|
[Sponsors] |
October 9, 2007, 08:25 |
What will happen if
|
#1 |
Guest
Posts: n/a
|
lots of people are putting their life into CFD
but suppose within 5-10 years somehow advanced hardware comes and it can solve the problems in real time with DNS approach with high grids Then all the models and techniques will become redundant IT is dangerous OUR careers depend on slow machines........... PLs comment |
|
October 9, 2007, 08:59 |
Re: What will happen if
|
#2 |
Guest
Posts: n/a
|
hmmm... So u feel that there would have been more jobs if the fastest processor was 4846 family???
|
|
October 9, 2007, 09:46 |
Re: What will happen if
|
#3 |
Guest
Posts: n/a
|
The reason there is a CFD market at all is because of high speed parallel computers. Without them, nobody would wait 1 month to get the solution to a simple problem. If your goal in life is to be an expert mesh maker, then maybe you should worry. If your goal is to be an engineer who can contribute to better designs through the understanding gained by analysis and insight then I wouldn't worry very much about faster computers.
|
|
October 9, 2007, 09:50 |
Re: What will happen if
|
#4 |
Guest
Posts: n/a
|
RKT the great joker.
During a symposium Intel show us our view of the futur about HPC. And you know what ? Turbulence is a great chalenge for us. If we consider a PetaFlop system (system does not exist for civil research today!) you can't direct resolv all the degrees of turbulence! If we consider a 10 Petaflop system, it's impossible Maybe, we can supposing that with 1000 Petaflop, you can use DNS for Insdutrial problem! But, today, their are no numerical code and no librairie etc ... which can run on more than 100 000 procs. DNS is necessary the futur but not before a long time ... And DNS use a model: Navier Stokes model! :x |
|
October 9, 2007, 09:56 |
Re: What will happen if
|
#5 |
Guest
Posts: n/a
|
Much, much faster computers would open computation to many, many more problems, problems that may even be unimaginable today due to geometric amd material complexities and computing economics. In this way, faster computers have created more jobs for creative people.
|
|
October 9, 2007, 10:19 |
Re: What will happen if
|
#6 |
Guest
Posts: n/a
|
The answer is simple. History will repeat itself. Models will get bigger. In the 10+ years I've been associated with CFD, 3D models have always needed to run in days; 1D models in minutes.
|
|
October 9, 2007, 10:24 |
Re: What will happen if
|
#7 |
Guest
Posts: n/a
|
with increasing power of computation, our tendency to include more details are increased too. Note that universe is multi-scale: macro, meso, micro, nano, molecular, atomic, sub-atomic (electronic),
incluading each scale in real world applications needs a great jump in power of computers. Nowadays, we have problem to simulate macroscale phenomena ! read the following paper to judge about your (funny) prediction (5-10 years !!!): Moore's Law and Numerical Modeling, Journal of Computational Physics 179, 698â€"703 (2002). |
|
October 9, 2007, 10:49 |
Re: What will happen if
|
#8 |
Guest
Posts: n/a
|
I'm totaly agree with you Jonas!
|
|
October 10, 2007, 13:49 |
Re: What will happen if
|
#9 |
Guest
Posts: n/a
|
Agreed Jonas... very true.
We have a long, long way to go before we saturate computing power requirements. Try simulating an automobile - with each fin, each louver at appropriate mesh density. Need a few lifetimes to get that one. At the moment, a 3d model of such configurations takes many months to model & simulate. mw... <www.adthermtech.com/wordpress2> |
|
October 10, 2007, 13:51 |
Eratta
|
#10 |
Guest
Posts: n/a
|
... Try simulating an automobile radiator...
|
|
October 14, 2007, 11:09 |
Re: What will happen if
|
#11 |
Guest
Posts: n/a
|
you still need the guy who will be able to analyse the results. The poor guy will have more results to analyse and data to collect.
During an experiment the density can vary a bit from literature (dust, humidity, temperature) so you CFD results will be slightly offset. By how much?. For example what is the sensitivity on your results if you change the density a wee bit. Are there more differences in the results due to approximation of the inlet velocity or from the choice of turbulence models? If you have faster computer you will be able to answers those question: what if this, what if that |
|
October 15, 2007, 01:59 |
Re: What will happen if
|
#12 |
Guest
Posts: n/a
|
WHAT I MEANT TO SAY IS HARDWARE CAN IMPROVE TO THE EXTENT WE CAN USE DNS FOR PRACTICAL PROBLEMS.
SOMEHOW I FELT THAT ALL THE NEWBEES FIND IT VERY EXCITING TO TELL OTHERES THAT MY CODE TAKES ....HOURS AND COGLOMOROVE SCALE.. IT IS NOT ALWAYS IMPORTANT TO PUT BILLIONS AND BILLIONS OF GRIDS AND DISCRETISE THE WHOLE UNIVERSE. FOR FANATICS THEY CAN KEEP DOING WHAT THEY ENJOY IN THEIR LABS.... WHAT I EXACTLY WANTED TO INITIATE WAS A DISCUSSION HOW HARDWARE ADVANCEMENT WILL RELIEVE US FROM UN-NECESSARY APPROXIMATIONS. AS SIMPLE AS THIS... |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
interFoam - where does the magic happen? | lindstroem | OpenFOAM Running, Solving & CFD | 9 | August 27, 2019 23:11 |
How could it (2D LES) happen? | Daniel WEI | Main CFD Forum | 7 | August 5, 2008 11:36 |
why this ERROR happen?? | prayskyer | CFX | 6 | September 19, 2006 04:36 |
why does this error happen? | prayskyer | CFX | 1 | July 23, 2006 07:31 |
What will happen | guanjian | Phoenics | 1 | September 19, 2001 15:38 |