2022 F1 car model

Why F1’s “face-melting” computing power is the game-changer behind its new 2022 rules

2022 F1 season

Posted on

| Written by

Formula 1 will today reveal the first full-size model of a car designed to its radical new technical regulations for the 2022 season.

The new rules, which were intended for introduction this year but delayed as a cost-saving measure due to the pandemic, are the product of extensive research into how to make it easier for F1 cars to race closely together.

This is a problem Formula 1 has tried to tackle before, with limited success. In 2009 new rules devised by the Overtaking Working Group were introduced to improve the racing. Its success can be gauged by the fact just two years later F1 resorted to another, more drastic, overtaking aid in the form of the Drag Reduction System.

Reducing downforce, and therefore the turbulence which makes it hard for cars to follow each other closely, has long been seen as key to improving racing. But F1 has faced the repeated problem that attempts to cut downforce have been thwarted by the immense development power of the teams.

F1’s computing performance is “insane” – Smedley
But now it has two advantages on its side: The rules increasingly restrict how much development teams can do, while F1 can harness impressively new levels of computing power, which one of the people running it describes as “face-melting”.

Former Williams and Ferrari engineer Rob Smedley is now Formula 1’s director of data systems. He says F1’s development power is “orders and orders of magnitude bigger” than what teams have access to under the aerodynamic testing restrictions. This was crucial to devising new rules which would finally pin teams down to producing cars which can race each other closely without sacrificing F1’s high performance levels.

“That’s why we at Formula 1 were able to support in such a crucial way the FIA in defining these regulations. We were able to go outside of that aerodynamic test restriction and we were able to say, right, this has all got to be about compute power.”

While teams’ design efforts are focused on the performance of a single car in isolation, Smedley says F1 knew it needed to simulate the behaviour of two cars in close proximity. The cost of doing this in a full-scale wind tunnel would be prohibitive, but Computational Fluid Dynamic simulations offered a solution.

“If you want a two-car CFD – one car following the other, full-car CFD simulation – and you want to be able to rapidly iterate across lots of design concepts, this is going to have to be quicker than what we could do under the aerodynamic test restrictions, where typically half [scale] of a single car will take you four hours.

Advert | Become a RaceFans supporter and go ad-free

“We were into four days, when we first started off, four days to do a single simulation or a single design iteration, which is just not efficient enough. So that’s why we partnered with AWS.

“Finding that solution, the thing just runs like a machine now, but [this] to me is a really, really impressive stat and a really impressive way that we’ve been able to use the partnership to such great effect: We are down to like seven hours now per simulation, per design iteration, which is just insane.

“If you’d have told me that when I was in the team environment, even going back, technology moves so fast. If you were talking about that five years ago, I’d have just thought you were mad.”

Formula 1’s simulations of cars built to the new rules – and the effect the aerodynamic wake of one has upon another – involved the analysis of over 550 million data points.

“When we were pushing through the peak number of jobs in one day we were at seven-and-a-half million cells to calculate across the CFD mesh and around about 7,000 cores we were spinning up at the same time in the EC2 service,” says Smedley. “That is just face-melting compute power.”

Advert | Become a RaceFans supporter and go ad-free

The rules, which are ultimately defined by the FIA, went through further development in conjunction with teams. F1’s chief technical officer Pat Symonds, who has decades of experience with championship-winning F1 teams, was at pains to weed out any areas of the rules which could have unintended consequences.

2021 F1 car wind tunnel model, Circuit of the Americas
An earlier model of the new design was presented in 2019
“Pat Symonds and his team when they were developing the 2022 car they did lots of rounds and iterations of ‘poacher-gamekeeper’,” said Smedley.

“So you get the first proper design iteration out that you’re happy with, the first iteration of the rules, and then it’s getting people to read through that and see where the loopholes are and then tightening it again and using the CFD and redesigning for certain aspects of it. And then doing this whole process again and going through the loop a number of times to try to get to a reasonably tied down aerodynamic rule set.”

A smaller wind tunnel model of the planned car was presented into 2019 before plans for its introduction had to be put on hold. The design includes several changes specifically intended to aid close racing including simper upper surfaces, the use of large ‘tunnels’ under the floor to generate downforce and strakes over the front wheels to further improve the airflow for a following car.

