Games & input lag.

Yes I'm aware of the Gamasutra article, but in all fairness to Mick West (without whose work I doubt we'd be doing this at all), using the latency controller board gives more accurate results than his method. There's a bit more of a story about Halo 3 related to our Natal article, which I'll be mentioning in the DF article.

Mick is correct in that firing and jumping are two entirely different measurements though. Which you would expect really - the act of jumping physically takes longer than the act of pulling a trigger. This is part of what I'm trying to say vis a vis Killzone 2. Latency in controller response is often deliberate, but I don't think there's any excuse for adding deliberate lag on pulling a trigger in a shooter (and that's where I get a 166ms measurement for KZ2 using Mick's technique).

If you are gettint 100ms on halo3 and 166ms on kz2, you are doing something wrong.
Please upload your files on sendspace or something so I can prove you wrong :)
 
Ben Heck's tech only works with the wireless controller, which means I can only use it for PC and Xbox 360. I do have an adaptor for running the 360's wired pad with the PS3, but Ben just couldn't get the wired controller working with the latency board.

It's a real pain because it would be really useful in the double-buffer vs triple-buffer debate. Perhaps it is something we could replicate on PC instead with other games.

I've been conducting FPS analysis on my captured clips, then super-imposing the camera shots over the captures. The idea is to get some idea of frame rate and its impact on response.

The biggest eye-opener thus far has been that my aged Dell 2405FPW introduces three frames of lag at 720p (ouch!) and two at 1080p. Also interesting is that the accepted measurement of Halo 3 lag being 133ms appears to be somewhat exaggerated. I am reading it at closer to 100ms.

You should be able to do that by using D3DOverrider just fine. Make sure you're not using SLI though, as triple buffering won't work properly. I'd also recommend messing with the "render ahead" setting in your GPU drivers, as this is often said to be the cause of mouse lag in a lot of PC games. It'd be nice to see the quantitative effects of that setting.

On the PC side, I'd recommend conducting tests both with and without SLI enabled as that could have the potential to introduce latency. Like I suggested before using UE3 games and limiting your framerate to 30fps and comparing it to a 60fps test could be interesting to see the direct affects of framerate on input lag, without introducing any other potential variables.
 
I don't think we really know if the couple extra frames of delay demonstrated in Killzone 2 is actually input delay/latency, or whether it's more an animation or general code-related issue.

Because I've found that when I fire a weapon after not firing a weapon after a while, I do notice a very noticeable slight delay in the weapon before it responds. But then, after the initial round fired, I test the firing delay again and the gun surprisingly feels responsive... i.e., there's very little delay in subsequent button taps. It seems like the guns take a shot or two to "warm up" before becoming responsive.

I'm using a CRT, so basically no added input delay on my end whatsoever.


Digital Foundry: Killzone 2 Latency Test
 
grandmaster said:
100ms, which is effectively the lowest possible lag for a 30fps game
Come again? The ideal input-lag in a 30fps game would fall inside 50-83ms(so 60+ average). Just because most modern games add an extra frame or more, to exploit easy paralelism, it doesn't make it the "lowest possible" scenario.
 
My perception was completely wrong about Geometry wars 2, I guess it's a mix of frustation and the little replay when you die that peaced me... I'm not good enough... Like with guitar it's always easier to question the tools (ie the guitar) than your skills... Anyway nice article :)
 
But it is worth pointing out that many people criticised Killzone's controls to the point where Guerrilla Games issued a patch. I have measured Killzone lag at at a baseline of 166ms which will rise still further when the frame rate drops. I can't use my new technique for the measurement, but I did enough tests using the old technique to establish that this is the real result.

I wonder if you can really call it a patch, as it's a setting you can switch on or off in the options screen. And that also means you can compare the old situation with the new easily. My personal impression is that they've focussed a lot on realism, and found that too many people complained that it wasn't like CoD4 or Resistance 1, and so they tweaked the configuration to be more 'twitchy' by default. This may just be a matter of removing a 'weight' factor from the aiming as anything else (that kind of difference is definitely factored into this game) - the guys at GG seem to be near anal-retentive about realism on those aspects.

Of course it could also be that they just reduced input lag somehow and allow you to switch between the old and the new. But it will probably be really hard to do a proper comparison here. You'd have to test most of the different guns, with both the new setting on and off, to see what's really changed between the two.

