Guidelines for Selection of CFD Hardware
CFD simulations required the following things should be considered while purchasing any hardware for CFD simulations. Apart from the investment of CFD Software as given in the page, CFD engineers need to think suitable hardwares for CFD simulations. High end simulations need more capital investment for CFD simulations.
- For industrial simulations, they should think of at least 12 core CPUs to finish the simulation in less time
- You can select number of cores or threads as per your budget
- CFD simulations are typically CPU-intensive, so a powerful multicore processor is crucial.
- High-end CPUs from Intel (e.g., Xeon) or AMD (e.g., Ryzen Threadripper or EPYC) can be expensive but offer excellent performance.
- Prices for high-performance CPUs can range from several hundred to several thousand dollars.
- Random access Memory is essential for loading and writing of large size of data during the simulation
- CFD simulations often require a significant amount of RAM to handle large datasets and mesh sizes.
- Depending on your simulation needs, you may need anywhere from 16 GB to 256 GB for most cases.
- The cost of RAM modules depends on capacity and speed, but it’s a relatively affordable component compared to CPUs and GPUs.
- To handle large mesh sizes, higher RAM is required. The following thumb rules must note:
- 5 Million cells: 4 to 8 GB RAM
- 10 Million Cells: 16 GB RAM
- 30 Million Cells: 64 GB RAM
- 50 Million Cells: 256 GB RAM
- 100 Million cells: 522 GB RAM
- 500 Million cells: 1 TB RAM
- For complex multiphysics simulations like multiphase combustion modleing,you would need 10 GB RAM for per unit millions cells.
Graphics Card (GPU)
- the workstation must have discrete graphics cards or separate GPU units to handle geometries of more CAD features
- As per ANSYS FLUENT requirement, 4 GB graphics is required. But 8 GB is recommended for high end simulations
- For high-performance computing (HPC) machines or supercomputers, GPUs are connected with each node or CPU
- While CFD simulations primarily rely on the CPU, some CFD software packages support GPU acceleration for specific tasks.
- High-end GPUs from NVIDIA (e.g., Quadro or GeForce RTX) can be beneficial for GPU-accelerated CFD, but they can also be very expensive.
- Prices for high-end GPUs vary widely but can range from several hundred to over a thousand dollars as mentioned in the post.
- Your PC should have space for the storage of simulation files
- Fast and ample storage is essential for storing simulation data, input files, and results.
- Consider using a combination of SSD (Solid-State Drive) for speed and HDD (Hard Disk Drive) for storage capacity.
- SSDs are faster but more expensive per gigabyte compared to HDDs.
- How much simulation files take space
- OpenFOAM can create multipe files you can merge files after simulations on the HPC
- ANSYS or Converge has “.h5″ file format which occupies less space.
- ANSYS FLUENT files occupy around 1 GB of space for 1 million cells
- Example: 10 M cells will take 10 GB space for combined case and data files for combustion or multiphase simulations
- Open FOAM files would take more space than ANSYS FLUENT unless you use file compression tools
- For data storage, you should have at least 100 GB of space in the drive
Maintenance Issues of Hardware
- Keep in mind that the specific hardware requirements can vary depending on the complexity of your CFD simulations and the software you use.
- It’s a good idea to consult with experts or the software vendor’s recommendations for hardware specifications.
- Additionally, workstation vendors like Dell, HP, Lenovo, and others offer pre-configured workstations optimized for CFD simulations, which can save you time and ensure compatibility.
Workstations for CFD simulations
The following are the most advanced workstations for industrial applications.
Dell Tower 7920 Workstation
- This one the scalable workstations. we can expand RAM, Data Storage, number of Graphics card
- Multi-core processor :
- Intel® Xeon® Silver/ Gold Processors,
- Number core you can select as per availability of core in processors
- For 24 cores, there are 48 threads
- RAM 64 GB DDR4, expandable to 3 TB
- Graphics cards:
- 5 GB Nvidia Quadro P2000
- You can customize the graphics cards
- SSD storage: 480 SSD
- Sata HD : 2 TB
Lenovo Think station P920
- Multi-core processor: Intel® Xeon® Gold Processors, 24 cores, 48 threads
- RAM: 64 GB DDR4, expandable to 3 TB
- Graphics cards: 16 GB Nvidia Quadro RTX A4000
- Storage: 480 SSD, 2 TB Sata HD
How can you buy a workstation?
- New workstations are expensive, you have to invest at least 60000 to 10,00,000 INR ( 5000 – 12000 USD) for computer hardware. The cost of commercial CFD solvers is also there.
- To save cost and electronic wastage, you can buy a refurbished workstation at a cost of 25% – 50% of new model
- If you want to buy a refurbished workstation at the lowest rate in India, you can contact us: firstname.lastname@example.org