The Only Thing That Counts Is Me

On Sunday, BMW scored its first 1-2 finish in Formula One since breaking its partnership with the Williams team and going it alone (by buying and absorbing Peter Sauber's team). It was a milestone for the team. Not only did they achieve their goal of winning an F1 race this year (less than halfway through the season) but their second car also came in second in the race, allowing them to score the maximum points possible.

Their driver, Robert Kubica, is now leading the Drivers' World Championship and they are 2nd in the Constructors' World Championship, only 3 points behind Ferrari and 17 points ahead of McLaren-Mercedes. Most people believe that BMW are unlikely to win either Championship this year but they have made incredibly quick progress in the last few years since leaving Williams.

In 2006, BMW finished 5th in the Constructors' World Championship with 36 points. 50 points behind Honda, over 70 points behind McLaren and over 160 points behind Ferrari and Renault. In 2007, BMW finished 3rd, over 100 points behind Ferrari and McLaren (if McLaren's disqualification is not counted) but over 50 points ahead of Renault in 4th. It was already a remarkable advance over the previous year for BMW but this year, they came in 1-2 at the Canadian Grand Prix and lie only 3 points behind Ferrari at the moment.

The normal convention in F1 is that as a team improves, so does its competitors so from season to season, everybody gets better more or less at the same rate and therefore the relative positions of the team remain fairly static. Ferrari and McLaren will always be the front runners and Williams, Toyota, Honda and Red Bull will always be in the middle of the pack. So how did BMW leap from mid-packers to challenging the front runners? Many people are beginning to think that the answer is Albert2.

In simple terms, the performance of an F1 car depends most heavily on 4 main factors - driver, tires, engine & aerodynamics. BMW has good drivers but not great ones. Neither Robert Kubica nor Nick Heidfeld have ever won the Drivers' World Championship. They are well regarded but generally considered not to be in the same class as Michael Schumacher or Fernando Alonso. As for tires, Bridgestone is today the only tire provider in F1. All F1 cars therefore use the same tires and BMW does not have a tire advantage over any other team. As far as engines are concerned, the BMW powertrain is highly regarded but probably not noticeably superior to other teams like Ferrari, Renault or Toyota. MacLaren gets their engines from Mercedes but unlike BMW who built the engines substantially in-house, Mercedes bought over Ilmor, a specialist F1 engine maker and started by using essentially re-badged Ilmor engines. Anyway, the point is that as good as BMW Power is, that's not the advantage. BMW's advance over the last few years can probably be traced to aerodynamics but Albert2 isn't an aerodynamic package or a person. Albert2 is a computer. In fact, one of the world's fastest supercomputers.

Aerodynamics is basically the airflow over a car. The sleeker the car is, the less wind resistance it encounters and the faster the car can go given the same amount of engine power. But apart from going fast in a straight line, an F1 car needs to be able to negotiate corners as quickly as possible without skidding off the road due to centrifugal forces. To do that the car is designed with a front wing and a back wing that generate large amounts of "downforce" and cause the car to stick to the road despite the high centrifugal forces during cornering. An F1 car at high speed generates so much downforce that even if the road and the car were upside down, the car will still stick to the road in defiance of gravity.

About 35% of an F1 car's downforce is generated by the back wing and about 25% is generated by the front wing. The remainder is generated by other parts of the car, every centimetre of which has to be carefully designed to produce the right balance of speed and stability. To complicate matters further, since the straights and curves of each race track are different, an F1 team must adjust the aerodynamics of their car for each race to give the car maximum performance for that particular racetrack.

When it comes to designing an aero-package, F1 teams normally go through a 3 stage process. First, Computational Fluid Dynamic ("CFD") simulations are undertaken on a computer to determine probable good wing and body contour designs. Secondly, scale models of the designs are built and put into a wind tunnel for testing. The BMW team generally build 60% size models for testing in their wind tunnel in Hinwil, Switzerland. Their tunnel can generate winds of over 300 kilometres per hour! Finally, after extensive wind tunnel testing, the F1 car is built and tested for real on racetracks.

It is well known that there is a gap between what CFD simulations can tell you and what the wind tunnel, using real wind of course, tells you. It is also well known that there is again another gap between wind tunnel simulations and performance of the real car on a real racetrack. Sometimes, irregularities which show up on the racetrack don't show up in wind tunnel tests or CFD simulations. If this happens, a team can sometimes go through an entire F1 season without figuring out what's wrong with their car's aero package and the consequences for their race performance would be drastic. A famous example was the Jaguar R3. For this reason, many teams would prefer to avoid drastic new aerodynamic designs from one season to another. Instead, they look for incremental improvements over last season's package.

