I'm an engineer (did undergrad in Automotive Engineering and doing a Masters in Aerospace) and I've got some explaining to do...
Simulations are by their very nature a complex and fickle breed. Basically there are two types of them: training simulations and engineering simulations.
Training simulations are simulators designed to run in real-time so that they can be used for either driver/pilot training or for very loose and low fidelity design work. They generally use look-up tables for values from values calculated elsewhere (torque/power values or tyre slip curves or aerodynamic data for example) and plug them in to linearised mathematical models to produce a fair approximation of the real thing. They need massive computers to run them. The PS3 does have a fair amount of mathematical power for pure number crunching so it can do an okay job at this but the results are never going to be that accurate. They will use lower-order approximations of the physics that are going on; removing any parameters that don't have too large an effect on the fundamental response of the car (for instance in my current field when we're approximating the roll response of an aircraft to a control input we ignore the coupling of roll and yaw so as to just have a one degree of freedom system that is perhaps 98% accurate) which reduces workload a great deal.
On the other side of the fence are engineering simulators. These are the real deal and in my current field of aerospace design it can take hours for these systems to be run on massive computer mainframes just to produce a few seconds of flight data. They tend to produce their values on the fly using either CFD or the alternatives for different parameters. These are nearly 100% accurate in real life and can completely replace prototype testing but they are a far cry from training simulators in terms of fidelity.
So, you will never be able to do a real comparison between what is fundamentally an entertainment product compared to real life and the people who do go on about physics to no end just come across as cute to me.
As an aeroengineer I can see that the main things that GT5 gets wrong are aerodynamics of both the high downforce cars (the F1 cars are all wrong) and the drafting effects....but these ARE very CPU intensive phenomena so it's understandable they've reduced them a lot - particularly when you've got multiple cars drafting each other*
*as an aside when I was doing some CFD work earlier in the year the addition of more cars in to the flow field made the calculation time more than quadruple - it took even the lowest fidelity solution hours to solve just one iteration. Can you imagine trying to do that at least 10Hz to make it workable for a game?
Simulations are by their very nature a complex and fickle breed. Basically there are two types of them: training simulations and engineering simulations.
Training simulations are simulators designed to run in real-time so that they can be used for either driver/pilot training or for very loose and low fidelity design work. They generally use look-up tables for values from values calculated elsewhere (torque/power values or tyre slip curves or aerodynamic data for example) and plug them in to linearised mathematical models to produce a fair approximation of the real thing. They need massive computers to run them. The PS3 does have a fair amount of mathematical power for pure number crunching so it can do an okay job at this but the results are never going to be that accurate. They will use lower-order approximations of the physics that are going on; removing any parameters that don't have too large an effect on the fundamental response of the car (for instance in my current field when we're approximating the roll response of an aircraft to a control input we ignore the coupling of roll and yaw so as to just have a one degree of freedom system that is perhaps 98% accurate) which reduces workload a great deal.
On the other side of the fence are engineering simulators. These are the real deal and in my current field of aerospace design it can take hours for these systems to be run on massive computer mainframes just to produce a few seconds of flight data. They tend to produce their values on the fly using either CFD or the alternatives for different parameters. These are nearly 100% accurate in real life and can completely replace prototype testing but they are a far cry from training simulators in terms of fidelity.
So, you will never be able to do a real comparison between what is fundamentally an entertainment product compared to real life and the people who do go on about physics to no end just come across as cute to me.
As an aeroengineer I can see that the main things that GT5 gets wrong are aerodynamics of both the high downforce cars (the F1 cars are all wrong) and the drafting effects....but these ARE very CPU intensive phenomena so it's understandable they've reduced them a lot - particularly when you've got multiple cars drafting each other*
*as an aside when I was doing some CFD work earlier in the year the addition of more cars in to the flow field made the calculation time more than quadruple - it took even the lowest fidelity solution hours to solve just one iteration. Can you imagine trying to do that at least 10Hz to make it workable for a game?
