Grad School Flow Simulation PC Build

Click For Summary
The discussion centers on the need for a powerful PC build suitable for running fluid mechanics simulations, particularly in the context of pursuing a PhD in mechanical and aerospace engineering. While the current desktop setup, featuring an AMD Ryzen 3 3200G processor, Nvidia GTX 980 graphics card, and 16GB of RAM, is adequate for basic simulations, there is a consensus that more advanced computational fluid dynamics (CFD) work will require significantly more processing power. Recommendations include investing in a high-performance graphics card, increasing RAM, and considering portable workstations like HP Zbooks. Furthermore, users are advised to leverage remote high-performance computing clusters for extensive simulations, as serious CFD tasks often necessitate 100+ cores, while local machines may suffice for smaller runs and pre/post-processing tasks.
AJSayad
Messages
27
Reaction score
6
Hey everyone,

I'm coming to the end of my Mechanical Engineering B.S. degree and I've been in the process of applying to mechanical and aerospace PhD programs. I want to do research in fluid mechanics; I have some expereience in hypersonics from a Research Expereince for Undergraduates NSF program so I'm thinkning about sticking to that specific field for now.

I was wondering if anyone has any experience in what types of PC builds would be strong in running flow simulations. My current desktop can run some simulations well and is great for what I need it to be for now, but when I get to grad school I feel as though I may need something a little more powerful (or maybe not, I'm not too sure haha). Any advice on what I should look for, build around, or any insight in general on the topic would be awesome! Below are my current desktop specs.

Thanks for the help!

OS: Windows 10 (64 bit)
Processor: AMD Ryzen 3 3200G
Graphics card: Nividia GTX 980
RAM: 16GB (x2 chips)
Motherboard: ASRock B450M
 
Computer science news on Phys.org
You should do pre and postprocessing on your local machine, and do the real work on a remote high performance cluster. You should get access to one when you enroll in a program that needs serious CFD work. So I would invest in a good graphics card, and lots of memory. Maybe an hp zbook so it's portable... And two good large monitors.
I have 16 cores locally, but I almost never use it for simulations, mainly for meshing and setup/small runs. For serious CFD work, you will need 100+ cores anyway. But if you do research on 2D geometries, I guess you can get away with 16 cores.
 
  • Informative
  • Like
Likes david2, AJSayad, berkeman and 1 other person
That sounds great thanks for the input! I didn't realize it could take up to 100+ cores, I'll defintely look into some options and plan for pre/post processing locally.

Thanks for the help!
 
After playing around a bit with making star maps, I found another video which introduced me to another trick that can be done using geometry nodes. This one involves fluids. For background, Blender has a fluid physics simulator, which, as the name suggests simulates the behavior of fluids. What this person did was to use geometry nodes to convert the fluid into a point cloud. Then he added a turbulence force field to "stir it around". He was then able to extract the velocity information...

Similar threads

Replies
4
Views
2K
Replies
4
Views
2K
Replies
2
Views
1K
Replies
5
Views
5K
Replies
7
Views
6K
Replies
2
Views
2K
Replies
2
Views
2K
Replies
9
Views
3K