Kazunori Yamauchi Says Tire Physics Remain One of Gran Turismo’s Biggest Challenges

[Jordan]

This is the discussion thread for an article on GTPlanet:

Kazunori Yamauchi Says Tire Physics Remain One of Gran Turismo’s Biggest Challenges​

 

[FMW]

Can GT7 even handle complicated tire models like the ones seen in iRacing or LMU?? The CPU that can be thrown on a PC is vastly different, so curious what the limitations are on PS5.

 

[05XR8]

…He says community feedback has drawn his attention to cars he might otherwise have overlooked, including the continuing appetite for ’90s Japanese machinery…

I approve of this message.
Keep checking the Suggestions to vote and/or add the cars you want to see in Gran Turismo. He’s looking.

 

[Theufcveteran]

This is the discussion thread for an article on GTPlanet:

Kazunori Yamauchi Says Tire Physics Remain One of Gran Turismo’s Biggest Challenges​

Thanks for sharing this and making the article. Thanks to evo as well for the interview with Kaz!

 

[M Stark]

Can GT7 even handle complicated tire models like the ones seen in iRacing or LMU?? The CPU that can be thrown on a PC is vastly different, so curious what the limitations are on PS5.

That’s a very good question.

What I gather from the article is that Kaz takes physics very seriously, but the tire model is still guesswork.

I think it’s very frustrating we can’t get a good tire model in GT. I can feel Kaz’s frustration too from what he’s saying. He made 2 partnerships with tire manufacturers, Michelin and now Dunlop and from what I understand along the lines is none of them is willing to share any actual data with PD. “Notoriously secretive” and “mysterious” is how Kaz characterized them 🙄 Smh There goes the Dunlop partnership I suppose.

 

[ZedNinetySix]

Tire physics are the single biggest challenge of every sim racing game. And have almost always been.

To be brutally honest no sim racing game has got tire physics right yet, all are within some sort of ballpark often using physics tricks outside of actual tire deformation geometry to portray certain behaviours.

Nobody has got it right yet.
And likely not for a while to come because in the real R&D world tyres are the single LEAST understood part of the car despite how far we have come.

The limitation is not really any game engine, the limitation is our current real world collective R&D on tires themselves and how hard it is to predict how they will behave in the most dynamic range of stresses. Pressure, temperature, atmosphere, surface, how the air spins and flows within the tire, slip angle, different compounds, different side walls, different structures, the grooves, the abrasion ect its endless.

They are the single most complex part of any car.

And any tire manufacturer who believes they have a better understanding of tires over their peers, will try and keep their R&D as secretive as they can, to have an edge over their peers.

 

[M Stark]

Tire physics are the single biggest challenge of every sim racing game. And have almost always been.

To be brutally honest no sim racing game has got tire physics right yet, all are within some sort of ballpark often using physics tricks outside of actual tire deformation geometry to portray certain behaviours.

Nobody has got it right yet.
And likely not for a while to come because in the real R&D world tyres are the single LEAST understood part of the car despite how far we have come.

The limitation is not really any game engine, the limitation is our current real world collective R&D on tires themselves and how hard it is to predict how they will behave in the most dynamic range of stresses. Pressure, temperature, atmosphere, surface, how the air spins and flows within the tire, slip angle, different compounds, different side walls, different structures, the grooves, the abrasion ect its endless.

They are the single most complex part of any car.

And any tire manufacturer who believes they have a better understanding of tires over their peers, will try and keep their R&D as secretive as they can, to have an edge over their peers.

From my understanding, that’s what I gather too.

For a developer to get some real data on tires they either have to somehow get some insider info from a tire manufacturer; lots of asterisks on how they could obtain that information since apparently none of them are too keen in sharing. Or, effectively develop their own tire R&D from scratch, which is ridiculous. In any other case it’s educated guesswork at best.

Honestly, it’s a pretty frustrating predicament.

 

[FMW]

That’s a very good question.

What I gather from the article is that Kaz takes physics very seriously, but the tire model is still guesswork.

