This forum contains threads to discuss teams themselves. Anything not technical about the cars, including restructuring, performances etc belongs here.
The construction sites are the familiar ones. McLaren has tackled her, but can't get her out of the way overnight. No investment had been made in the Woking team location for more than ten years. It takes time to build a new high-tech simulator and wind tunnel, like Mercedes and Ferrari have. The simulator should be the first to go online this year. The wind tunnel is planned for the end of 2022. https://www.auto-motor-und-sport.de/for ... latz-vier/
Whilst the wind tunnel will be operational this year, it's not going to be calibrated and mapped out to where it's useful for probably 3-4 months later.
The construction sites are the familiar ones. McLaren has tackled her, but can't get her out of the way overnight. No investment had been made in the Woking team location for more than ten years. It takes time to build a new high-tech simulator and wind tunnel, like Mercedes and Ferrari have. The simulator should be the first to go online this year. The wind tunnel is planned for the end of 2022. https://www.auto-motor-und-sport.de/for ... latz-vier/
I suspect that whilst the teams will have similar tech for their simulators and similar code, they will all have developed their own CFD based algorithms that the code would use to create the racing and environmental feedback that you might call crude CFD.
Is there a limit on processing power for a simulator and at what point can it start to be considered an extension of CFD? Is there a consideration for that in the rules?
Sims don't do CFD, they just do a lookup table from the windtunnel data.
I thought the simulator wasn’t going to be ready till mid 2023?.. I assume the wind tunnel may help with the second development phase of the 2023 car, assuming the team can achieve correlation within the 3-4 month time frame..
The more I look into things like these, the more mad I get that Daniel is performing the way he is.
It just doesn't make sense. The guy was a rockstar when he was at RedBull, and he made Ocon look like a low class driver on their time together at Renault.
He joins McLaren and he gets completely beaten by Lando on a short track like Austria while having a power advantage over him.
Please Daniel, for the love of god, get back to performing the way you used to.
It makes perfect sense.
Here's the radio of Austria(click watch on Youtube).
He complains about the front end all race. He hasn't figured out how to drive around it.
I’m not McLaren, and I’m going to assume they’re smarter than I am (probably safe assumption) but if I were them I would be trying radical setups on DR’s car. He doesn’t like understeer, so take off some (maybe a lot) of rear wing and try to make the car pointier. That’s just an example, there’s many things they could do. Get experimental with the braking maps. Something.
If we look at Renault 2020, he was not remarkably better than Hulk the year before and was only marginally better than Ocon until they radically overhauled the setup between the first and second Silverstone races. And the effect was like a light switch in terms of DR’s performance.
If you try a radical setup and it fails you will finish 13th or 14th. At the rate the season is going, following the “standard”setup direction seems to only yield 8th-11th anyways. Maybe you will miss out on 10-20 points over a handful of races, but if you flip the light switch you can gain that back in 2. Maybe they are/have been doing that. If so I haven’t seen anything about it.
"You can't argue with stupid people, they will drag you down to their level and beat you with experience"
- Mark Twain
I thought the simulator wasn’t going to be ready till mid 2023? (or maybe they meant it wasnt going to be functional till mid 2023 because they're waiting for the new windtunnel to achieve correlation).. I assume the wind tunnel may help with the second development phase of the 2023 car, assuming the team can achieve correlation within the 3-4 month time frame..
The more I look into things like these, the more mad I get that Daniel is performing the way he is.
It just doesn't make sense. The guy was a rockstar when he was at RedBull, and he made Ocon look like a low class driver on their time together at Renault.
He joins McLaren and he gets completely beaten by Lando on a short track like Austria while having a power advantage over him.
Please Daniel, for the love of god, get back to performing the way you used to.
It makes perfect sense.
Here's the radio of Austria(click watch on Youtube).
He complains about the front end all race. He hasn't figured out how to drive around it.
I’m not McLaren, and I’m going to assume they’re smarter than I am (probably safe assumption) but if I were them I would be trying radical setups on DR’s car. He doesn’t like understeer, so take off some (maybe a lot) of rear wing and try to make the car pointier. That’s just an example, there’s many things they could do. Get experimental with the braking maps. Something.
If we look at Renault 2020, he was not remarkably better than Hulk the year before and was only marginally better than Ocon until they radically overhauled the setup between the first and second Silverstone races. And the effect was like a light switch in terms of DR’s performance.