We’ll have to wait until next year to see whether Formula 1’s latest attempt to aid overtaking will succeed where others have failed. But the design will be presented today has benefitted from technology which Smedley and his team could only have dreamt of a few years ago.

2022 F1 season

Browse all 2022 F1 season articles

Author information

Keith Collantine
Lifelong motor sport fan Keith set up RaceFans in 2005 - when it was originally called F1 Fanatic. Having previously worked as a motoring...

Got a potential story, tip or enquiry? Find out more about RaceFans and contact us here.

61 comments on “Why F1’s “face-melting” computing power is the game-changer behind its new 2022 rules”

  1. He says F1’s development power is “orders and orders of magnitude bigger” than what teams have access to under the aerodynamic testing restrictions.

    Exactly, there’s no pride in outrunning Usain Bolt if you chain the poor guy to the starting blocks.

    F1 got stuck in the mid-naughts because the rule makers didn’t understand technology and were scared of teams upgrading their poultry “supercomputers”. (with complete disregard for the sponsorship opportunities lost)

    1. I’m 100% onboard with the idea of a “poultry supercomputer”.

      1. Especially the Red Bull iteration which already had wings.

      2. “We’ve combined the power of 27,894 chickens to complete a game of tic-tac-toe in under 70 milliseconds. And it wins almost every time.”

      3. That gave me a chuck-chuckle.

      4. Their power is measured in terrorFLAPS.

    2. You could say the regulators were chicken.

      1. Come on they need at least one feather in their cap

        Reply moderated
    3. It’s paltry nor poultry

      Reply moderated
      1. Cluck cluck cluck

  2. Happily, Smedley’s face doesn’t appear to be melted, merely well tanned.

  3. Has Smedley become to spokesperson for AWS in F1?

    1. I thought that was Crofty?

    2. Bring back the bottom screen graphic.

    3. It all sounds a bit press release, which is a bit sad. I can’t help but feel they’re approaching this in a totally backwards way – they’re hoping to hamstring aero development by creating more and more restrictive rules. Why not use their super computer (sounds incredibly naff 90s film) to come up with a way of measuring and restricting dirty wake?

      One idea mentioned in motorsport magazine a few years back was to mandate a maximum level of downforce, which could be wind tunnel tested. The idea is that teams could adopt whatever aero design they wanted but the advantages would be found in efficiency i.e. least drag (assuming everyone reached the maximum allowed downforce level). The consequence of this would then be low drag cars, that created less dirty wake, while retaining the performance ‘big aero’ delivers.

    4. Absolutely no faith in Smedley. The aws graphics are useless. Anyone watching the action can instantly read the peking order better than the aws graphics.

  4. Reducing downforce, and therefore the turbulence which makes it hard for cars to follow each other closely, has long been seen as key to improving racing.

    Hold on… this means that making the cars slower allows better following and racing?

    1. Biskit Boy (@sean-p-newmanlive-co-uk)
      15th July 2021, 12:41

      … it’s a revelation! who knew?

      1. Shocking right. Less speed less aero, only aws could have figured that out.

    2. @krichelle It’s less about the speed & more about how you produce it.

      You could probably have cars that offered identical peak performance/laptimes to what we have now but which were better in terms of racing if the way they were achieving that performance was been generated in a different way.

      It’s the same misconception people have with dirty/turbulent air. A car that produces less or even no downforce at all will still produce a lot of dirty air because anything travelling through the air at 200mph is going to create a ton of turbulence. The problem therefore isn’t that the cars are producing so much turbulent air but more how sensitive the aerodynamics are in terms of been affected by it.
      You could put a car from say the 1970’s behind a car from 2021 & it could run close to it because a car from the 70’s isn’t as aerodynamically sensitive to that turbulent air. And on the flip side you could put a car from 2021 behind a car from the 70’s & it would still struggle to run as close to it.

      If you look at Indycar for example. On ovals they still struggle to follow closely despite having way less aero in the low drag setup & relying more on the underwing/ground effects because at 220+mph the cars are still very sensitive to the turbulent air. And that is why you still see cars just understeer straight into the wall if a car ahead moves up the track & takes all the air off the front end of the car behind. In that situation you can ever sometimes see the car behind just swap ends due to the sudden shift in aero balance.

