A client engaged in research into the safety of construction work approached us about introducing a machine for simulating soil movement using Ansys Rocky. The client wanted to introduce a new machine because the specifications of their current workstation could no longer keep up with the simulation content.
|
[Current machine specs] OS: Windows 10 Pro |
Specifically, the simulation will be performed using a method called DEM/SPH, and a total of 100 to 300 million particles simulating sand and water will be handled. This large number of particles requires a GPU with high processing power and large memory capacity. In the current environment, a 10-second simulation requires a total calculation cost of about 4 days, so the goal is to perform this simulation dozens of times in succession to get closer to real experimental data.
The specifications considered by the customer for the simulation content are as follows.
|
[Expected specifications] OS: Windows 10 Pro |
Considering future spec expansion and porting the currently used GPU to a new machine, the customer would like a case that allows for the addition of one more GPU. In addition to the RTX 1, the RTX 4090 Ada is also considered as a GPU candidate, and the customer has asked whether the GPU memory capacity has a significant effect on the simulation execution time when comparing simulations that handle a large number of particles.
Based on the above discussion, we proposed the following structure:
| CPU | Intel Xeon W5-3435X 3.10GHz (3.0GHz at TB4.70) 16C/32T |
| memory | Total 256GB DDR5 4800 REG ECC 32GB x 8 |
| storage | 2TB SSD S-ATA |
| Video | NVIDIA Geforce RTX4090 24GB |
| network | on board (1GbE x1 /10GbE x1) |
| Housing + power supply | Tower type housing + 1000W |
| OS | Microsoft Windows 11 Professional 64bit |
Single and double precision
When choosing a GPU, it is important to consider whether the calculations you will be performing are single or double precision.
Provided by ANSYSANSYS ROCKY GPU Buying GuideIn the case of Geforce RTX 4090, if the shape of the particles to be handled is spherical, a GPU with high single precision is recommended. If the shape of the particles collapses or the shape is complex, double precision is recommended. In this case, the customer's simulation mainly deals with spherical particles, so a GPU with high single precision is fine. The Geforce RTX XNUMX selected by the customer is a GPU that is strong in single precision calculations, so it matches the purpose.
Relationship between GPU memory capacity and simulation execution time
Rocky GPU Buying Guide | Ansys KnowledgHere is the description of GPU memory:
If you intend to run very large cases, with millions of particles, you should consider GPUs with larger memory size.
Therefore, if your calculations fall within the 4090GB range of the RTX 24's GPU memory capacity, then the RTX 4090 will be fine.
Comparing the RTX 4090 and RTX 6000 Ada, the RTX 4090 has a higher clock speed but fewer CUDA cores than the RTX 6000 Ada. Therefore, in terms of theoretical parallel calculations, the RTX 6000 Ada has a single-precision performance of 91.1 TFLOPS, while the RTX 4090 has a performance of 83 TFLOPS, which is slightly inferior.
When comparing the benchmarks of these two GPUs, the RTX 2 has a higher score, but this is likely due to the GPU's operating clock. To maintain a high clock, the RTX 4090 requires 4090 times the power consumption of the RTX 6000 Ada.
Therefore, in cases where peak performance occurs momentarily, such as in games, the benchmark for the RTX 4090, which can run at a higher clock speed, will have a higher value. When high load is applied to many cores for a long period of time, it can be difficult to maintain a high clock speed due to heat and power issues, so the results may differ from benchmarks measured at typical peak speeds.
Support for adding GPUs
It is equipped with a 4090W power supply unit in anticipation of adding one RTX 1 in the future. If you add an RTX 1600 to the configuration in this example, you must use a 4090V power supply environment due to power consumption. Please note that you can only install up to two GPUs (RTX 200) in total.
In addition, we have added a 200V power cable for use in a 200V power environment.
In addition, whether or not you can install the RTX 2080Ti GPU you own will depend on the size of the card and the shape of the physical cooler. If the RTX 2080Ti you plan to transplant is an internal exhaust model, we do not recommend transplanting it for use due to heat dissipation issues in the machine itself. We apologize for the inconvenience, but please be forewarned.
■ Keywords・About Ansys Rocky Ansys Rocky is a discrete element method (DEM)-based software that can simulate powder behavior with high accuracy. In addition to modeling non-spherical particles and high-speed calculations using GPUs, it also supports coupled analysis with Ansys software. |
|
■ Click here for details and inquiries about this PC case
* Please enter the name of the case or your desired conditions. |
