is a third party free lance writer who carried out all of those tests independently. He's also the person who made and designed SMTT v1 and v2. Those extensive tests were carried out with kit he borrowed worth >100,000 euros and was very complicated and difficult to achieve. There is no explanation of what prad now use to measure input lag at all and it's not detailed anywhere. They slso dont seem to respond to emails asking them, i (and others i know) have tried before. You can be pretty sure they don't use that same setup and really a lot of trust is being placed on assumptions and based on the word "oscilloscope" :)

I'm not saying it isn't necessarily accurate, maybe they have found a decent alternative method which is reliable. My point is just that we don't know what method is being used and yet it seems some people take it as absolute gospel
Click to expand...
Point taken!

As philm2004 said, it's a shame that it's so hard for a consumer to get reliable numbers for such an important display property. 9 people out of 10 will assume advertised pixel response time is how to compare 'lag' between screens, and it's hard to blame them for doing that.

Maybe it'll take a manufacturer knowing their product has good latency and advertising it, casting more light onto the subject for more people?
 
i dont mind if reviwes come out with different numbers, as long as the monitors still can be compared. in the example- the 27" Hp is faster than the 27" dell. that is all a consumer needs to know in the end .
 
geok1ng said:
i dont mind if reviwes come out with different numbers, as long as the monitors still can be compared. in the example- the 27" Hp is faster than the 27" dell. that is all a consumer needs to know in the end .
Click to expand...

thats not really the point though, if you bought a car you would want to know its top speed, not just that it was faster than another car and slower than another. And ultimately the information is still wrong.

Anyway what does being "faster" mean. The HP ZR2740 may have a lower processing lag the reason being that it has no osd or any other inputs other than a dvi, but the quoted response time is 12ms GtG and in comparison to other montitors such as the dell u2711 ( which has a quoted GtG time of 6ms) there are significant effects such as motion blur.

So there is no such thing as a monitor being "faster" than another one. You have to look at all the individual parts, and compare things equally.
 
You must log in or register to reply here.
Back
Top