If you try a radical setup and it fails you will finish 13th or 14th. At the rate the season is going, following the “standard”setup direction seems to only yield 8th-11th anyways. Maybe you will miss out on 10-20 points over a handful of races, but if you flip the light switch you can gain that back in 2. Maybe they are/have been doing that. If so I haven’t seen anything about it.
I don't know I recently saw a f1 article where Lando himself says he likes more oversteer in the car than Daniel, Daniel seems to be losing the time when the car is at low fuel...watching the onboards from Austria on the last stint through each corner Lando certainly seems far more confident/reactive and carries much more speed through the corners..
I also don't think this car is understeery, if anything its very front loaded*.
The more I look into things like these, the more mad I get that Daniel is performing the way he is.
It just doesn't make sense. The guy was a rockstar when he was at RedBull, and he made Ocon look like a low class driver on their time together at Renault.
He joins McLaren and he gets completely beaten by Lando on a short track like Austria while having a power advantage over him.
Please Daniel, for the love of god, get back to performing the way you used to.
It makes perfect sense.
Here's the radio of Austria(click watch on Youtube).
He complains about the front end all race. He hasn't figured out how to drive around it.
I’m not McLaren, and I’m going to assume they’re smarter than I am (probably safe assumption) but if I were them I would be trying radical setups on DR’s car. He doesn’t like understeer, so take off some (maybe a lot) of rear wing and try to make the car pointier. That’s just an example, there’s many things they could do. Get experimental with the braking maps. Something.
If we look at Renault 2020, he was not remarkably better than Hulk the year before and was only marginally better than Ocon until they radically overhauled the setup between the first and second Silverstone races. And the effect was like a light switch in terms of DR’s performance.
If you try a radical setup and it fails you will finish 13th or 14th. At the rate the season is going, following the “standard”setup direction seems to only yield 8th-11th anyways. Maybe you will miss out on 10-20 points over a handful of races, but if you flip the light switch you can gain that back in 2. Maybe they are/have been doing that. If so I haven’t seen anything about it.
There is no indication that this hasn’t happened already… Radical setups could end up (most probably) been slower for the same driver, as well as not allowing for tire management or not allowing for enough energy to be placed on the tires (and therefore operating out of the window)… I wouldn’t be surprised if some of this has been tried already within reason.
The construction sites are the familiar ones. McLaren has tackled her, but can't get her out of the way overnight. No investment had been made in the Woking team location for more than ten years. It takes time to build a new high-tech simulator and wind tunnel, like Mercedes and Ferrari have. The simulator should be the first to go online this year. The wind tunnel is planned for the end of 2022. https://www.auto-motor-und-sport.de/for ... latz-vier/
I suspect that whilst the teams will have similar tech for their simulators and similar code, they will all have developed their own CFD based algorithms that the code would use to create the racing and environmental feedback that you might call crude CFD.
Is there a limit on processing power for a simulator and at what point can it start to be considered an extension of CFD? Is there a consideration for that in the rules?
Sims don't do CFD, they just do a lookup table from the windtunnel data.
Interesting. Even some PC games have some simple CFD in them now but for things like smoke and it is carefully chose to be just a small amount of the graphics that are rendered this way. I'm surprised they don't use a map (or lookup as you say) from the wind tunnel and CFD box and then perform simple CFD on the fly, which should be possible now. Obviously you will not get close to full realtime CFD for many many years.
Eh, it's not really CFD they have zero interest in them being remotely accurate bar looking good.
It's usually a set of filters or just a couple of what are basically large snapshot images of vectors with local collision detection that are then raytraced into spheres with the texture overlaid. It's the only way to do it remotely quickly enough on a GPU. It's nothing like being useful or accurate for anything apart from looking interesting.
The lookup table is far more accurate, same as they do for tyres.
Interestingly most games went away from lookup tables for tyres because they can't get accurate data easily and now use a mix-mash of a couple different models depending on the car speed to fudge it. It's why you get die hard sim fans running heavily modded versions of old games like RBR because the tyre feel is more realistic than a lot of modern games.
Eh, it's not really CFD they have zero interest in them being remotely accurate bar looking good.
It's usually a set of filters or just a couple of what are basically large snapshot images of vectors with local collision detection that are then raytraced into spheres with the texture overlaid. It's the only way to do it remotely quickly enough on a GPU. It's nothing like being useful or accurate for anything apart from looking interesting.
The lookup table is far more accurate, same as they do for tyres.