I think it’s very frustrating we can’t get a good tire model in GT. I can feel Kaz’s frustration too from what he’s saying. He made 2 partnerships with tire manufacturers, Michelin and now Dunlop and from what I understand along the lines is none of them is willing to share any actual data with PD. “Notoriously secretive” and “mysterious” is how Kaz characterized them 🙄 Smh There goes the Dunlop partnership I suppose.

I don't blame any of the tire manufactures he's approached not sharing actual data. Michelin, Pirelli, Dunlop and Goodyear are all very competitive and tight lipped. Many have tried to do "How it's made" videos on tire production, especially race tires and all of them have said "HELL TO THE NO" LOL. It's like secret magic sauce.

I play a lot of LMU. I do really like what the've done with the tire model, but even that studio is constantly tweaking the model.

From my understanding, that’s what I gather too.

For a developer to get some real data on tires they either have to somehow get some insider info from a tire manufacturer; lots of asterisks on how they could obtain that information since apparently none of them are too keen in sharing. Or, effectively develop their own tire R&D from scratch, which is ridiculous. In any other case it’s educated guesswork at best.

Honestly, it’s a pretty frustrating predicament.

Race tires are so secretive that Pirelli takes both used and unused F1 tires back into their possession after a race weekend. From there those tires go back to a secret location so they never leave sight. I imagine any of their competitors would love to direct them.

 

[happycorey]

I approve of this message.
Keep checking the Suggestions to vote and/or add the cars you want to see in Gran Turismo. He’s looking.

It's great to know what they are still considering. It seems like there are so many suggestions and not enough targeted differences. Thing that can still provide unique experiences for their effort. Especially without seeming repetitive among the sea of GTRs

 

[Cosbuster]

Can GT7 even handle complicated tire models like the ones seen in iRacing or LMU?? The CPU that can be thrown on a PC is vastly different, so curious what the limitations are on PS5.

Both games you mentioned run fine on the PC equivalent of a PS5 CPU, in excess of 60fps, so it's not really a hardware limitation at this point.

 

[Fuzzi0n]

No surprise with the tyres being the hardest thing to simulate because it's been the main sticking point for sim racing for decades. It's incredibly hard to get any real world data and even if you do can you build a model that simulates it correctly in almost any condition, Niels Heusinkveld has some good videos talking about the challenges. Add to that you might also want to tweak to to make it more accessible to people on different input devices and then you have to balance realism with making something accessible and fun. Interesting to see they've been a bit surprised by what cars people want in the game and how that may have differed from what they thought would be best.

 

[Rocket Racer]

What I gather from the article is that Kaz takes physics very seriously, but the tire model is still guesswork.

Adding on to what @ZedNinetySix has said:

I've spent six or seven years of my professional life dealing with the modelling and simulation of driving dynamics. Any tire model is guesswork.

My old colleagues at a certain German OEM might have more elaborate guesswork supported by larger amounts of data, but still guesswork at the end of the day.

There's a reason one of the popular established models is literally called "Magic Formula", it's just a super hard problem to solve, where elegant physics-based approaches won't get you far, as opposed to the mentioned chassis kinematics.

 

[Nakano219]

From the interview:

'Then you need not just tracks, but beautiful environments. For example to reproduce the Nordschleife – the Nürburgring – took us four years.’

Is that the first time we've had a Kaz figure on track development time? I know the 8-9 months for a car is relatively common knowledge, but that's I think the first I've heard on the track side.

 

[Haitauer]

I approve of this message.
Keep checking the Suggestions to vote and/or add the cars you want to see in Gran Turismo. He’s looking.

Like we dont have 90s Japanese machinery already? 😂 How about european and american 90s and 00s? They are so very biased towards Japan the comments.

To be brutally honest no sim racing game has got tire physics right yet, all are within some sort of ballpark often using physics tricks outside of actual tire deformation geometry to portray certain behaviours.

Yes - Id say most of the time games use many things outside the actual tires to make the "tire" feel certain way.

