[Mr.ZIMULAXION]

Hi all,

I hope you are doing very well and are at the edge of publishing some world-beating simulations.

I have a question - I am interested in building a PC for casual gaming, home office work, and pre-processing simulation work (i.e., meshing using ANSYS workbench, ANSYS Fluent, etc), postprocessing, and preparing presentations. I have gotten the latest PSU ATX3.0 is now left with the motherboard and processor.

I am eyeing the AMD AM5-7950x and Intel i9-13900K. As I am a casual gamer (not full-time as I have lots of work to do just like all of you), I want to utilize this newly built computer to allow me to use it for work while I am not using it for gaming.

The application I am using is

• ANSYS Workbench CFD Meshing tool (I want to max out the parallel cores say 14/16 cores) - estimated mesh size 3 - 7 million.
• ANSYS Fluent - I won't be running for long period, most probably 6 hours - 12 hours to test case stability before submitting to HPC.
• Matlab post-processing or processing (in parallel)
• ANSYS Mechanical - very minor as I have just started, please note that I have a university XEON machine with 10 cores (W-1290CPU), and 128 GB RAM ECC at home. And I am aware that ANSYS Mechanical is Cache heavy or RAM heavy? But Mechanical is not my main application I probably will use it less than 1%.
• I also use SolidWorks, PTC Creo, ANSYS DesignModeller and etc.

The main thing I would like to get out of this system is to shorten the time for meshing in parallel mode and perhaps check the model setup for several hours before submitting it to HPC.

Could you please advise me if I am better off with 7950x, I want to avoid doing all the complicated BIOS core-disabling activity for Intel i9-13900K for load distribution and etc, and switch it back when I am going to play the game. This is too much hassle for me.

However, if someone could advise me in the era of 2023, with the new windows 11 and the latest version of ANSYS, would it be the case where the process has already been so automated that the Intel i9-13900K no longer needs complicated manual inputs (going to BIOS to switch off P cores or E-cores) to use a stable parallel feature in Workbench or Fluent? In essence, by clicking say 16 cores, the system will be smart enough to invoke all the E-cores for task?

Thanks

Hope to hear your opinions.

MrSimulation

[wkernkamp]

I think the two processors are very close in performance and a good choice. In fact both are overkill for gaming. Gaming performance will in most cases be bottle-necked at the GPU.



For the workloads you list, it seems to me that the heaviest load will occur during the ansys CFD trial runs. These runs on the maximum number of cores will tend to bottleneck at the memory bandwidth. Both CPUs have dual channel DDR5. However the intel CPU can be tuned somewhat higher (about 7200 MT/s) than the AMD (solid out of the box 6000 MT/s). To get the higher 7200 MT/s memory speed you need to carefully select the right memory modules and they can be expensive.



You indicate that you don't want to be distracted by complications such as Performance versus efficiency cores. I presume that you are also not interested in memory tuning. That means that you are better off with the AMD processor. Make sure you get a 6000 MT/s memory kit.

[Mr.ZIMULAXION]

Thanks for your inputs! I have decided to get the 7900x3d which was launched recently and pair it with a TUF X670E mobo.

I have a question: from my understanding, the 7900x3d stated that 2 sticks of RAM can support up to 5200 MHz (default). However, based on the ability of the MOBO and the AI overclocking, etc. Will the 6000 MHz (DDR5) RAMs be stable even if I purchase from a good brand say G.Skill Trident, because the CFD is run for say 6-12 hours trial run while I am sleeping at night? It shouldn't be the case where I have to tune the voltage etc to achieve a stable system?

Everything is done by the MOBO software?

[flotus1]

Memory overclocking is one of those "tinker-y" topics. Especially with AMDs latest CPUs. Once you have more than one rank per channel, things become difficult. I.e. you probably won't run DDR5-6000 with 4 DIMMs installed.
The "X3D" models are something I would not recommend at this point. Thanks to AMDs questionable decision to put the extra L3 cache only on one CCD. It's exactly what you did not want.
Either wait for the 7800X3D, or just save some money with the non-X3D chips.