Tame the Mountain: TT Isle of Man Launches March 6 on Consoles

" If you’re playing your games at 30 FPS then the amount of time your display is sitting on one frame is 33.3 milliseconds. This means that when you move your mouse to aim at a target, it will take at least 33.3 milliseconds before you even start to see the cursor move. This delay is halved to 16.65ms at 60 FPS and so on. Of course this is without considering other factors such as networking latency (in multiplayer games) and monitor response time which can add additional latency. " from here : http://www.technologyx.com/featured/understanding-frame-rate-look-truth-behind-30v60-fps/

Dont get me wrong, im not try to make this TT game bad.
It's only because of 30fps in a racing game, which is a no go criteria.

That quote is not really relevant to framerate though. An average person has a response time of around 1/4 of a second (you can test yours here: https://www.humanbenchmark.com/tests/reactiontime/), so 1/30 or 1/60 has no effect what so ever on your ability to react to events on the screen.

And network latency and monitor response time is obviously the same regardless of the framerate.

60 fps provides a smoother experience, that's really all there is to it.
 
It's turned out to be highly addictive.
I hope they add supertwin class after sidecars.
All the included tracks are good. I used career mode to find my way around the game.
Going to concentrate on the TT now, time trialling MD's R6, figured I'm about the same weight as him, much less fit though. Ha.
I've still got a few handling issues with the game, front end washing out, rear letting go for no apparent reason, but then I look at the top times, must be me. :-)
A couple of times I've thought 'Had a pint in that pub' then immediately crashed. Must concentrate.
 
Last edited:
That quote is not really relevant to framerate though. An average person has a response time of around 1/4 of a second (you can test yours here: https://www.humanbenchmark.com/tests/reactiontime/), so 1/30 or 1/60 has no effect what so ever on your ability to react to events on the screen.

And network latency and monitor response time is obviously the same regardless of the framerate.

60 fps provides a smoother experience, that's really all there is to it.
This is ever so reminiscent of the "the human brain can't see more than 24 fps" argument that frothed away up until a couple of years ago, when current VR tech first started to take hold. Now we know that a latency of more than 10 ms is detrimental, and more than 1 ms is noticeable!!

Reaction time is one thing, but the bulk of reaction time is in taking action once you've "recognised" it needs to be taken (or perhaps just reflexing). What's more, it's well known that reactions can be trained - I have been able to get mine down to around 100 ms when I really wanted to - adding 15 - 30% to that figure through frame latency alone is going to be noticeable. On top of all of that, you can read the road ahead, parse several seconds' worth of dynamics information from the vehicle to predict what it will do next, and adjust your controls ahead of time - before you need to "react" - and will have in effect still been in total control based on visual feedback alone. The finer that information, i.e. in relation to the kinds of information we obtain in the real "analogue" world, the better - always.

Getting information into your brain faster is always a good thing, but most of the time you don't actually need to directly react to anything, the visual feedback simply gives you a better sense of the systems you are in control of (something our brains are innately good at, subconsciously), and the more fluid and crisp and otherwise connected that sensation is, the better the overall experience is - especially as far as general action-type experiences are concerned. Even phones target this kind of interactivity in their interfaces today.

Basically it helps to cement the physical actions we give as input with the resulting (visual) output.
So once games can approach that 10 ms input - > output figure, perhaps then we can draw a line under the whole deal. Not that it should stop us enjoying ourselves before then, and that indeed has always been the art of the action video game.
 
rear letting go for no apparent reason

Some of the bumps simply can't be taken at full speed, and some won't allow you to even breath on the brakes when going over. There's one or two places I've learned so far where I need to shift the weight forward or backward on the bike, even though there are no visible cues.
 
This is ever so reminiscent of the "the human brain can't see more than 24 fps" argument that frothed away up until a couple of years ago, when current VR tech first started to take hold. Now we know that a latency of more than 10 ms is detrimental, and more than 1 ms is noticeable!!

It may be noticeable when it's about a framerate drop, but not when we're talking about a constant 30 fps framerate, because the events take place at a steady pace and you're not aware of them until you see them on the screen. So for practical purposes there is no delay.

A framerate drop is something else, because then you have a steady flow of events that is suddenly interrupted.

And I doubt that the 1 ms claim is true. The delay between clicking the mouse and the sound reaching your ears is between 1 and 2 ms. I certainly can't notice that gap.

Reaction time is one thing, but the bulk of reaction time is in taking action once you've "recognised" it needs to be taken (or perhaps just reflexing). What's more, it's well known that reactions can be trained - I have been able to get mine down to around 100 ms when I really wanted to - adding 15 - 30% to that figure through frame latency alone is going to be noticeable.

It's irrelevant what the bulk of reaction time is. When playing a game you see an event and then you need to react to it. It's the whole process that counts. And 30 ms won't be any more noticeable just because you've trained your reaction time down to 100 ms. Especially not when playing at a steady framerate where you're not even aware of any delay.

On top of all of that, you can read the road ahead, parse several seconds' worth of dynamics information from the vehicle to predict what it will do next, and adjust your controls ahead of time - before you need to "react" - and will have in effect still been in total control based on visual feedback alone. The finer that information, i.e. in relation to the kinds of information we obtain in the real "analogue" world, the better - always.

Nope. You can read the road just as well at 30 fps and there are no situations where you need reaction times below even a quarter of a second. Why? Because your brain is able to predict when an event is going to take place, based on what it has seen in the past. You don't actually need to see the precise moment when your bike is at the braking point, because your brain simply takes the preceding events and use them to predict the precise moment when it's going to occur.

60 fps looks smoother and is more pleasing to watch, but it doesn't make you perform better in any way.
 
Last edited:
It may be noticeable when it's about a framerate drop, but not when we're talking about a constant 30 fps framerate, because the events take place at a steady pace and you're not aware of them until you see them on the screen. So for practical purposes there is no delay.

A framerate drop is something else, because then you have a steady flow of events that is suddenly interrupted.

