I think you meant 30FPS.
Digital Foundry Article Technical Discussion Archive [2010]
- Thread starter liolio
- Start date
-
- Tags
- digital foundry
- Status
That was painfully obvious. Good thing Logitechs better mices have had option to set polling rate since many years ago. Got it set to 500Hz, 2000DPI for my G5 v2 lasermosue.
Eh,. I'm still doubting people can tell 33 or 66 ms of difference., probably even 166. Some HDTV's probably add 100 ms of lag, but nobody in general notices, unless I think they've been told too. If you buy a new Samsung TV with tons of image processing and lag, I dont think most people even know. Also game modes on TV often have no real effectiveness.
I found this http://www.humanbenchmark.com/tests/reactiontime/index.php
I also remember hearing that human reaction time is about a quarter second, or 250 ms. It seems odd to me then that less lag than that could really have much effect.
It's because it's cumulative. You get a visual signal. At this point you want to respond. From wanting to respond to actually responding physically, say, by pressing a button, takes for instance 215ms (the median score here: http://www.humanbenchmark.com/tests/reactiontime/index.php).
This is the baseline. Any lag from a game is added to this. So a game that has 200ms + lag is approaching half a second between something happening on screen and your ability to react to it. That's a lot, and you can see how minimising the game part to as low as possible a response time is beneficial to your experience.
Of course there are mitigating factors, such as the human ability to anticipate. How effective this can be greatly depends on the game you are playing and how it is presented. Say that you're playing table tennis - you usually don't even watch your bat as you hit the ball, you just follow the ball's trajectory and while you are moving your arm / hand / wrist to hit it, you are already looking at where you want the ball to go.
Now if there is lag, the ball will be further along its path when you intend to hit it. As a response, you can almost automatically start hitting the ball sooner, as if the ball was further in its trajectory than you were visually guessing it to be. In 2D, it is hard to know exactly anyway, so you will barely notice that this is a little off from reality. As long as the speed with which the ball travels allows for enough time to make this adjustment, you're fine.
Interaction isn't just about reaction. The best example is computer music. You know exactly what notes you are to play when, with no reactionary response at all, but if there's a difference of even only 20 ms or so, it sounds wrong, and these errors as you try to correct to the rythmn of what you hear can accumulate to throw your whole timing off. It's the discrepency between intending an action and when it occurs that is offputting. In a shooter a small delay between pulling the trigger and the gun firing doesn't matter that much, so I don't notice lag. In FIFA where you have to hold the button for a long pass, it can become extremely frustrating when you want to pass NOW but have to wait half a second by which point the attacker is upon you. Likewise in a timing based game, you have to adjust your timing so isntead of recognising when is the right time to jump, you have to recognise the right time to press the button so that the character jumps at the right moment. Smaller lag means less adjustment and a more natural game.
An app that compares reactions to timing would be interesting. Have a reaction challenge where the user has to press a button according to a cue and time the delay. Then have a challenge where things are happening at a given speed, such as objects scrolling towards a crosshair, and the user ahving to hit the button when the object is centered under the crosshair and time how many ms they are from the ideal. Might be better not to do that with visual position as that'll be influenced by spacial perception. Maybe have the object move in steps, each step taking one beat, so it'll be obvious when under the crosshair will happen.
Edit: You could take this a step further and have an adjustable lag so users can try different lags and see how it affects them. Maybe even have a simple game with adjustable lag to see what people can and can't get used to.
Edit: You could take this a step further and have an adjustable lag so users can try different lags and see how it affects them. Maybe even have a simple game with adjustable lag to see what people can and can't get used to.
dragonelite
Veteran
I know some folks used this webgame to test reaction speed.
http://www.bbc.co.uk/science/humanbody/sleep/sheep/reaction_version5.swf
Silent_Buddha
Legend
Add to that the ability to configure the lag between button press and system acknowledging button press (to simulate controller lag).
That way you can establish a baseline for Display lag + player reaction lag + default baseline controller lag. Then introduce the controller lag variable to see what differences in controller lag are perceptable to a person.
Of course, it'll be essential to display something on screen to indicate it accepting the button press. As we won't have things that would be noticeable in games. For example in games sensitive to controller lag, you'd have things like timing and executing Blocks (fighting game example), or jumps (platformer example), or shooting using a single frame of reference (twitch FPS example), etc.
