|
[Sponsors] |
October 2, 2003, 09:24 |
PC for CFD
|
#1 |
Guest
Posts: n/a
|
Hi all,
Could some one please just recommend the type of PC that that I need for a multiphase (Gas,solid, liquid) simulation. So far the only information I have is that the PC should have at least: -17" screen, -1 GHz processor speed, -1 GB RAM, -30 GB HDD Some additional advise please.... thanks, Eldoret |
|
October 2, 2003, 10:04 |
Re: PC for CFD
|
#2 |
Guest
Posts: n/a
|
It depends on your mesh size is. The rule of thumb is: 1 million Mesh = 1 Gb of ram, of course the the faster your procesor speed the better.
|
|
October 2, 2003, 11:36 |
Re: PC for CFD
|
#3 |
Guest
Posts: n/a
|
Hi Eldoret,
You may as well get as much RAM as the PC will take. I also recommend getting a high quality PC, don't go cheap. A CFD solver is a demanding application and will burn out cheap motherboards. Furthermore, the solver is very memory intensive, which means that RAM and a memory bus speed will have a significant effect on overall performance. Error checking RAM is also recommended. Lastly, get a decent graphics card. For OpenGL performance, the nVidia cards seem to perform well, but most high end cards will do. There is a good discussion of this on the CFX Community Forum. If you are a licensed CFX user, you can gain access to the forum from the CFX Community Site. Regards, Robin |
|
October 3, 2003, 03:31 |
Re: PC for CFD
|
#4 |
Guest
Posts: n/a
|
Hi Strudl, I am thinking of putting in 2GB RAM because i want to have as many cells as my computer can allow. Preferably 300,000 cells and above. My current budget is about US$2500.
Robin, Yes, I have an access to the Site, I will check it out. Many thanks. Eldoret |
|
October 3, 2003, 08:56 |
Re: PC for CFD
|
#5 |
Guest
Posts: n/a
|
By the way, the rule of thumb for the CFX5 solver is NOT 1 Gb for 1 million cells it is:
400Mb for 1 million cells For single phase 2-eqn turb model. For (inhomogeneous) Eulerian flow multiply that figure by the number of phases. |
|
October 3, 2003, 09:08 |
Re: PC for CFD
|
#6 |
Guest
Posts: n/a
|
Jon The mesh/ram ratio was given to me by CFX-UK, I am using CFX for my Ph.D research, beleive me I have had to halve my models just to get by on a 3.65Gb Ram computer, I have had a simultion going on for 7 days, swapping with the hard drive.
|
|
October 3, 2003, 10:17 |
Re: PC for CFD
|
#7 |
Guest
Posts: n/a
|
Strudl,
Thats is as maybe, but trust me, its correct. |
|
October 3, 2003, 10:18 |
Re: PC for CFD
|
#8 |
Guest
Posts: n/a
|
Jon,
OK then. |
|
October 3, 2003, 16:14 |
Re: PC for CFD
|
#9 |
Guest
Posts: n/a
|
Actually, memory usage is very grid and physics dependant. There is no universal rule of thumb.
-Robin |
|
October 6, 2003, 04:25 |
Re: PC for CFD
|
#10 |
Guest
Posts: n/a
|
Hi all,
allow me to make a do loop here: .._GO TO_. Rao's & Strudl's latest posts on "Memory Problem in Solver".& "Memory alocation", respectively.. _IF_ .....<...> _END_ |
|
October 19, 2003, 14:28 |
Re: PC for CFD
|
#11 |
Guest
Posts: n/a
|
You should try doing a PhD with only 256mb of memory!
|
|
|
|