And I doubt that the 1 ms claim is true. The delay between clicking the mouse and the sound reaching your ears is between 1 and 2 ms. I certainly can't notice that gap.



It's irrelevant what the bulk of reaction time is. When playing a game you see an event and then you need to react to it. It's the whole process that counts. And 30 ms won't be any more noticeable just because you've trained your reaction time down to 100 ms. Especially not when playing at a steady framerate where you're not even aware of any delay.



Nope. You can read the road just as well at 30 fps and there are no situations where you need reaction times below even a quarter of a second. Why? Because your brain is able to predict when an event is going to take place, based on what it has seen in the past. You don't actually need to see the precise moment when your bike is at the braking point, because your brain simply takes the preceding events and use them to predict the precise moment when it's going to occur.

60 fps looks smoother and is more pleasing to watch, but it doesn't make you perform better in any way.
What you say would seem to contradict modern research.
If it were only based on reaction times as you state, then a framerate of 8 would suffice (I doubled it because Nyquist) - good luck hitting your braking points. It should be clear that having two points separated by 33 ms contain less information about the (relative) motion of an object (say a rival racer or apex) than three at 16 ms; more dots to join up, more information; better informed and more precise "reactions".

Sound and vision are processed by different parts of the brain, are processed and interpreted differently outright and have different effective "latencies" - for instance a reaction from a sprinter (to a starting gun) of less than 100 ms is considered a false start, but F1 drivers manage that all the time (with lights).

You cannot, for instance, separate two sounds that are about 100 ms or closer together - they sound as one. Flash a light twice 100 ms apart, and you'd see both flashes for sure (it's only 10 Hz). Interestingly, however, interaural time differences are of the order of a fraction of a millisecond and our brains extract useful (if not critical) spatial information from it - modeling it is very useful in VR, and by extension games in general. So do you allow your audio engine to handle such short delays, or not? Clearly, it's not just one cue we rely on with any of our senses.

The very fact is that a whole brace of feedback mechanisms - not just reaction times - all add up to make a difference to the experience (even if that effect is diminished by things like aging), so it should be something we work towards in time, as we have done for many years already. This concept of a discrete stream of slides effectively having zero delay is utterly false, though: its discrete nature is the very basis by which latency is measured.

In short: it's not (just) about reacting, it's about the feeling of connectedness created by several aspects of visual processing. Ask any professional gamer, higher framerates just feel better in action games - this then means they actually tend to "perform" better.

That doesn't mean I think this game being at 30 fps is abysmal in any way, as I already said.
 
Last edited:
Which difficulty are you guys using in career mode (and against AI in general)? The AI seems a little weird. On medium difficulty - assuming I don't crash, or at most once - there will usually be one somewhat competitive driver while the others are far behind.
On the more difficult courses (like sections of the Snaefell course) - where I still tend to crash a couple of times on 10-15 km stretches - I end up on the middle of the board at best, and barely win any money. It seems like on medium difficulty the AI doesn't drive particularly fast (I can often catch up with overtakers after a fall) but they don't seem to make mistakes either.

Maybe I'll just spend a few days practicing in time trial mode first to get to know the tracks and bikes.

On a marginally related note I just loaded up Ride 2 and took a couple of bikes out for a spin. It just feels so dull after TT - sense of speed, sound, bike behaviour. Ride 2 might be a bit more of a sim - and certainly wins in terms of content volume - but definitely not as much of a thrill where it really matters.
 
Which difficulty are you guys using in career mode (and against AI in general)? The AI seems a little weird. On medium difficulty - assuming I don't crash, or at most once - there will usually be one somewhat competitive driver while the others are far behind.
On the more difficult courses (like sections of the Snaefell course) - where I still tend to crash a couple of times on 10-15 km stretches - I end up on the middle of the board at best, and barely win any money. It seems like on medium difficulty the AI doesn't drive particularly fast (I can often catch up with overtakers after a fall) but they don't seem to make mistakes either.
I've been on Expert difficulty from the get go. I absolutely murder the AI on the original tracks. On the IoM however, I ended up 6th or 7th in the Supersport class. If I never had fallen off the bike (twice) I could have made the podium. If we're talking about per section(s), the AI is no competition at all.

On a marginally related note I just loaded up Ride 2 and took a couple of bikes out for a spin. It just feels so dull after TT - sense of speed, sound, bike behaviour. Ride 2 might be a bit more of a sim - and certainly wins in terms of content volume - but definitely not as much of a thrill where it really matters.
I finally caved in and bought RIDE2 in the latest PSN sale. Gave it a go yesterday around North West 200 after rage quitting TTIOM. It was pure borefest in comparison.
 
What you say would seem to contradict modern research.

Studies such as...?

If it were only based on reaction times as you state, then a framerate of 8 would suffice (I doubled it because Nyquist) - good luck hitting your braking points.

I never said that framerate is based on reaction time. I said that because reaction time is relatively slow, then a framerate of 30 is equal to a framerate of 60 in terms of your performance as a gamer.

[/quote]It should be clear that having two points separated by 33 ms contain less information about the (relative) motion of an object (say a rival racer or apex) than three at 16 ms; more dots to join up, more information; better informed and more precise "reactions".[/quote]

Not in a way that makes a meaningful difference.

Sound and vision are processed by different parts of the brain, are processed and interpreted differently outright and have different effective "latencies" - for instance a reaction from a sprinter (to a starting gun) of less than 100 ms is considered a false start, but F1 drivers manage that all the time (with lights).

This study (and those mentioned in it) shows that auditory stimuli gives a faster reaction time than visual stimuli. Why F1 drivers manage <100 ms "all the time" it's probably because they are allowed to gamble, while sprinters are not.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4456887/#!po=47.0000

How common are these <100 ms starts in F1? And are they more common than jump starts? If they're not significantly more common than jump starts then it could be argued that those occasions are simply drivers taking a chance and being lucky about it.

You cannot, for instance, separate two sounds that are about 100 ms or closer together - they sound as one.