Even then I'm not sure whether that will tell the whole story as you won't be able to check potential display variability (amount of information available in 30 frames a second versus 60 frames a second) which can greatly affect whether you successfully make a block in a fighting game, or notice a guy in a shooter, or jump at the exact pixel required for a platformer.
Regards,
SB
GT5 article is up....Nicely put article overall.
http://www.eurogamer.net/articles/digitalfoundry-gran-turismo-5-tech-analysis
Though I'd like to mention something, I feel that the game runs quite a bit smoother in other camera angles (most notable the hood cam & chase cam) as compared to the bumper cam. And my suspicion was confirmed after I saw the LoT framerate analysis where I saw the game drop frames whenever the person went into the bumper cam view. This is quite surprising considering that you don't have your car in view when you use the bumper cam.
EDIT: I see that the article missed giving the three display settings normal,flicker reduction & sharper a mention. Won't blame him though, the article is mammoth as it is.
http://www.eurogamer.net/articles/digitalfoundry-gran-turismo-5-tech-analysis
Though I'd like to mention something, I feel that the game runs quite a bit smoother in other camera angles (most notable the hood cam & chase cam) as compared to the bumper cam. And my suspicion was confirmed after I saw the LoT framerate analysis where I saw the game drop frames whenever the person went into the bumper cam view. This is quite surprising considering that you don't have your car in view when you use the bumper cam.
EDIT: I see that the article missed giving the three display settings normal,flicker reduction & sharper a mention. Won't blame him though, the article is mammoth as it is.
Last edited by a moderator:
Billy Idol
Legend
congratulations grandmaster!! what a great GT5 article and how much stuff you have put in - sensational!
3D analysis is very interesting as well - so you definitively recommend to increase parallax?
Mr. Leadbetter, I salute you!
3D analysis is very interesting as well - so you definitively recommend to increase parallax?
Mr. Leadbetter, I salute you!
Some standard cars actually look pretty good:
Anything below this quality level should've been scrapped or reworked on, cause some of them look hideous (there are much worse examples than this one). The thing is that most of them have this very low resolution textures with terribly jagged edges, like the window frame & arch in the 2nd picture.
Anything below this quality level should've been scrapped or reworked on, cause some of them look hideous (there are much worse examples than this one). The thing is that most of them have this very low resolution textures with terribly jagged edges, like the window frame & arch in the 2nd picture.
I'm convinced now this is a bug. In that picture you have a completely different quality in the surface facing the camera versus the back of the car. You've got the wheel arch flairs showing the polygons, and then these chunky pixelated arch rims. It's as if a different LOD texture is being applied to the model, with the texture for the car viewed from 100 feet away is being applied to the car up close. Explained that way, it actually makes sense, and explains the widely differing results and how such crap looking cars could get past PD's QA. If so, I expect this'll get patched. The DF examples show the standard cars still look pretty fabulous with no trace of this mess.
Yeah, haven't read all of it yet, because actually the game just got a new 1.02 patch and I've been testing the new car restriction options which are great. But he mentions he's still not covering everything.
Yep great article.
I think there are a few things I hoped they would touch a bit more.
For example I hoped to see more coverage on the damage modeling since it appears to have different levels of damage, varying on car or perhaps an event.
There are images floating around with cars having their bonnets completely wrecked. It has also been reported that damage affects handling.
The example in the DF article covered none. The car didnt seem to be affected by crashes, and the damage was minimal at best.
So I think there is still room to cover for that.
Another was headtracking. I hoped to see a video of that. They didnt specify what angles does the tracking recognize and they didnt show any video either of that. Is it horizontal tracking only? Does it zoom in and out?
Lastly I was hoping to read more on the AI behavior.
ShadowRunner
Veteran
I tried the head tracking and its pretty great imo. Uses the angle your head is tilted for looking left and right but also moving your head on the x/y is represented on screen so you can peek over the bonnet more as you come over a hill for instance.
Now that's what we were wanting!
no reference to Temporal AA??
Temporal AA is invisible on screenshot because is a temporal AA without blend but it's here! eye is better than HD capture card
standard mode is with temporal AAx2 (create flicker) and flicker reduction mode is without temporal AA
Last edited by a moderator:
- Status
Similar threads