      1. To add. There was actually some CFD modelling done some years ago which looked at F1 designs over the decades which gave a good idea of what sort of turbulence cars of the past were producing.

        https://blogs.sw.siemens.com/wp-content/uploads/sites/6/2019/09/LineUp-1.png

      2. One of the best comments on turbulence, wake and downforce I’ve ever read. Well said!!

        Reply moderated
    3. @krichelle, I was about to point out that paragraph until I saw your post. @gt-racer is spot on – the amount of downforce itself isn’t an issue, but DF generation.

  5. Quite environmental unfriendly, for a sport that promoyes to be green and carbon neutral… kinda like bitcoin

    1. Biskit Boy (@sean-p-newmanlive-co-uk)
      15th July 2021, 12:43

      which bit is environmentally unfriendly?

      1. The amount of power the datacentres draw is also “face melting”. Measured in gigawatts.

        A few of the AWS datacentres (3-4 of 13 regions) claim to carbon neutral. I’d be skeptical if that includes construction, or offsetting secondary impacts from staff and hardware production though, Greenpeace certainly dispute Amazon’s claims.

        Reply moderated
        1. 1.21 jigawatts wasn’t it :-)

      2. I suppose they mean spinning up a ton of cloud servers and burning through electricity to run these simulations, but given the context (a sport that travels the globe and literally runs on fossil fuels) I’m not sure how big a deal it really is.

        I haven’t researched it, but it’s possible that AWS gets its electric from renewable sources anyway.

        1. @sparkyamg you will find that Amazon is far less transparent about where it gets its power from than either Google or Microsoft, though the indication is that it also makes rather less use of renewable power than its rivals.

          Amazon gives a vague indication of their overall CO2 emissions, which they claim was about 60 million tonnes in 2020. They have indicated that, across Amazon as a whole, about 50% of their energy supplies come from renewable sources – whilst they have talked about eventually becoming carbon neutral as an entity, that is only in 2040 (with no published plans for how they intend to achieve that).

          Whilst there was a promise to shift their data centres to operating entirely on renewable power in the longer term, there are allegations that Amazon Web Services have quietly abandoned that goal and is lagging quite some way behind both Google and Microsoft.

          At best, maybe 50% of their power comes from either renewable sources or with measures to offset the carbon impact: at worst, some of their larger data centres are suggested to only have about 12% of their energy come from renewable sources.

      3. What does “environmentally friendly” mean anyways?

  6. i stopped reading at “So that’s why we partnered with AWS”

    1. Bring back the bottom screen graphic.

  7. I’m having deja vu — I’m sure I’ve heard the “face-melting computing power” quote before. Wasn’t Smedley at an Amazon event a couple years ago talking about all this?

    1. It is not deja vu, it’s an actual memory. Smedley was at AWS re:Invent 2019 (December 2019).
      The words “face-melting stats” are spoken at the 8:44 mark:

      1. https://www.youtube.com/watch?v=eBX7lPk5qmA
        Sorry for multiple posts — link didn’t go through in last post.

    2. @x1znet I think someone (or multiple, I recollect Brawn and Smedley at least) also did an interview as part of a piece for GP Racing magazine. In must have been nearly 2 years ago now, roughly in the summer of 2019.

  8. These CFD numbers are nonsensical and Rob Smedley is clearly just throwing out AWS marketing material. 7.5 million cells is a laughable number even by non-F1 standards. In F1 we use hundreds of millions of cells and don’t need much more than 500 cores in a few hours. I ran Formula Student car CFD cases with 20M cells on my 6 core desktop at home.

    1. @pozzino, well if you still have access to the software then you can answer me this quandary I’ve wondered in the past:
      If you were to fashion the rear wheel(s) into a fan, would you not be able to suck out the air from under the car/from the diffuser and shoot it out from the middle of the hub ? Wouldn’t that work ?

    2. Thanks for pouring a bit of cold water on the hyperbole bubble in that AWS PR message @pozzino!

  9. Not on hold, but postponed by a year, to be more specific.

  10. I’m excited about the new cars, but that excitement has now been tempered after learning that the same folks who bring us the Qualifying Pace graphics from the Free Practice sessions have determined how the cars will perform under the new designs.

