Jump to content

Help building workstation server for Finite Element Analysis

Hello there!

I was requested to buy a new workstation for my work, but I'm having a hard time since I'm not used to professional hardware. The only build I ever did was my personal gaming desktop last year. Anyway, the workstation would be a rack-mounted server that would be accessed remotely by two users that might perform finite element analysis (FEA) simultaneously. The analysis will be mostly structural, thermal and fluidic and the program will be ANSYS. My main doubts are:

More cores or faster cores?

One or two processors?

nVidia Quadro or Tesla?

Is ECC memory needed?

And, as for gaming PC, checking for compatibility is pretty straight forward, but I've been told that this is not the case with professional grade hardware compatibility, is it true?

I have a budget around US$ 5500.00 including shipping (to Brazil).

 

Any help will be very much appreciated, thanks

MoBo                           ASUS TUF GAMING X570 Plus/Br
CPU                             AMD Ryzen 5 3600x
CPU Cooler                 AMD Wraith Spire
Memory                       2x Corsair Vengeance LPX 8GB DDR4 3200MHz C16
SSD                             Samsung 850 EVO 500GB
SSD                             Kingston SSDNow V300 120GB
HDD                            Seagate Barracuda 3TB 7200RPM
Video Card                  EVGA nVidia GeForce RTX 2070 Super KO Gaming
Case                           Corsair Carbide 600Q
PSU                            Corsair CX750M
Optical Drive               ASUS DVD RW
OS                              Windows 10 Pro 64-bit
Keyboard                    Corsair K70 LUX Cherry MX Brown
Mouse                         Corsair Sabre RGB
Speakers                     JBL Peebles
Monitor                        LG 23MP55HQ 1080P 60Hz 5ms 23"

Link to comment
Share on other sites

Link to post
Share on other sites

2 minutes ago, VaPaL said:

Hello there!

 

I was requested to buy a new workstation for my work, but I'm having a hard time since I'm not used to professional hardware. The only build I ever did was my personal gaming desktop last year. Anyway, the workstation would be a rack-mounted server that would be accessed remotely by two users that might perform finite element analysis (FEA) simultaneously. The analysis will be mostly structural, thermal and fluidic and the program will be ANSYS. My main doubts are:

 

More cores or faster cores?

 

One or two processors?

 

nVidia Quadro or Tesla?

 

Is ECC memory needed?

 

And, as for gaming PC, checking for compatibility is pretty straight forward, but I've been told that this is not the case with professional grade hardware compatibility, is it true?

 

I have a budget around US$ 5500.00 including shipping (to Brazil).

 

 

 

Any help will be very much appreciated, thanks

 

 

look on supermicros website they doo some good barebones rackmount systems

6600K - ASUS Z270i Gaming ITX - 8GB Corsair  Vengence LPX DDR4 2400MHZ - EVGA 1070SC - 120GB HyperX Savage SSD - CX430 PSU:|

PSU tier list- 

 

Link to comment
Share on other sites

Link to post
Share on other sites

-For the kind of work you're doing, more cores is better 

-Two is better than one obviously, so if you can squeeze a second CPU and a dual socket mobo in there it'll be worth it

-i think a Tesla will be a better bet since Quadros re better off for cuda developers and such.

-I think ECC is essential in your case.

 Crust : Intel Core i5 4690K @ 4.4Ghz 1.45v  |  MotherboardMSI Z97 MPower  |  Fruity FillingMSI GTX 960 Armor 2Way-SLI |  CoolingNoctua NH-D15  |  RAM : 16GB Corsair Vengeance 1600Mhz | Storage : 2xSamsung 840 EVO 500GB SSDs Raid-0  |  Power Supply : Seasonic X-Series 1250W 80+Gold  |  Monitor : Dell U2713HM 27" 60Hz 1440p  |                                                                                                                                           

Link to comment
Share on other sites

Link to post
Share on other sites

So I ended up with this build, just a little bit over budget. What do you guys think?

 

CPU: 2x Intel Xeon E5-2640 v4 (10 cores, 2.4GHGz, 25MB Cache)

Motherboard: Intel Server Board S2600CW2R

GPU: nVIDIA Quadro M2000

Memory: Kingston ValueRAM 128GB DDR4 2133MHz ECC (8x16GB)

HD: Seagate ST8000AS0002 Archive HDD 8TB, 128MB Cache

SSD: 480GB Kingston SV300S37A/480G

PSU: Corsair CX750

MoBo                           ASUS TUF GAMING X570 Plus/Br
CPU                             AMD Ryzen 5 3600x
CPU Cooler                 AMD Wraith Spire
Memory                       2x Corsair Vengeance LPX 8GB DDR4 3200MHz C16
SSD                             Samsung 850 EVO 500GB
SSD                             Kingston SSDNow V300 120GB
HDD                            Seagate Barracuda 3TB 7200RPM
Video Card                  EVGA nVidia GeForce RTX 2070 Super KO Gaming
Case                           Corsair Carbide 600Q
PSU                            Corsair CX750M
Optical Drive               ASUS DVD RW
OS                              Windows 10 Pro 64-bit
Keyboard                    Corsair K70 LUX Cherry MX Brown
Mouse                         Corsair Sabre RGB
Speakers                     JBL Peebles
Monitor                        LG 23MP55HQ 1080P 60Hz 5ms 23"

Link to comment
Share on other sites

Link to post
Share on other sites

  • 4 years later...

I understand that this thread is quite old, but wanted to put some more information for reference of anyone who comes across the thread's subject. FEA solvers utilize computational hardware differently. An implicit based solver (ANSYS is primarily based on this method) uses more memory than cores. For this reason faster CPU's are better than number of CPU's. Also, as stated above, memory capacity of the CPU is a big factor as implicit solvers are very memory intensive. Implicit solvers can also take advantage of GPU power very well, but the code must have some type of feature that will enable the use of the GPU capabilities.

 

Explicit solvers are able to run more quickly with a higher number of cores, but you can not take the full advantage of this without a parallel processing capability. There are many LS-Dyna conference papers that discuss this trade off using their SMP vs MPP solver codes. Basically, you have marginal gains using an SMP solver with more than 8 cores, but can take better advantage of more cores when using the massive parallel processing code (MPP).

 

To best answer the question posed, one needs to do the research on the specific solver method to be used most often and how that uses computational hardware. A good balance can be achieved with modern rack mounted servers to tackle either solver type, but license limits also can be a factor in buying something with a lot of cores.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×