Just tested that. I can easily distinguish when they're 25 ms apart. I attached an mp3 file so you can test it yourself as well. First are two 50 ms 440 hz tones with a 50 ms gap in between. Second are two 25 ms 440 hz tones with a 25 ms gap in between. At the end there is a single 1 second 880 hz tone just added there to make Audacity know when to end the track.

Flash a light twice 100 ms apart, and you'd see both flashes for sure (it's only 10 Hz). Interestingly, however, interaural time differences are of the order of a fraction of a millisecond and our brains extract useful (if not critical) spatial information from it - modeling it is very useful in VR, and by extension games in general. So do you allow your audio engine to handle such short delays, or not? Clearly, it's not just one cue we rely on with any of our senses.

As interesting as interaural time differences are, it' s completely irrelevant to framerate.

The very fact is that a whole brace of feedback mechanisms - not just reaction times - all add up to make a difference to the experience (even if that effect is diminished by things like aging), so it should be something we work towards in time, as we have done for many years already. This concept of a discrete stream of slides effectively having zero delay is utterly false, though: its discrete nature is the very basis by which latency is measured.

Of course it makes a difference to the experience, 60 fps is smoother. But it doesn't make any difference to your ability as a gamer. You don't miss out on any significant information at 30 fps and there is no noticeable delay.

In short: it's not (just) about reacting, it's about the feeling of connectedness created by several aspects of visual processing. Ask any professional gamer, higher framerates just feel better in action games - this then means they actually tend to "perform" better.

Which is what I've said all along. 60 fps is smoother, but that's also all there is to it. I'm sure it can feel awkward to go to 30 fps when you're used to 60 fps but once you are used to 30 fps you'll perform equally. There is no difference caused by delay.
 

Attachments

  • 100-50ms.zip
    151.9 KB · Views: 12
http://www.gameblog.fr/tests/2992-tt-isle-of-man-xb1-pc-ps4

7/10

TT Isle of Man est un bon jeu de course de moto, réaliste et très exigeant. À condition d'être complètement amoureux de l'épreuve mythique de l'île de Man, il peut même devenir un indispensable. Si c'est votre cas, vous pouvez ajouter un point à la note, sinon, vous lui pardonnerez sans doute moins son contenu "monomaniaque" et ses quelques défauts, notamment d'IA. Quoi qu'il en soit, on ne lui enlèvera pas ses qualités principales, à savoir les sensations de vitesse et de danger d'une conduite "réaliste". Tout cela sans risquer sa peau.

+ ON AIME
  • Des motos aux comportements assez réalistes.
  • Une licence officielle.
  • Les Side Cars qui arrivent gratuitement.
  • Des sons de moteurs enivrants.
  • Une sensation de vitesse grisante.
  • Le tracé de l'île de Man impressionnant de précision.
  • Des motos parfaitement modélisées.
  • Un contenu mythique...
- ON N'AIME PAS
  • ... mais trop léger et exclusif.
  • Une carrière bancale.
  • Des catégories manquantes de la course réelle.
  • Une IA limite.
  • Un multi en retrait.
 
I play on Ps4 Pro for two days ! I'm hypnotized by this game ! It's beautiful, fluid and after a few hours in settings 100% simulation, it becomes a pleasure all the time !

Yesterday I finished my first 60 kms a smug smile on my face ...

And even the other tracks are clever !

After a great WRC 7, Kylotonn (french studio ^^) proves that he is going in the right direction !
 
Studies such as...?

It's predominantly proprietary research for VR, where vendors tout the importance of lower latency (of which frame delay is a part). They also all agree that the findings would benefit general interactive environments, not just VR headsets.

Carmack wrote a lengthy piece a couple of years ago.
www.twentymilliseconds.com/post/latency-mitigation-strategies/
Abridged:
http://oculusrift-blog.com/john-carmacks-message-of-latency/682/

nVidia are working on very low latency displays.
https://www.roadtovr.com/nvidia-demonstrates-experimental-zero-latency-display-running-at-17000hz/

In terms of object tracking, whether that be on an ordinary display or via a headset, latency is extremely important.
https://www.researchgate.net/public..._Lag_and_Frame_Rate_on_Various_Tracking_Tasks

I never said that framerate is based on reaction time. I said that because reaction time is relatively slow, then a framerate of 30 is equal to a framerate of 60 in terms of your performance as a gamer.

Not in a way that makes a meaningful difference.

Citation needed. It makes a difference to me, maybe not for you, and generally gamers are somewhat split on this - where it matters, in competition, consensus is that higher framerates help, amazingly even at the expense of visual acuity (resolution).

This study (and those mentioned in it) shows that auditory stimuli gives a faster reaction time than visual stimuli. Why F1 drivers manage <100 ms "all the time" it's probably because they are allowed to gamble, while sprinters are not.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4456887/#!po=47.0000

They don't mention how they calculated the input - output latency of the audio and graphics systems of the "laptop" in order to frame and work out actual reaction times. In a typical platform, audio and video latency will be different (perhaps significantly so), and video is usually somewhere above 100 ms end-to-end - I struggle to get my audio latency below 50 ms without special drivers. Their software makes use of standard platform driver layers, but it is at least internally precise.
http://www.millisecond.com/products/inquisit5/timing.aspx

That was also an issue with the reaction time test you linked to previously, where I could get reaction times of 100 ms (accidental premature clicks) without even seeing the screen change colour. I can also get consistently faster results (about 80 ms faster) on this test: https://f1-start.glitch.me. A friend once built a reaction time tester using electronic components, and reactions were generally faster on that too.

In addition, some auditory reactions are faster than others, like reflexes or other subconscious "reactions" (see "Auditory Evoked Potentials"). Crucially, it depends on what exactly is heard, and, for example, the brain stem can filter these out (a twig snapping) before your brain even gets chance to truly "analyse" it in its cortical regions. The same part of the brain stem handles certain visual reflexes, too - notice that F1 drivers rely on their peripheral vision in their reflex tests, for example.