For a developer to get some real data on tires they either have to somehow get some insider info from a tire manufacturer; lots of asterisks on how they could obtain that information

I dont think the data is the key. It is physics anyways, and the multipliers can be tested with trial and error. The problem is to implement these infinite calculations to code. In real world developing tyres is an inside to out side problem. You already have the physics, and you can calculate and test what works with them. In simulation the problem is from outside to inside - you try to make the tire work so the rest of the physics can match the aimed outcome. In simulation you need people who understand physics and coding AND have vision to see how these could be combined. Its a rare thing that cant fully be trained. You need mechanically and physically oriented people who understand software world. Two things that dont mix so often in the same peraon.

Both games you mentioned run fine on the PC equivalent of a PS5 CPU, in excess of 60fps, so it's not really a hardware limitation at this point.

Yes - I dont think its a hardware issue, but rather the limitations of ISA or operating system, that might not be as nimble to run different kind of software codes as PC is.

 

[ScottPuss20]

The thing about tyres in real life is that they don't have a "refresh rate". When they're contacting the surface of the road (whether that be gravel, tarmac, dirt or snow), they are constantly gripping without any delay. You'll never truly be able to capture all the intricacies on a digital system that relies on data input and output. However humans DO have a refresh rate so if data is being relayed fast enough, it's impossible to perceive the changes occuring. The trick is knowing specifically what the most important pieces of data are and relaying that in a way which feels engaging and fun.

 

[Rocket Racer]

The thing about tyres in real life is that they don't have a "refresh rate". When they're contacting the surface of the road (whether that be gravel, tarmac, dirt or snow), they are constantly gripping without any delay. You'll never truly be able to capture all the intricacies on a digital system that relies on data input and output. However humans DO have a refresh rate so if data is being relayed fast enough, it's impossible to perceive the changes occuring. The trick is knowing specifically what the most important pieces of data are and relaying that in a way which feels engaging and fun.

It's not a digital delay of course, but tires do take their time to build up grip, because they pretty much have to deform internally before actually transmitting forces. Look at a slo-mos of dragster tires under acceleration for an extreme example. It's the same in the lateral direction.

The sample rate can be a limitation, but is really not the core of the problem, you are usually far away from the Nyquist threshold in an area where discrete and continuous transfer functions are similar enough.

The problem is rather the myriad of degrees of freedom, with distinct nonlinear transient characteristics, and many of them interacting with each other. I'm not really qualified to go into much more detail but that's the general idea.

The second problem is the tire being the starting point of any macroscopic handling characteristic, so any weird quirk of your tire model quickly gets amplified.

 

[MroczMR]

Can GT7 even handle complicated tire models like the ones seen in iRacing or LMU?? The CPU that can be thrown on a PC is vastly different, so curious what the limitations are on PS5.

iRacing can run on PC with far slower CPU than PS5 have so its not about hardware. LMU also doesn’t have huge requirements and is going to be on consoles.

This is more a matter of software.

 

[Old_Setzer]

The game is on PS4, a hardware from 2014, it is fair not forget that.

Although I don't think it's just a hardware issue.

"Yamauchi also repeated the now-familiar figure that a single car takes roughly eight months to complete. That covers both the graphics model and the physics work required to make it behave like the real thing, compared to roughly two hours per car back in the original Gran Turismo days."

Wow!

 

[ArronWestley]

I'd say game balance is an even bigger one, because the nerfs to certain cars in 1.71 were hilarious and relegated them to uselessness.

iRacing can run on PC with far slower CPU than PS5 have so its not about hardware. LMU also doesn’t have huge requirements and is going to be on consoles.

This is more a matter of software.

GT7 runs on PS4 base. You can't run iRacing nor LMU on a PS4.

 

[Met Entso]

Then maybe create less cars and put more months in physics development?

 

[Talon16]

Like we dont have 90s Japanese machinery already? 😂 How about european and american 90s and 00s? They are so very biased towards Japan the comments.