Did you mention which of these two settings you actually used by the way? Because if you had the game before that update, it doesn't enable this new setting automatically. Only people who play it for the first time post that update get this setting on by default.
 
Does anyone know where you can get input lag measurements for your set?

I have a Sony Bravia XBR4 and was wondering how much input lag it introduces to the equation.
 
Does anyone know where you can get input lag measurements for your set?

I have a Sony Bravia XBR4 and was wondering how much input lag it introduces to the equation.

I've been doing a lot of research on this lag issue recently, as I got a new HDTV and then was told on a forum that it might be laggy for games. Concerned, I set out on a fact finding mission.

The best resource I found seems to be a very large thread on Shoryuken.com forums. The Streetfighter guys, so you know they would know lag. There's also some threads on AVS forums.

But to make a long story short, there seems to be very very little hard info or data on LCD lag. Nor any easy way for an average joe to test their own set. Virtually every test has been deemed unreliable by some members or other at Shoryuken. Their are lag tests in many Guitar Hero and Rock Band games built in. Once again though, many deem them innaccurate. There is the method running a timer on a CRT and LCD at the same time to see how much the LCD lags. Again, many say this is not accurate either. There's a method of filiming with a 60 FPS+ camera, this seems to be deemed the most accurate, but I doubt most people can do this. There also seems to be no clear recommendations on which HDTV models have low lag. There's just amazingly little hard info out there. Another issue is, many times pinning down what particular panel a LCD uses is nearly impossible. Sometimes the panel used may vary on the exact same model of TV.

Some things said to help, hooking your console up by VGA (easiest on 360 which has a VGA cable), will often set your TV to PC mode, which is basically supposed to be a no processing at all mode or close to it (you basically want your TV to process the image as little as possible to reduce lag). But as always, this can vary by set and nobody has any hard info on it. If your TV has in any way a PC mode via HDMI as well, you may want to try to use it. So basically PC mode/VGA MAY help, but it may not, depending on model (and good luck trying to get info on any particular model). That's the frustrating thing with so many lag solutions. Also, many people complain that the picture looks like crap on PC mode, which I guess is supposed to be the idea.

Game modes are good, and surely reduce lag, but they apparently do not turn off all image processing, and again, results vary by manufacturer, etc.

PC monitors with TN panels probably have the least lag, theoretically. But, who wants to game on a tiny screen imo. This might be a good solution for a SF tourney, but I dont think the average joe wants to give up his 40+ HDTV just to reduce a little lag.
 
Now the real reason I bumped this thread, I came across some very interesting info in the Shoryuken thread, they said the biggest factor in reducing lag is to run the console at the TV native res. The scaler in the console is said to be much faster than the one in the TV. Feeding the TV a non native res is supposedly a sure way to invite lag.

Now, it appears the PS3 has problems outputting 1080P due to it's scaler issues, and it depends on the game. So apparently, the PS3 could have an endemic lag problem on 1080P native LCD's if I'm understanding this right. On the Shoryuken forums for example, it was discovered that the PS3 version of SF4 has an extra frame of "inherent" lag vs the 360 (I believe 4 frames on PS3 vs 3 on 360) which was possibly considered due to this 1080P scaling thing. I thought this was interesting and I thought Mr Leadbetter might be interested in such things given his past lag tests.

Am I right or way off on this? I dont own a PS3 anymore so I'm not sure what output options it has. Does it have 1080i on all games? If so, would 1080i (PS3) to 1080P (TV) mean no TV scaler lag?

I guess another option touted on Shoryuken for all gamers, was to simply stick to 720P TV's. Of course though, the big/nice sets arent in 1080P, though I think they can go to 37". (Yes I'm aware it's really 1366X, but I think this would still reduce lag I guess).

Also, I recall a Sony TRC to make all games have a render mode at 960X1080, then scaled the 960 to 1920 via the working half of the scaler, being discussed on B3D. Did that ever happen and is it currently in place?
 
If a PS3 game supports 1080i, then it supports 1080p. And you are right in that running at native resolution of the screen knocks off a frame or two of lag.

My Dell 2405FPW (1920*1200) is a pretty laggy screen. If I feed it 720p, the display eats up three frames of lag. Alternatively, if I feed it 1080p, it is "only" two.