In the case of BMW, at the end of the 2006 season (where they finished 5th), they already had a fast supercomputer called Albert doing CFD and a state of the art wind tunnel in Hinwil. When it comes to investing in more hardware, the conventional wisdom is that since wind tunnels are more accurate than CFD simulations, it is better to invest in a second wind tunnel if the money is available. However, instead of doing that, BMW chose to invest in a new more powerful supercomputer which they called Albert2. Team Principal, Dr. Mario Theissen, explained, "We are not going to follow the route of other teams and build a second wind tunnel. We see the future is in testing simulation. Having two wind tunnels makes no sense to us. We want to be at the top of CFD and in order to be there we need a strong tool. With Albert2 we have what we need to become the top in aerodynamics."

Albert2 uses 1,024 Intel processor cores and has a total memory of 2,048 GB. It has a maximum power of 12,288 GigaFlops i.e. 12,288,000,000,000 calculations per second. At the time of its launch in December 2006, Albert2 wasn't just the fastest supercomputer in F1. It was the fatest supercomputer in industrial use in the whole of Europe.

Many believe that BMW's rise to 3rd in the Constructors' World Championships in 2007 was significantly helped by the power of Albert2. More important than just coming in 3rd, BMW also closed the points gap to Ferrari from 160 points in 2006 to 100 points in 2007.

However, as good as their performance was in 2007, a 100 points gap was still a very large gap to close. Incremental improvements to BMW's 2007 car, the F1.07, would not allow them to close that gap in 2008 since Ferrari and McLaren would also be making incremental improvements to their previous year's cars. Therefore, for the 2008 car, the F1.08, BMW decided to take a risk. As Technical Director Willy Rampf explained, "this approach is the only option if you want to close the gap to the frontrunners. You don’t get to the top just by being conservative. The benchmark in Formula One is on a level that forces you to take calculated risks. And that’s exactly what we did."

However, conventional wisdom held that the abrupt sculpting of the middle section of such 3 piece wings meant that the downforce produced would be highly irregular. The middle section would be very sensitive and would unpredictably switch from producing lots of downforce to producing little downforce in different parts of a track curve or corner. Indeed, during initial test drives, both BMW drivers expressed serious concern about the F1.08. Nick Heidfeld's opinion was that "we have not moved closer to McLaren or Ferrari" while Robert Kubica admitted that, "The car is too unstable."

This was a very serious concern because it meant that there was a gap between what the CFD and the wind tunnel was telling them and how the car was actually performing on the racetrack. It appeared that parts of the new car were interacting with each other in aerodynamic ways that had not been anticipated and had not been experienced with the 2007 car. Rampf said, "There is no question that the F1.08 is clearly faster than the F1.07 but it is more complex and, for the drivers and the engineers, more difficult to find the full potential at all times and in all conditions."

Even after the season started in March, BMW has continued to develop their car faster than their competitors. In the 2nd race of the season (Malaysia), Kubica's fastest lap during the race was 0.516sec slower than eventual winner Kimi Rakkonen's Ferrari. In the 4th race of the season (Spain), that gap had narrowed slightly to 0.436sec. By the 5th race (Turkey), Kubica's fastest lap was now only 0.114sec slower than eventual race winner Filipe Massa of Ferrari. In the 7th and most recent race (Canada), Raikkonen's fastest lap was only 0.152sec quicker than Kubica's and this time, it was Kubica who won the race. These results suggest that BMW have closed the gap to Ferrari dramatically.

It appears that the raw computing power of Albert2 is gradually allowing BMW to tame the tricky 3 piece front wing design and start to unlock the full aerodynamic potential of the F1.08. CFD is a very calculation intensive operation. It consists of breaking up airflow into small elements and calculating how those elements interact with each piece of the car they come into contact with as well as how they interact with each other. It is estimated that Albert2 has the power to break up airflow into hundreds of millions of elements and calculate interactions for all these elements with the car parts and with each other. The more elements you can calculate with, the more accurate your CFD model becomes and the more subtleties about aerodynamics you can discover.

You can be sure that somewhere in Ferrari and McLaren, there are very high level people wondering if the raw computing power of Albert2 is allowing BMW to consistently develop their car faster than the wind tunnels at Ferrari and McLaren can. The remainder of this F1 season will be very interesting indeed.