Framerate is, therefore, only a small part of the total input-output latency issue, but minimising its contribution goes quite a long way to improving the situation. If you have an end-to-end latency of 100 ms, going from 30 Hz to 60 Hz is not going to be as noticeable as it would be on a low-latency system (implying framerate will be the bottleneck to response, not the platform), which is now perfectly possible thanks to VR. E.g. a 50 ms latency and 90 Hz refresh target in current VR means the end to end latency at 30 and 60 Hz are 72 ms and 55 ms respectively - a substantial relative difference.

How common are these <100 ms starts in F1? And are they more common than jump starts? If they're not significantly more common than jump starts then it could be argued that those occasions are simply drivers taking a chance and being lucky about it.

I saw it on TV, I can only find one clip that mentions 100 ms (here), I also found 200 ms regarding Bottas' claimed jump start. Perhaps I was mislead / misremembered, although it's probably where I got the sprinter comparison! The point I was making is that there is a difference in response according to the type of cue, even within the same sense.

There is also some confusion as to what is being counted as a reaction in research. Sometimes it's motor response, sometimes it's just "recognition" somewhere in the brain. They have even timed and traced the "signal paths", which is pretty cool.

Just tested that. I can easily distinguish when they're 25 ms apart. I attached an mp3 file so you can test it yourself as well. First are two 50 ms 440 hz tones with a 50 ms gap in between. Second are two 25 ms 440 hz tones with a 25 ms gap in between. At the end there is a single 1 second 880 hz tone just added there to make Audacity know when to end the track.

Yes I performed similar tests when I first came across that nugget, but the figure pertains to continuous sounds within a full soundscape - if a change in both occurs within 100 ms, it is perceived (esp. by untrained ears) as being linked and therefore the same sound / source. The analogy to a "continuous stream" of frames should be obvious.

You can take that clinical test all the way down to individual samples perceived at 2-3 ms separation. But it kind of sounds like a drop of water hitting a rigid surface, which would ordinarily be perceived as one sound, whereas I know that it's two ;)
It's also related to the perception of spaces, where the reflections arriving before 100 ms have one perceptual role, and those after have a different role. Again, this is a cognitive grouping of (learned) associations between sounds. Also again, it varies.

As interesting as interaural time differences are, it' s completely irrelevant to framerate.

It represents encoded information that would be lost if the audio system did not have fine grained outputs enough to represent them, information that would be useful to the player. A maximum ITD of ~600 microseconds is still much less than 2 milliseconds, which itself is a very optimistic figure for typical listening environments. Most games update sound controls at video rate, but sound hardware has been doing more subtle stuff for output mixing for decades - because it makes a perceptual difference. Those games that decouple the audio inherently sound more precise, or "smooth".


Of course it makes a difference to the experience, 60 fps is smoother. But it doesn't make any difference to your ability as a gamer. You don't miss out on any significant information at 30 fps and there is no noticeable delay.

Define "significant". Much of in the zone "reactions" are subconscious, it's not just about spying your braking points. For example, the very limit of vehicle control is reading subtle movements in pitch, roll and yaw, missing those cues (often near the periphery of vision), or ignoring them altogether if they don't make sense (because of temporal aliasing), will be a disadvantage. At the very peak at least. For everyone else, it's just a nicer experience.

Which is what I've said all along. 60 fps is smoother, but that's also all there is to it. I'm sure it can feel awkward to go to 30 fps when you're used to 60 fps but once you are used to 30 fps you'll perform equally. There is no difference caused by delay.
We agree it feels different, I'm simply stating that that "feel" has more significant implications than you seem to want to admit.

A really interesting experiment would be to buffer the video output gratuitously, so that the output stays a fixed 30 fps, but is delayed by increasing amounts. No doubt that would get frustrating to control very quickly. When we touch an object, we expect it to respond immediately, not a tenth of a second later - that's the world our brains evolved in.

Another cool experiment would be to put shutter glasses on athletes and see if they hurt themselves... :)
 
Last edited:
The ride games to me have always been lack luster, no sense of speed or any raw bike power and they turn too slow..

Also dam you amazon no ps4 copies left of TT...grrrrr

I picked it up for the Xbone and it plays great.... I like having the "triggers" for gas/brake over the ps4 controls.
 
where it matters, in competition, consensus is that higher framerates help, amazingly even at the expense of visual acuity (resolution)

Hmmm... in the context of this game, you're constantly looking quite far down the road. I sit about 2m away (probably more like 1.5m when I'm leaning forward) from a 42" screen, and seem to spend most of my time focusing on a relatively small area in the middle of the screen. As things in the distance aren't moving that much, relative to things in the foreground, I'd rather have the resolution I think. Things tend to move faster, the closer they get, there comes a point where these things are simply too close to do anything about when you're hurtling towards them at 150mph on a bike... By the time they're moving the quickest (and therefore making best use of the frame rate), they're about to run off the edge of the screen, and it doesn't matter how well you perceive them, it's too late.

_____________

Anyhow.. off the topic of frames per second...

Managed to get into the 21's last night with a 166km/h lap. Still some way off my 18'13" target for a 200km/h lap.... but then I'm still throwing myself at the scenery about 15 times per lap!

I have to say, I've long been aware of the TT, and on paper how dangerous it is, I've even seen some of the crashes before and thought... oh man, that's unlucky... but now I have a much greater understanding of the course and the speeds involved I think it's phenomenal that there aren't more deaths... and you really must be ******** nuts to hit it at maximum attack in real life. I also think it's miraculous that in this age, it's still an event that happens - and where I am normally pretty loose with my opinion on track safety, it does make even me question why there aren't more attempts to make it safer.
 
Hmmm... in the context of this game, you're constantly looking quite far down the road. I sit about 2m away (probably more like 1.5m when I'm leaning forward) from a 42" screen, and seem to spend most of my time focusing on a relatively small area in the middle of the screen. As things in the distance aren't moving that much, relative to things in the foreground, I'd rather have the resolution I think. Things tend to move faster, the closer they get, there comes a point where these things are simply too close to do anything about when you're hurtling towards them at 150mph on a bike... By the time they're moving the quickest (and therefore making best use of the frame rate), they're about to run off the edge of the screen, and it doesn't matter how well you perceive them, it's too late.

