The Random Thread - Buffalo Wild Wings edition

[Johnny T]

It's a heat exchanger problem and the fluid mechanics part has problems converging. It might converge if i allow more iterations before it stops or i'll try to mess with the mesh size and type. In any case, i've getting that often today, the simulation converges quite nicely with a particular geometry, but when i try to change the position of the inlet and outlet it doesn't like it. There is also the fact that the original file was designed to be as light as possible on computer resources since it's for students and it has to run on any of the uni's computers.

Anyways, i'll have to play with the simulation parameters as the problem might be an inadequate solver, simply a number of maximum iterations too low for the max convergence error i've chosen, a mesh that is too coarse. It could be a number of things. I've actually managed to have it converge, but i get a warning that i might need another set of iterations so i'll tryto have it converge without that warning. The good thing is that the values i got do make sense, not like when i had a velocity profile that was ridiculous.

Ahh ok, if in doubt, throw more stuff (iterations) at it.

Don't worry it'll go down soon as my laptop will be lying in pieces because if this continues, i'll go berserk and start flippin tables.

On the bright side, i can say that the 920xm has been put to good use recently, since the latest version of the software i use has been made hyperthreading aware (they found out it did nothing for it so they limited the number of threads to the number of physical cores) i can oc it a little farther. I can't go all out as i would in games though as COMSOL is almost as bad as prime95 with an equivalent amount of threads when it comes to the load it puts on the CPU. I could literally run the simulation i'm running now as a stress test.

Hah, I have a T61 in pieces here...

[tijo]

Well, i'm giving it two more shots then to bed. The thing is, it might take a good 30mins before i get the error message which means i'll go to bed late. Good thing i'm staying home tomorrow. I have some reports to correct (uguu) but i can still do that from home with a nice cup of tea at least.

Converging, yay! I used the brute force approach, change the iterative GMRES solver for direct MUMPS and bam it converged. The RAM usage kinda doubled so it's a big no no for some of the uni's computers, but not a problem for me Might as well try a finer mesh for kicks.

Alright, done for tonight and the verdict is the configuration i was trying is actually worse in terms of pressure drop, equivalent in terms of flow uniformity and the amount of heat transfered is actually lower by close to half. So the simulation is working, but the actual geometry is crap. Oh well, i'll be able to say that it's a bad idea.

[Vogelbung]

Present-day CFD of any consequence on a notebook... Jeez. Wish you luck.

*Pats the on-demand cluster of Z800s*


-waits for anime CFD reference to make in-thread conversation relevant-
._.

[tijo]

Well, it's not like it's anything heavy, i mean now with a little tweaking it takes about 15 mins for the simulation to run on my 920xm. I have friends who do some real intensive stuff and it isn't on workstations, it's on dedicated HPC servers, that's what i'd call intensive.

Tried to find a CFD anime reference, but didn't find any T_T

[Vogelbung]

Well, it's not like it's anything heavy, i mean now with a little tweaking it takes about 15 mins for the simulation to run on my 920xm. I have friends who do some real intensive stuff and it isn't on workstations, it's on dedicated HPC servers, that's what i'd call intensive.

Tried to find a CFD anime reference, but didn't find any T_T

Luckily, I managed to 'sell' (or rather, long-term wet-lease as it were) my cluster when I shut down most parts of my business. Which was very lucky timing as I wouldn't have known what to do with it (and obviously also didn't want to pay to continue to host it).

For my more limited needs these days though, an ad hoc bunch of Z800's will suffice.

[Datamonger]

Random post. I will not allow this thread to die.

[Tsunade_Hime]

Random post. I will not allow this thread to die.

._.

-points at Monger-

[Vogelbung]

._.

-realises thread has to achieve typical TRT level equilibrium-

-dons anime outfit-

[Shobster]

What.
10whattheheckisthatchar.

[Datamonger]

Eddie Murphy as Kinomoto Sakura.