My thoughts exactly when I read that. IMO, over the course of Gran Turismo the Japanese marketed cars have been way oversaturated. It's common knowledge at this point that PD are very bias towards Japanese cars. How many Miatas and GTRs does the game need with other manufactures are being left in the dust? Do we need the R32, R33, and R34? H3ll yes! Do we need 87 versions of the R34? No.

And this is coming from someone who loves 90s JDM cars. I'd rather see a Gen 3 RX-7 out and about over a Ferrari. But let's call a spade a spade here and at least recognize the bias. The best part is, Kaz/PD (maybe even Sony?) don't even hide it, which, despite how annoying it can be, I actually respect.

And if they're going to add more 90s JDMs I wouldn't hate to see the Talon/Eclipse (all 4G63 gens) make an appearance. BRB, on my way to the car wish-list thread as I may be pushing my luck

 

[Haitauer]

I'd say game balance is an even bigger one, because the nerfs to certain cars in 1.71 were hilarious and relegated them to uselessness.

People really should remember the game is more than Gr3 BoP. This update has wrecked havoc on sportmode due to BoP setups, but those are temporary and people should look behing just that when evaluating this update. I still think current Gr3 daily race with RS/RM is pretty bad, but like I said - go try Gr3 without BoP and with sport tyres or even RH. Its day and night difference to current daily C..

 

[Voodoovaj]

Can GT7 even handle complicated tire models like the ones seen in iRacing or LMU?? The CPU that can be thrown on a PC is vastly different, so curious what the limitations are on PS5.

I feel like people are missing the elephant in the room.

"Tire Physics"

Which tire? which tire are we basing this on? The Michelin? The Dunlop? The Pirelli? Which of those? Which model in their range? I don't think it's as much of an issue of the computational power, but which computation you have to throw at it, and how many of them with a field of 20 cars.

For instance, you are racing the 512BB (with 1976 Michelin XWX tires) against the Ferrari 296 (with Michelin Pilot Sport Cup 2 RK2) - which tire will you model? Both? Just one? Which one?

Most racing series have 1 series wide tire because the difference between 2 tires can change a car's performance drastically.

The game has always had 1 tire model and the grip is a variable, but that isn't how tires actually work. The construction of the tire matter enormously, so a proper one size fits all approach will always lead to compromise. If the game had 1 generic racing tire, it would be an easy mountain to climb.

 

[Seikenfreak]

Still need some more Skylines as well. Just sayin I'm sure a lot of people, especially the drift crowd, would appreciate some of the lower trim models that have different engines and/or RWD etc. They also look different. But people like to just lump them all together as one car. Do we need like.. Special Midnight Purple and Millennium Jade editions? Nur or M-spec? Etc. No, of course not.

And on the subject of bias towards Japanese cars.. Well for one, they are a Japanese company. And second, who's to say that it isn't much easier for them to get access or licensing, plus mint condition original Japanese cars vs an American or Euro car? I'd be willing to bet the Japanese auto manufacturers have a much closer and easier relationship with PD.

That being said, there are SO many cars I wish they would add. I think of new stuff every day. This morning in bed, for some crazy reason, I thought of a Pontiac Fiero GT. Do I like that car? Not really. Would it be fascinating to have and use in GT? Absolutely. That and tracks are my two biggest frustrations. How long does it take them to make a fantasy/original circuit vs a real one? And original circuits are more often than not, not fun IMO. Mostly the newer ones from last few games. And then they have other original circuits that we know are good and fun, but they don't add those.

I think it would be a fun challenge/demo for them to show like.. With todays modern software and tools, what could a car look like if they had 2 hours to make it? lol Would be like an indie game-jam thing. Would it look/sound any better than a GT2 car? I'd be curious.

On the subject of tires.. Eh. It will never be perfect. I just want the game to handle "realistically", whether that's done by "vibes" for them or Kaz's personal experience etc. It feels subjective and unknowable. I'm more interested in concrete subjects like cars and tracks.. and how can they add more, faster.