_____________

Anyhow.. off the topic of frames per second...

Managed to get into the 21's last night with a 166km/h lap. Still some way off my 18'13" target for a 200km/h lap.... but then I'm still throwing myself at the scenery about 15 times per lap!

I have to say, I've long been aware of the TT, and on paper how dangerous it is, I've even seen some of the crashes before and thought... oh man, that's unlucky... but now I have a much greater understanding of the course and the speeds involved I think it's phenomenal that there aren't more deaths... and you really must be ******** nuts to hit it at maximum attack in real life. I also think it's miraculous that in this age, it's still an event that happens - and where I am normally pretty loose with my opinion on track safety, it does make even me question why there aren't more attempts to make it safer.
Most of the men and women who race there have either lived there for a while or visit on a regular basis just to learn the course, they drive around in cars with ex racers like Milky Quayle who give them tips and the benefit of their experience.
I've visited a few times to watch, I usualy spend each day at a different part of the course, I've ridden around it on road bikes too and I'm always stunned at the pace the racers go.

I still haven't completed a lap without some kind of incident. Maybe tonight, hopefully. I'll post my lap time if I can get around it without crashing. At the minute I lap at the 20 mins mark on the 600's. Going to experiment without 'assists' over the weekend too and maybe split the brakes, let you's know how I get on. :-)
 
Last edited:
It's predominantly proprietary research for VR, where vendors tout the importance of lower latency (of which frame delay is a part). They also all agree that the findings would benefit general interactive environments, not just VR headsets.

VR is an entirely different thing though, and it doesn't have the same requirements as gaming on a traditional screen.

In terms of object tracking, whether that be on an ordinary display or via a headset, latency is extremely important.
https://www.researchgate.net/public..._Lag_and_Frame_Rate_on_Various_Tracking_Tasks

And what does that article say? I requested the full text, but I guess it takes some time before I receive it.

Edit: Found it elsewhere. It doesn't show any significant difference between 60, 30 or even 20 fps.

Citation needed.

Citation provided. From the very same study, actually.

IMG_1140.PNG

This is the result of the tracking test. With a target frequency factor of 1, the error is pretty much consistent between 60, 30, 20 and 10(!) fps.

With a frequency factor of 3, there is no significant difference between 60, 30 and 20 fps.

As you might know, this study is from 2001, when some computer interfaces sometimes operated at down to 2 fps (one of the test subjects were actually experienced in working under such circumstances), and the conclusions are primarily based on the lower end of the framerates - that's where it makes a noticeable difference.

It makes a difference to me, maybe not for you, and generally gamers are somewhat split on this - where it matters, in competition, consensus is that higher framerates help, amazingly even at the expense of visual acuity (resolution).

Citation needed ;)

They don't mention how they calculated the input - output latency of the audio and graphics systems of the "laptop" in order to frame and work out actual reaction times. In a typical platform, audio and video latency will be different (perhaps significantly so), and video is usually somewhere above 100 ms end-to-end - I struggle to get my audio latency below 50 ms without special drivers. Their software makes use of standard platform driver layers, but it is at least internally precise.
http://www.millisecond.com/products/inquisit5/timing.aspx

So basically you think that the scientific community got it all wrong when the consensus is that we react faster to sound than to vision? You place more importance on the difference in false start rules between F1 and sprint?

In addition, some auditory reactions are faster than others, like reflexes or other subconscious "reactions" (see "Auditory Evoked Potentials"). Crucially, it depends on what exactly is heard, and, for example, the brain stem can filter these out (a twig snapping) before your brain even gets chance to truly "analyse" it in its cortical regions. The same part of the brain stem handles certain visual reflexes, too - notice that F1 drivers rely on their peripheral vision in their reflex tests, for example.

And reacting to a start pistol wouldn't be a reflex? What exactly are you implying? And how is it relevant to the discussion?

Framerate is, therefore, only a small part of the total input-output latency issue, but minimising its contribution goes quite a long way to improving the situation.

If it's only a small part, then it doesn't. And if those 0.017 seconds are a big part, then there's not much to improve.

I saw it on TV, I can only find one clip that mentions 100 ms (here), I also found 200 ms regarding Bottas' claimed jump start. Perhaps I was mislead / misremembered, although it's probably where I got the sprinter comparison!

I couldn't find any sensible data on F1 starts either, so let's just drop it from the conversation.

Yes I performed similar tests when I first came across that nugget, but the figure pertains to continuous sounds within a full soundscape - if a change in both occurs within 100 ms, it is perceived (esp. by untrained ears) as being linked and therefore the same sound / source. The analogy to a "continuous stream" of frames should be obvious.

You made the comparison to a flashing light though, and this audio file is a good analogy to that.

Define "significant".

Important and noticeable.

Much of in the zone "reactions" are subconscious, it's not just about spying your braking points. For example, the very limit of vehicle control is reading subtle movements in pitch, roll and yaw, missing those cues (often near the periphery of vision), or ignoring them altogether if they don't make sense (because of temporal aliasing), will be a disadvantage. At the very peak at least. For everyone else, it's just a nicer experience.

Why makes you think you would miss those cues at 30 fps?

It would be much more distracting to have a sudden framerate drop, so for that reason it's better to go for a steady 30 fps than gamble with a 60 that might not hold up when the scenes get busy.

We agree it feels different, I'm simply stating that that "feel" has more significant implications than you seem to want to admit.

Provide support for the statement and I'll admit it. Until then it's just a statement.

A really interesting experiment would be to buffer the video output gratuitously, so that the output stays a fixed 30 fps, but is delayed by increasing amounts. No doubt that would get frustrating to control very quickly.

That's exactly what they did in the study you shared. Beyond 0.1 seconds the errors started to increase.

When we touch an object, we expect it to respond immediately, not a tenth of a second later - that's the world our brains evolved in.