    1. It’s like they polished their silicon into a crystal ball.

    2. @g-funk ..or the ‘tyre performance’ shabacle

  11. strakes over the front wheels to further improve the airflow for a following car

    Do they improve the airflow for the car with the strakes? otherwise it would be a shame if they turned out to be terribly fragile, fell off, and made your own car much harder to pass :)

    1. I thought the idea behind the strakes was to reduce the lifting effect open wheels have when travelling at speed. I see they only appear to cover half the front wheel with a strake, which must add turbulence because the air has generated lift on the uncovered half while the covered half has either no lift of less lift, consequently you’d get a vortex forming as a consequence of these two regions where differences in generated lift apply.

      1. I think that having a small winglet / strake there would actually shape and deflect part of the wake of the turning wheels there @drycrust, possibly preventing that from forming a large plume

        1. @bascb Thanks, that sounds plausible.

  12. I just hope all these brave, chest beating articles about petabytes and thousands of cores and such are remembered and bookmarked for when we see eventually see the results on track. I sense silliness in our future, and I love silliness.

  13. pastaman (@)
    15th July 2021, 19:24

    I bet the cost of all those EC2 instances is face-melting as well

    1. Cheaper than building and maintaining your own iron though

    2. I’m making a lot of assumptions about what exactly they are doing but the list price for ~7000 cores on AWS would be about $300 per hour. Any “AWS partner” would be paying a lot less than that.

  14. Worth saying that, unless the correction is from F1, when he says “where typically half [scale] of a single car will take you four hours”, scale is incorrect in this context – the mesh is symmetric so you only run one half, cutting the car down its x axis.

    Reply moderated
  15. Coventry Climax
    16th July 2021, 0:14

    F1 designed a car. Wow. What for? I thought it was the teams (except AM, ofcourse) that were supposed to design their cars, to then compete in the manufacturer’s championship that F1 actually is. So unless I’ve missed the part where it says that the teams have to race this car, which would effectively kill F1, there is no guarantee whatsoever, that there won’t be teams coming up with cars that conform to the regulations, yet look or at least behave completely different. OK, maybe not completely, but the point is, that the new regulations may aim for certain goals, but we won’t know if these goals have been met, until the first couple of races. Given the FIA’s history, I would not be surprised if things do not quite work out the way they planned. But then there’s always the option of mid-season technical directives ofcourse, to punish teams that have been smarter than the FIA. Yep, you’ve guessed right; I’m sceptical, to say the least.

    1. As far as I get it, the idea is that this way the F1 team have a good understanding of what the car does – they talk about how they were able to simulate whether it does what they want from it (be able to closely follow without too much loss of downforce) and how it helped them look at potential loopholes.

      I would add, that it enables them to add bits teams come up with to evaluate whether those ideas work against those stated goals and close potential loopholes etc.

    2. I agree CC, Formula One the business owner of the series, is converting it to a spec car series by dictating the FIA’s technical regulations. What a waste! Just have Dallara run up a car and be done with it. This sham of having ten teams all manufacture the same car and then power it by the same PU is a “face-melting” absurdity in financial and sporting terms.

      Reply moderated
  16. So excited to see the racing next year and applaud f1 for going out on a limb to create a set of rules designed to facilitate closer wheel to wheel racing.

    My only disappointment is the width of the front wing and size and length of the nose….

    Wish the nose was more 2004 Ferrari or 91 jordan or even 93/94 Williams with front wing end plate ending inside the front tyres.

    Reply moderated
  17. I feel like they have failed to take into account that teams might want make a car that actually has higher trailing turbulence as long as that does not impact their own performance. This would make it harder for competitors to follow and overtake.

    Do the 2022 rules actually dictate things like Reynolds number or just what components are allowed?

    1. I think they defined areas where no bodywork can be @yitzchak.

      And they state here that they will look at what teams bring to their cars – probably by doing what all teams do ASAP after they see a new interesting bit on competitors cars, i.e. making a 3D model and simulate what effect it has – to see whether those bits go against the stated goals. If so, they would work with the FIA to close any “loopholes” they find.

Leave a Reply

Your email address will not be published. Required fields are marked *

All comments are moderated. See the Comment Policy and FAQ for more.
If the person you're replying to is a registered user you can notify them of your reply using '@username'.