 

[MIE1992]

Still need some more Skylines as well. Just sayin I'm sure a lot of people, especially the drift crowd, would appreciate some of the lower trim models that have different engines and/or RWD etc. They also look different. But people like to just lump them all together as one car. Do we need like.. Special Midnight Purple and Millennium Jade editions? Nur or M-spec? Etc. No, of course not.

Plus there are still at least a couple very special Skyline models seen in previous installments, like the Nismo Skyline GT-R LM Road Going Version, and the Skyline BLRA-3 from the first generation that was designed by Giovanni Michelotti. Plus there’s the Skyline GT-R Autech Version 40th Anniversary, too.

 

[art3mis]

Here's what I'm struggling with: Why did PD feel the extreme need to change the physics. I haven't heard mass complaints. I haven't seen "pros" saying "wow this is bad" so why keep iterating when it was working fine?

It's like the IT worlds obsession with continually changing user interfaces. I don't need a new 'Ribbon' for Excel I need Excel to be the same it's always been. Add a new feature sure then change the UI but like MS changes things just to change things.
I feel PD does exactly the same.

 

[Theufcveteran]

Here's what I'm struggling with: Why did PD feel the extreme need to change the physics. I haven't heard mass complaints. I haven't seen "pros" saying "wow this is bad" so why keep iterating when it was working fine?

It's like the IT worlds obsession with continually changing user interfaces. I don't need a new 'Ribbon' for Excel I need Excel to be the same it's always been. Add a new feature sure then change the UI but like MS changes things just to change things.
I feel PD does exactly the same.

To be fair, you could argue it's PD just trying to get closer to realism. I personally don't mind physics changes during the game's life, even multiple times, but I don't make a ton of tunes so I can understand why it would be annoying for folks who do.

 

[effenberg]

Here's what I'm struggling with: Why did PD feel the extreme need to change the physics. I haven't heard mass complaints. I haven't seen "pros" saying "wow this is bad" so why keep iterating when it was working fine?

It's like the IT worlds obsession with continually changing user interfaces. I don't need a new 'Ribbon' for Excel I need Excel to be the same it's always been. Add a new feature sure then change the UI but like MS changes things just to change things.
I feel PD does exactly the same.

Maybe because GT, although considered a "good game", is considered "worse" than at least 5~10 other sim games?

What Kaz/PD are doing is wonderful. Despite the relative success, they are still trying to improve and to aim for the top.

 

[05XR8]

Like we dont have 90s Japanese machinery already? 😂 How about european and american 90s and 00s? They are so very biased towards Japan the comments.

Oh, I agree about the cars from Europe the USA. GT2 car list would be bonkers in GT7. Kaz probably means the variety of 90s Japanese machinery. We’ve got the popular Lancers and WRXs and Skylines. Cars like the recent Toyotas are different to the amount of Japanese coupes that have been added.

Just need to look back from GT1-GT6 at lots of different cars from 1990s Japan. If we only got one Lancer, one Silvia, one Civic, etc. instead of each generation, replaced with a Mirage ZR, Stagea, Forester turbo, that’s more of the variety like past games(except the first GT). The utility vehicles in GT7 help the variety immensely.

I don’t know if Kaz thought with the longer time it takes to model cars, continuing the car lineage in the way GT Sport did, was the best thing forward for the franchise as it is now. Lots of member(including me) were and are saying if PD added the NA Roadster, where’s the NB roadster? If PD added the 1969 Camaro, where’s the 1970 1/2? Yet, PD add more variations and trims of a newer model.
We know it’s due in part to the manufacturer, but sometimes it seems Kaz forgets the old games and needs to be reminded of what cars made the past games popular. Again, the time to model cars means he has to balance what are added, without even getting into car sourcing and licencing phases.

 

[art3mis]

Maybe because GT, although considered a "good game", is considered "worse" than at least 5~10 other sim games?

What Kaz/PD are doing is wonderful. Despite the relative success, they are still trying to improve and to aim for the top.

I still don't understand why. GT will never be a hardcore sim like iRacing. It also won't ever be the most sim title on it's own platform.

It's as dumb as microslop