We can handle small delays without a problem though. If you play an instrument you need to start playing the tone a fraction of a second before it should sound. Same with video games, if there is a tiny delay we learn to press the button a fraction of a second before we need the action.

But 0.03 seconds is such a tiny delay that we don't even notice it.
 
Hey folks that LeanGP simulator looks awesome, i want one of those.And the price isn't bad either.I wonder if they ship to the UK and how much for delivery.Shame it hasn't got a clutch lever though.Might be a good idea to add that to the next prototype.I know none of the bike games have the use of a clutch, that i know of.Would be good if that were an option in the game so it can be mapped to this simulator.
 
Last edited:
VR is an entirely different thing though, and it doesn't have the same requirements as gaming on a traditional screen.

Not exactly, but vendors agree that changes required to rendering in general to benefit VR would have a positive impact on gaming overall, for basically the same reasons: responsiveness. Maybe also something something share price. Who knows.
And what does that article say? I requested the full text, but I guess it takes some time before I receive it.

Edit: Found it elsewhere. It doesn't show any significant difference between 60, 30 or even 20 fps.

Citation provided. From the very same study, actually.

View attachment 720559

This is the result of the tracking test. With a target frequency factor of 1, the error is pretty much consistent between 60, 30, 20 and 10(!) fps.

With a frequency factor of 3, there is no significant difference between 60, 30 and 20 fps.

As you might know, this study is from 2001, when some computer interfaces sometimes operated at down to 2 fps (one of the test subjects were actually experienced in working under such circumstances), and the conclusions are primarily based on the lower end of the framerates - that's where it makes a noticeable difference.

Well except that their concept of the "zero delay" case they provide is erroneous in the context of end-to-end latency in a computer system, as I already stated. Regardless, from your pre-existing experience, would you notice a change from 20 to 30 fps? What about 10? I'm not using it to cite 60 Hz being "superior", only their finding that frame delay is analogous to naked lag, which itself you were denying.

The limitations of that test include the fact that it was run at no higher than 60 Hz (probably hardware-limited), so it cannot reproduce information above 30 Hz without aliasing, but its major flaw is its false attribution of "zero lag". Remember I said that typical latency is of the order of 100 ms - that's 10 Hz equivalent ;)

In other words, the tests in the range of 10 Hz plus are probably unreliable (at least in the context of what they claim it shows), however the small changes in performance over that range imply that relative changes in delay of the order of "only" 17% are noticeable.

Modern methods of measuring lag include high-speed photography, with the catch-phrase: "motion to photons". This kind of analysis has been the game changer for VR this time around, and I'd argue it re-frames the discussion regarding latency for ordinary games as well.

(Sorry, I assumed the .pdf was reachable from that link)

Citation needed ;)
c.f. all the guides to maximising framerates ahead of LAN parties / competitions. Granted, that is more prevalent for FPSs than sims, but I am arguing from a general perspective - as always. Notice that @MatskiMonk already stated a preference to prioritise resolution in racing sims, and I think I would seek a balance between definition and response myself. In fact, I always have, but I would prefer to have the maximum of both in every type of action game.

I recall some interesting discussions regarding iRacing, but I'm no longer a member. I did find this, though:
https://steamcommunity.com/app/241560/discussions/0/34096318669939263/

Even at 70 Hz, hitting braking points is not guaranteed - it's recommended to run faster than that. Of course, with the square relationship between speed and stopping distance, I guess every little ounce of precision helps, even with our powers of anticipation (how do you know you hit your intended braking point if it falls between frames? That's important feedback you don't want to lose to aliasing).

Wipeout springs to mind at this point.

So basically you think that the scientific community got it all wrong when the consensus is that we react faster to sound than to vision? You place more importance on the difference in false start rules between F1 and sprint?

If the "consensus" is reached using the same flawed deployment of software (notice the software devs state that low-latency serial hardware is available, but that paper states they used the standard keyboard...), then maybe. I saw some surprisingly and incredibly shoddy science when I was in research - often it was simply a case of people not being aware of something, and all the existing "science" had to be re-done (fun times). Maybe people assumed computers respond instantaneously, because speed of light.

I'm not an expert in the field, I'll admit, and, as I've said, I've seen other figures for tracing actual neuron firing events which also depend on a lot of things. So "it depends" is my stance, as it was from the start.

And reacting to a start pistol wouldn't be a reflex? What exactly are you implying? And how is it relevant to the discussion?

If it's only a small part, then it doesn't. And if those 0.017 seconds are a big part, then there's not much to improve.

I couldn't find any sensible data on F1 starts either, so let's just drop it from the conversation.

You made the comparison to a flashing light though, and this audio file is a good analogy to that.

My point is that it varies according to what information is being perceived by the brain and other organs. Also the fact that much of it isn't a case of focused reaction as it is reflex or other subconscious processes, which have different mechanisms and different timings and so, for example, a "pure reaction time" (starts) or "focused attention / anticipation" (braking points) is maybe not the best basis to use outright for what framerate is ideal.

As above, that 0.017 seconds is noticeable.

Maybe a sprinter does rely on reflex to get out of the gate quickly, rather than any real cortical process (like most processing of the content of sounds requires).
What do you suppose would happen if some comedian shouts "bang!" at the Olympics? Aside from them being escorted out, I mean...

Important and noticeable.

It is both to me. Is that really the limit of discussion here? If so, fine: I prefer higher framerates and I think everyone should have access to them! :)

Why makes you think you would miss those cues at 30 fps?

Aliasing. Also, apparently (and relatedly), simple physics: distance = speed * time. As in the above link, try converting frames per second to metres per frame for some of your favourite speeds.

It would be much more distracting to have a sudden framerate drop, so for that reason it's better to go for a steady 30 fps than gamble with a 60 that might not hold up when the scenes get busy.

Maybe, but why "gamble" at all? Why not target 60 Hz in the first instance? It's the same process for 60 as it is for 30. Doesn't really say anything about whether 60 Hz is beneficial from an interactivity point of view, just shows how it's an industry inertia issue. My favourite. ;)

Provide support for the statement and I'll admit it. Until then it's just a statement.