I will be returning to input lag, and hopefully getting some PS3 measurements that match the accuracy of my 360 ones in the next few weeks.
 
Also, I recall a Sony TRC to make all games have a render mode at 960X1080, then scaled the 960 to 1920 via the working half of the scaler, being discussed on B3D. Did that ever happen and is it currently in place?

Doubt it, most recently Dragon Age: Origins didn't support 1080p/i modes of any kind.

And if the game is sub-HD (like Ghostbusters at 960x540), a 960x1080 mode would be impossibly expensive.
 
I guess another option touted on Shoryuken for all gamers, was to simply stick to 720P TV's. Of course though, the big/nice sets arent in 1080P, though I think they can go to 37". (Yes I'm aware it's really 1366X, but I think this would still reduce lag I guess).

Or commercial grade LCD's that are used for displaying info or whatnot in malls etc. Those have no post processing implemented whatsoever, but they naturally don't look as good as mid-range to high end sets.

You have to keep in mind that the tests were done by competitive fighter players who are extremely picky, sometimes needlessly if you're looking at 2-3 frames. Sometimes, there is the issue of placebo effects but there are some out there who are really that sensitive. IIRC, I was told that SF Alpha Anthology on the PS2 has 3 frames of lag.

With more frame sensitive fighters like Tekken (especially Tekken), and VF, even using a wireless controller shows some lag if you're sensitive enough. Just frame moves can be tricky to pull with any lag. Stick to game mode, which still has some post processing on that you can turn off manually, and get some long USB cables and you should be fine.

I actually prefer the look of PC modes in general since it's the purest signal you can get. I've only got my Dreamcast hooked up that way, and even those old games still look pretty good IMO.
 
Or commercial grade LCD's that are used for displaying info or whatnot in malls etc. Those have no post processing implemented whatsoever, but they naturally don't look as good as mid-range to high end sets.

You have to keep in mind that the tests were done by competitive fighter players who are extremely picky, sometimes needlessly if you're looking at 2-3 frames. Sometimes, there is the issue of placebo effects but there are some out there who are really that sensitive. IIRC, I was told that SF Alpha Anthology on the PS2 has 3 frames of lag.

With more frame sensitive fighters like Tekken (especially Tekken), and VF, even using a wireless controller shows some lag if you're sensitive enough. Just frame moves can be tricky to pull with any lag. Stick to game mode, which still has some post processing on that you can turn off manually, and get some long USB cables and you should be fine.

I actually prefer the look of PC modes in general since it's the purest signal you can get. I've only got my Dreamcast hooked up that way, and even those old games still look pretty good IMO.


The TV's for malls and such are called digital signage, and as much as I can tell at the Shoryuken thread, it pretty much turned out to be a myth that they have any lower lag. In fact they are optimized for high viewing angles.

Also, to the wireless controllers, I believe those were tested as having very very little or even claimed zero added lag versus a wired controller. At least that much is comforting. http://www.shoryuken.com/showthread.php?t=165659&p=5773657&viewfull=1#post5773657
 
Last edited by a moderator:
Also, to the wireless controllers, I believe those were tested as having very very little or even claimed zero added lag versus a wired controller. At least that much is comforting. http://www.shoryuken.com/showthread.php?t=165659&p=5773657&viewfull=1#post5773657
Seems from there a guy who creates arcade sticks for lowest-possible lag found wireless is no laggier than wired. And this is a very honest assessment because he enters the thread saying wireless is rubbish and anyone thinking it was the equal of wired was a moron!
 
I found no measurable difference between a 360 wired controller and the wireless one, and neither did Mick West in the original Gamasutra feature. Nice to see that the ultra-hardcore players are reaching the same conclusions though.
 
Guess I'll give wireless a try then. I've yet to play any fighter with a wireless controller or stick (pad Tekken player here)

I forgot to comment on the issue or supposed advantage of having a 720p native set (they're not since all of them are 1366x768), the issue still stands and depends from one manufacturer to another. In fact, my old AQUOS had some noticeable lag than my current XBR4.

Incidentally, the best settings I came up with after consulting many people and doing tests for months were based on game mode. Turning off all post processing is VERY beneficial in some higher end sets.
 
Back
Top