CFD Reliability

[Belatti]

I have tried to put a sport proto (a guy helped me to model it) in a CFD soft, but my dual core note pad time to mesh it (2cm mesh) was 60hrs and the time for solving it was 250hrs...

Imagine if I try to run 3 tests with different rear wings and diffuser configurations...


I guess it doesnt make sense to try with a bigger mesh... imagine a "rounded" sport proto to be calculated like a.... prism!


I´ll have to wait till AMD or Intel sells a 32 core 5 Ghz CPU with 1 Terabyte of RAM for less that $1000

[BrendanMcF]

Friend of my wife works for [large PC manufacturer] doing corporate sales, two of her clients are F1 teams. They regularly order 40-50 new high end servers a month, and as the PC company are a technology partner with one of them, they get them free, and in 4-5 days not the 2-3 weeks that their paying customers have to wait.

These are mostly destined for CFD work.

[JamesH]

Originally Posted by BrendanMcF

Friend of my wife works for [large PC manufacturer] doing corporate sales, two of her clients are F1 teams. They regularly order 40-50 new high end servers a month, and as the PC company are a technology partner with one of them, they get them free, and in 4-5 days not the 2-3 weeks that their paying customers have to wait.

These are mostly destined for CFD work.

Blimey. That's a lot of oomph.

What happens to the old machines when they fall out of the back of the server farm? Got any spare?

That said, the current crop of graphics cards now have the ability to parallel process non-graphics but compute intensive tasks, so if the CFD software people take advantage of that, the compute times on normal desktops will drop dramatically.

See http://en.wikipedia.org/wiki/OpenCL

Doing a search on OpenCL and CFD shows loads of stuff being done. From one of the tests...

"For a variety of mesh sizes, an average speed-up factor of roughly 9.5x
is observed over the equivalent parallelized OpenMP-code running on a quad-core CPU,
and roughly 33x over the equivalent code running in serial."

[JohnD]

"40-50 new high end servers a month"
These things go out of date, they don't wear out or go off, like fresh fruit.
Can any computer guru explain why anyone needs over 500 hispec PCs a year?

"They get them free" Could it be that selling them on is a nice little earner?
Off the back of a lorry at GP circuits -"get yer luvly fresh PCs 'ere! Never seen a byte!"

John

[icemachine]

In my company the latest and greatest comps go to the CFD and FEA guys, who hand them down to the Designers, who hand them down to the Test or Project Engineers. I doubt there is anyone in this company other than students working on something older than 2 years and we're just a small R&D group of a large autoparts company

[Joaquin Topiso]

Originally Posted by BrendanMcF

The biggest problem with CFD is that it is by definition an inexact science.

Same for tunnels, unless you match Reynolds number, which you can't at the moment with current 40-50% tunnels.

[R59]

So far it looks like Nick Wirth's car is the best of the new boys. We'll see how it develops of course.

[old man]

Good article with interesting side bar:

http://www.autosport.com/features/article.php/id/2652


I am keen to see the "proof of the pudding" but not sure how we, as mere spectators, will be able judge, any views you technocrats?

[old man]

As is clear, there were no answers to the question posed above and now, half way through the season the on track answers have not been exactly clear but Lotus, who admit to a combination of CFD and wind tunnel work, would appear to be leading the new boys. This may be down to budget but in the article link above Wirth intimates that the use of CFD design is the only way they can compete with the budget they have got.

So, the question now is, is the CFD only point proven now that we have tasted the pudding or is it clear that you need actual, as well as virtual testing data for success?

[Notso Swift]

I waould also argue that MG has more current "information" than NW as to what works for a modern F1 car


At least we know what not to do... have no budget what so ever... HRT come on down

[old man]

Interesting comments from Rubens B on the Williams tunnel on the Autosport website. He feels they are now getting abetter translation of wind tunnel data for actual driveability on the track, he mentions the relationship between straight line speed and aero and cornering. I seem to recall problems with some LPM cars in that relationship and note that Peter Ellary has joined Lotus. For these new teams the testing ban is the real problem, worse for them that the established teams with lots of track experience.

As you say Notso, MG vs NW there could be significant

[LCM]

Originally Posted by R59

So far it looks like Nick Wirth's car is the best of the new boys. We'll see how it develops of course.

Think we've got the answer to that now....................

[old man]

Well now, it seems Virgin have been forced into a rethink and the early posts on here that pointed out that CFD was just a tool in the box were always going to be correct. This is not the end of the CFD story of course but I for one conclude that practical testing of the results of computer modelling is essential, a lesson I shall take into the day job.

[tristancliffe]

It could be that CFD on its own is actually the best option nowadays, but the people that used the CFD weren't good enough. Maybe a wind-tunnel Virgin, designed by the same people, would actually be slower. We'll probably never know (and in any case, I'm playing devil's advocate here and I do believe that validating one's results in another medium is critical to avoiding problems).

[GolddustMini]

Originally Posted by tristancliffe

It could be that CFD on its own is actually the best option nowadays, but the people that used the CFD weren't good enough. Maybe a wind-tunnel Virgin, designed by the same people, would actually be slower. We'll probably never know (and in any case, I'm playing devil's advocate here and I do believe that validating one's results in another medium is critical to avoiding problems).

Hopefully one of the first things that symonds tells them to do is to get to moral or something and see if the figures match, atleast then they'll have an idea of where they are

[LCM]

Quiet on here isn't it?

Where are all those who boldly asserted that pure CFD was the only way and that Virgin were assured of success?

[tristancliffe]

I don't think anyone asserted that, did they?

[LCM]

Originally Posted by tristancliffe

I don't think anyone asserted that, did they?

Really!

How do you know that nobody asserted that and can you produce the evidence?

[GolddustMini]

just realised that on my post above "moral" should read "MIRA" stupid phone, corrects me all the time!