Well ditto for your statement. Again, is this the level we're going to work on?

That's exactly what they did in the study you shared. Beyond 0.1 seconds the errors started to increase.

As mentioned, no they didn't. They started with a baseline 100 ms delay, and only added to it from there. But at least, as I said, it disproves the idea that a stream at 30 fps has no delay. Thus implying it can be improved upon, from a responsiveness and controllability point of view. The errors started to increase significantly above 100 ms because the zero wasn't zero, it was 100. 100% is probably a good definition of "significant".

We can handle small delays without a problem though. If you play an instrument you need to start playing the tone a fraction of a second before it should sound. Same with video games, if there is a tiny delay we learn to press the button a fraction of a second before we need the action.

That's true, and that's commented on a lot in the research as well, we can adapt very well in perceptual terms (probably related to tying together distant events in sight and sound - it takes sound 100 ms to travel 34 metres) - it doesn't mean we are still operating optimally. It doesn't mean that reducing that delay doesn't feel better; consensus is that it does feel better, in fact. It doesn't necessarily prevent us from enjoying it, either, which is, as I said, a part of the art of the video game.

Indeed as I mention that, I did happen across some discussion of old arcade emulators going to extreme measures to hack into render processes and driver gubbins on modern platforms to better approximate the crisp response of classic arcade games. Some of which ran at 60 Hz raw, hardware - probably little more than 17 ms total latency. In the 1970s. John Carmack once regaled of id Softwares early days of porting arcade games to PC (DOS), that involved the use of videoing the native machine and recreating its output (as opposed to its internal code), frame by frame, in order to reproduce its feel! No wonder he landed on "motion to photons" analysis.

I don't think the musical instrument analogy is accurate. The notes sound the very moment physics lets them (that's what makes them so appealing), and it is that that your brain keys into through various feedback mechanisms. I.e. you latch onto the timing of a particular point within your physical movements (of playing the note) as being the rhythmic "centre". Otherwise all notes start when you pick the guitar up. Also, interestingly, playing a note does not really involve a reaction as such (except maybe in Jazz), but you do subconsciously react to the instrument if, say, your grip shifts or if the controls are in a different position as you go to play them. You feel where they are before you play them and adjust your input accordingly. That would be analogous to my point of there being effectively hidden feedback involved.

But 0.03 seconds is such a tiny delay that we don't even notice it.
Research shows that most people notice a delay down to 0.02 seconds*, even on a flat screen with the bare minimum of interactivity: a mouse cursor. That's a little more than one frame at 60 Hz, which doesn't leave much room for I/O and driver latency. So to make room for it and still hit that 20 ms target (well, 50 realistically today), you minimise frame delay by maximising framerate - simple.

* which is to say, most people notice an improvement in responsiveness up to that point, some go beyond it - that change in qualitative terminology (delay -> response) probably reflects a change in the (effectiveness of certain) mode(s) of perception once you reduce lag beyond a certain point (100 ms?), which may be related to what I was talking about with my "it depends" comments.
 
Last edited:
Not exactly, but vendors agree that changes required to rendering in general to benefit VR would have a positive impact on gaming overall, for basically the same reasons: responsiveness. Maybe also something something share price. Who knows.



Well except that their concept of the "zero delay" case they provide is erroneous in the context of end-to-end latency in a computer system, as I already stated. Regardless, from your pre-existing experience, would you notice a change from 20 to 30 fps? What about 10? I'm not using it to cite 60 Hz being "superior", only their finding that frame delay is analogous to naked lag, which itself you were denying.

The limitations of that test include the fact that it was run at no higher than 60 Hz (probably hardware-limited), so it cannot reproduce information above 30 Hz without aliasing, but its major flaw is its false attribution of "zero lag". Remember I said that typical latency is of the order of 100 ms - that's 10 Hz equivalent ;)

In other words, the tests in the range of 10 Hz plus are probably unreliable (at least in the context of what they claim it shows), however the small changes in performance over that range imply that relative changes in delay of the order of "only" 17% are noticeable.

Modern methods of measuring lag include high-speed photography, with the catch-phrase: "motion to photons". This kind of analysis has been the game changer for VR this time around, and I'd argue it re-frames the discussion regarding latency for ordinary games as well.

(Sorry, I assumed the .pdf was reachable from that link)


c.f. all the guides to maximising framerates ahead of LAN parties / competitions. Granted, that is more prevalent for FPSs than sims, but I am arguing from a general perspective - as always. Notice that @MatskiMonk already stated a preference to prioritise resolution in racing sims, and I think I would seek a balance between definition and response myself. In fact, I always have, but I would prefer to have the maximum of both in every type of action game.

I recall some interesting discussions regarding iRacing, but I'm no longer a member. I did find this, though:
https://steamcommunity.com/app/241560/discussions/0/34096318669939263/

Even at 70 Hz, hitting braking points is not guaranteed - it's recommended to run faster than that. Of course, with the square relationship between speed and stopping distance, I guess every little ounce of precision helps, even with our powers of anticipation (how do you know you hit your intended braking point if it falls between frames? That's important feedback you don't want to lose to aliasing).

Wipeout springs to mind at this point.



If the "consensus" is reached using the same flawed deployment of software (notice the software devs state that low-latency serial hardware is available, but that paper states they used the standard keyboard...), then maybe. I saw some surprisingly and incredibly shoddy science when I was in research - often it was simply a case of people not being aware of something, and all the existing "science" had to be re-done (fun times). Maybe people assumed computers respond instantaneously, because speed of light.

I'm not an expert in the field, I'll admit, and, as I've said, I've seen other figures for tracing actual neuron firing events which also depend on a lot of things. So "it depends" is my stance, as it was from the start.



