|
[Sponsors] |
November 11, 2020, 02:02 |
Entry level desktop workstation for CFD
|
#1 |
New Member
Upeka
Join Date: Sep 2019
Location: Sri Lanka
Posts: 8
Rep Power: 7 |
I am thinking of buying a Workstation with the following specs for entry-level CFD Simulations. With the budget, this is what I can go for. I will need to do transient simulations to find drag and lift variation of a cylinder at high Mach numbers (M=1). Want to know whether this type of machine can handle such a simulation (Can wait for a day or two to get results)
Core i7 (8 cores, 16 threads) RAM = 32 GB (16*2 - 3200 Hz, 4 slots available in motherboard)) |
|
November 11, 2020, 08:16 |
|
#2 | |
Senior Member
Sayan Bhattacharjee
Join Date: Mar 2020
Posts: 495
Rep Power: 8 |
Quote:
RAM looks okay. Not sure about Intel CPU. AMD Ryzen CPUs are significantly cheaper and perform just as well. However you won't observe high performance in AMD CPUs if your software uses Intel MKL library. Intel MKL intentionally hobbles itself when running on AMD CPUs. You're most probably gonna need a GPU for high speed computation (if your CFD software is capable of using GPUs) and for proper visualization. |
||
November 11, 2020, 08:26 |
|
#3 |
Senior Member
Joern Beilke
Join Date: Mar 2009
Location: Dresden
Posts: 540
Rep Power: 20 |
This will be a nice PC for entry level CFD work. Try to fill all RAM slots and not only two.
GPU computing is very much overrated in the field of CFD, so don't care about it. |
|
November 12, 2020, 10:06 |
|
#4 | |
Senior Member
Sayan Bhattacharjee
Join Date: Mar 2020
Posts: 495
Rep Power: 8 |
Quote:
This is an interesting observation. I'm not able to understand why GPU computing would be overrated for CFD. My favorite CFD presentation was that of SpaceX's combustion simulation on GPU - https://youtu.be/vYA0f6R5KAI Even NASA's FUN3D was upgraded to use GPUs - https://youtu.be/TyXhmqjGSj0 I guess many commercial CFD solvers only have OpenMP and MPI support and the supercomputers in different national computation centers aren't going to buy new GPUs. But for a single desktop user, GPUs do provide significant advantages. Last edited by aerosayan; November 12, 2020 at 10:07. Reason: fixed link |
||
November 12, 2020, 10:09 |
|
#5 |
Super Moderator
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,428
Rep Power: 49 |
Name one, with an example that applies to OPs specific use-case. And the software he intends to use.
|
|
November 12, 2020, 11:52 |
|
#6 | |
Senior Member
Sayan Bhattacharjee
Join Date: Mar 2020
Posts: 495
Rep Power: 8 |
Quote:
OP previously mentioned that he's trying to learn CFD. Currently, the system requirements for the problem he's trying to solve is rather small. However in very near future he's definitely going to want to solve more complicated problems, or solve the current problem with more refinement. Some GPU enabled open source solvers : - RapidCFD : https://github.com/Atizar/RapidCFD-dev : "OpenFOAM fork running fully on CUDA platform." - Basic Euler Solver : https://github.com/joel-ho/BES : https://joelcfd.com/projects/bes/ - OpenHyperFlow2D : https://github.com/sergeas67/openhyperflow2d : "Parallel (C++/MPI/OpenMP/CUDA) research-educational CFD code for simulation 2D (flat/axisymmetrical) transient viscous compressible multicomponent sub/trans/supersonic reacting gas flow with RANS/URANS turbulence models." The second and third examples are for compressible flows, but since OP will do his simulation in the transonic region, they might be helpful. The first example is a fork of OpenFOAM, and they have a choice between a free and paid license. |
||
November 12, 2020, 17:42 |
|
#7 |
Senior Member
Lucky
Join Date: Apr 2011
Location: Orlando, FL USA
Posts: 5,762
Rep Power: 66 |
Setup looks fine. Any hex/oct-core cpu that can reach 3 GHz speeds coupled to DDR4 memory will do fine. 2 memory sticks is also fine if you're dual channel (4 sticks just increases capacity and not throughput unless you have a quad channel setup).
GPU computing is overrated, it's not even a good place to start at for an entry-level nor a budget level task. |
|
November 13, 2020, 06:48 |
|
#8 |
Senior Member
|
I second the suggestion to fill all the available slots for RAM in order to have better performances in a multicore environment.
GPU is probably nice to have in any case if you want to have a multi-purpose machine and you later want to dig into using it somehow but, as everyone told you, that's not where you start from in CFD. |
|
Tags |
computer parts, processors, ram requirement, workstations |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
decomposePar problem: Cell 0contains face labels out of range | vaina74 | OpenFOAM Pre-Processing | 37 | July 20, 2020 06:38 |
[snappyHexMesh] Error snappyhexmesh - Multiple outside loops | avinashjagdale | OpenFOAM Meshing & Mesh Conversion | 53 | March 8, 2019 10:42 |
[snappyHexMesh] Edge Refinement | fracasce | OpenFOAM Meshing & Mesh Conversion | 3 | December 2, 2017 14:30 |
[snappyHexMesh] Adding layers goes wrong with SnappyHexMesh | Elise | OpenFOAM Meshing & Mesh Conversion | 1 | April 22, 2013 03:32 |
[snappyHexMesh] Boundary layer in a pipe | Clementhuon | OpenFOAM Meshing & Mesh Conversion | 6 | March 12, 2012 13:41 |