Interestingly most games went away from lookup tables for tyres because they can't get accurate data easily and now use a mix-mash of a couple different models depending on the car speed to fudge it. It's why you get die hard sim fans running heavily modded versions of old games like RBR because the tyre feel is more realistic than a lot of modern games.
I'm not talking about F1 games or racing sims, I'm talking about the emergence of CFD in some video games. There is definitely some CFD beyond just the physics that you talk about but on racing sims where it would not be possible on home powered PCs. iRacing was developing or working towards developing CFD internal to iRacing with TotalSim and that started several years back. Where that is now I don't know but it is not inconceivable that it sits within a corporate platform or that it couldn't be used for a small specific area of the simulation to add realism.
iRacing talk about specifically moving away from the Aero mapping generated from previous generic CFD runs.
Last edited by mwillems on 20 Jul 2022, 14:18, edited 2 times in total.
That's because it's prohibitively expensive for them to generate or get that sort of aeromapping, at least to usable accuracy. A team has it at hand. Same as with tyre models.
The TotalSim stuff is not realtime CFD, they're just doing aeromapping using openfoam, same as many posters here do. so that iRacing can change the characterstics of the aero depending on following cars, yaw, setup, rideheights, etc. It's definately not a realtime process at all and is still using lookup tables for the game itself - you just don't have anything near the processing power required for that bar the simple stuff I mentioned earlier for effects - we're literally orders of magnitude in processing power away from even basic CFD in realtime in games.
That's because it's prohibitively expensive for them to generate or get that sort of aeromapping, at least to usable accuracy. A team has it at hand. Same as with tyre models.
The TotalSim stuff is not realtime CFD, they're just doing aeromapping using openfoam, same as many posters here do. so that iRacing can change the characterstics of the aero depending on following cars, yaw, setup, rideheights, etc. It's definately not a realtime process at all and is still using lookup tables for the game itself - you just don't have anything near the processing power required for that bar the simple stuff I mentioned earlier for effects - we're literally orders of magnitude in processing power away from even basic CFD in realtime in games.
Again I'm not saying it is yet, just that they are working towards it.
There are engineering platforms now that offer levels of Realtime CFD too and I have attached links.
I'm not suggesting that the whole sim would be CFD, but that there could well be aspects of the sim that use CFD to make the responses more realistic and to provide data. That processing power, due to the cloud, is now available and not as expensive as you'd think as you can pay by Compute Units, not just having to own and maintain a huge server farm.
That's still in terms of using massive amounts of computing power for a single vehicle. Imagine trying to do that on 12 cars whilst also running all the rest of the physics and graphics tasks on a single machine. It's not viable.
There's a massive difference in terms of 'near' realtime cfd using massive server clusters to running it actually realtime with interactions. You're talking needing to do simulations in 10-20ms (less really, because the GPU you're trying to do it with is flat out with rendering tasks, you probably want 1-2ms of overhead max) with some decent accuracy on massively complex models, not 2-3 or 30 seconds with huge cell sizes like those links.
You have to remember 0.1% of performance difference is a discrete differentiator in racing that wouldn't be any sort of issue for what they're working on.
Is it viable? Sure. At some point. But we're 10-20 years away from enough raw compute easily.
That's still in terms of using massive amounts of computing power for a single vehicle. Imagine trying to do that on 12 cars whilst also running all the rest of the physics and graphics tasks on a single machine. It's not viable.
There's a massive difference in terms of 'near' realtime cfd using massive server clusters to running it actually realtime with interactions. You're talking needing to do simulations in 10-20ms (less really, because the GPU you're trying to do it with is flat out with rendering tasks, you probably want 1-2ms of overhead max) with some decent accuracy on massively complex models, not 2-3 or 30 seconds with huge cell sizes like those links.
You have to remember 0.1% of performance difference is a discrete differentiator in racing that wouldn't be any sort of issue for what they're working on.
Is it viable? Sure. At some point. But we're 10-20 years away from enough raw compute easily.
I'm not talking about doing it on 12 cars, I'm talking about having a one car simulator to run test laps that can do discrete CFD outside of the rules and provide data. COmputing power is easily getting very close to being very complete. but even then I wasn't talking about CFD on the entire car, it is entirely possible to use CFD on certain aspects, so more than possible.
But the resolution you'd have on a simulator would be pretty much completely useless compared to your other models, it's not going to tell you anything.