My point is that it varies according to what information is being perceived by the brain and other organs. Also the fact that much of it isn't a case of focused reaction as it is reflex or other subconscious processes, which have different mechanisms and different timings and so, for example, a "pure reaction time" (starts) or "focused attention / anticipation" (braking points) is maybe not the best basis to use outright for what framerate is ideal.

As above, that 0.017 seconds is noticeable.

Maybe a sprinter does rely on reflex to get out of the gate quickly, rather than any real cortical process (like most processing of the content of sounds requires).
What do you suppose would happen if some comedian shouts "bang!" at the Olympics? Aside from them being escorted out, I mean...



It is both to me. Is that really the limit of discussion here? If so, fine: I prefer higher framerates and I think everyone should have access to them! :)



Aliasing. Also, apparently (and relatedly), simple physics: distance = speed * time. As in the above link, try converting frames per second to metres per frame for some of your favourite speeds.



Maybe, but why "gamble" at all? Why not target 60 Hz in the first instance? It's the same process for 60 as it is for 30. Doesn't really say anything about whether 60 Hz is beneficial from an interactivity point of view, just shows how it's an industry inertia issue. My favourite. ;)



Well ditto for your statement. Again, is this the level we're going to work on?



As mentioned, no they didn't. They started with a baseline 100 ms delay, and only added to it from there. But at least, as I said, it disproves the idea that a stream at 30 fps has no delay. Thus implying it can be improved upon, from a responsiveness and controllability point of view. The errors started to increase significantly above 100 ms because the zero wasn't zero, it was 100. 100% is probably a good definition of "significant".



That's true, and that's commented on a lot in the research as well, we can adapt very well in perceptual terms (probably related to tying together distant events in sight and sound - it takes sound 100 ms to travel 34 metres) - it doesn't mean we are still operating optimally. It doesn't mean that reducing that delay doesn't feel better; consensus is that it does feel better, in fact. It doesn't necessarily prevent us from enjoying it, either, which is, as I said, a part of the art of the video game.

Indeed as I mention that, I did happen across some discussion of old arcade emulators going to extreme measures to hack into render processes and driver gubbins on modern platforms to better approximate the crisp response of classic arcade games. Some of which ran at 60 Hz raw, hardware - probably little more than 17 ms total latency. In the 1970s. John Carmack once regaled of id Softwares early days of porting arcade games to PC (DOS), that involved the use of videoing the native machine and recreating its output (as opposed to its internal code), frame by frame, in order to reproduce its feel! No wonder he landed on "motion to photons" analysis.

I don't think the musical instrument analogy is accurate. The notes sound the very moment physics lets them (that's what makes them so appealing), and it is that that your brain keys into through various feedback mechanisms. I.e. you latch onto the timing of a particular point within your physical movements (of playing the note) as being the rhythmic "centre". Otherwise all notes start when you pick the guitar up. Also, interestingly, playing a note does not really involve a reaction as such (except maybe in Jazz), but you do subconsciously react to the instrument if, say, your grip shifts or if the controls are in a different position as you go to play them. You feel where they are before you play them and adjust your input accordingly. That would be analogous to my point of there being effectively hidden feedback involved.




Research shows that most people notice a delay down to 0.02 seconds*, even on a flat screen with the bare minimum of interactivity: a mouse cursor. That's a little more than one frame at 60 Hz, which doesn't leave much room for I/O and driver latency. So to make room for it and still hit that 20 ms target (well, 50 realistically today), you minimise frame delay by maximising framerate - simple.

* which is to say, most people notice an improvement in responsiveness up to that point, some go beyond it - that change in qualitative terminology (delay -> response) probably reflects a change in the (effectiveness of certain) mode(s) of perception once you reduce lag beyond a certain point (100 ms?), which may be related to what I was talking about with my "it depends" comments.

If you don't believe in that study, why don't you find one that shows that there are significant performance improvements when going from 30 to 60 fps?
 
If you don't believe in that study, why don't you find one that shows that there are significant performance improvements when going from 30 to 60 fps?
It shows that reducing latency is beneficial, and it shows that using a discrete representation of continuous information introduces an equivalent latency on their own, so using the minimum rate that is perceived as "smooth" is not enough.

Others (using motion to photons analysis behind closed doors) found that under 20 milliseconds is the target. You won't get that on modern platforms without a severe overhaul of the software architecture.

It also shows that reducing the latency from 117 to 100 milliseconds is barely noticeable, but noticeable nonetheless. Reducing it further with framerate alone is an exercise in diminishing returns. Eliminate all other sources of latency and 60 Hz is the bare minimum rate required for a truly responsive experience.


One thing I forgot to mention in relation to aliasing is the tuned frequencies of suspension. At 30 Hz, you won't (reliably) see / feel the effects of anything operating above 15 Hz. Remember I said that physics had to be decoupled from the graphics in order to improve precision? This is why.
 
Welp, 20'13" is my best lap of the full course so far. If I can stay on the bike for a full lap I think I'm on for a 19'00" lap... I've gone back to combined brakes because I wasn't happy with the controller mapping, and I've changed to one of the fractionally less powerful BMW S1000RR bikes.

It's highly addictive.
 
This is probably as good a video review of it I've seen so far. Can't wait to play this game now. Just sold a bunch of unplayed ps4 games on eBay in order to fund buying it.

I doubt you'll regret it.

Welp, 20'13" is my best lap of the full course so far. If I can stay on the bike for a full lap I think I'm on for a 19'00" lap... I've gone back to combined brakes because I wasn't happy with the controller mapping, and I've changed to one of the fractionally less powerful BMW S1000RR bikes.

It's highly addictive.

I found by doing each of the 12 sectors in time attack mode until I got familiar with them really helped me, (even though I'd played the Jester PS2 game) and then doing the full TT course, cut my time down to a 19.20ish on the cbr600 even with 6 or 7 crashes and riding at 85-95%.

I'm staying on combined brakes too (for the time being) and resorted to auto gears and chase cam, instead of helmet view and manual cogs.

It's still really frustrating to try and control the bike on sticks without binning it regularly and the R2 trigger needs a user calibration adjustment menu.
 
Back