As you might expect, CPU vendors are not willing to give up GPU, and there are many applications that are most promising. It should not be forgotten that GPU is not only intended for graphics performance, but there is also a growing trend to use it for general purposes. GPUs are still evolving, but it is important to remember that they are not just designed for graphical performance.
For example, Intel has released the Intel Integrated Graphics Integrated Core (IGC) and the Intel Open Compute Engine (FINE), which can combine multiple cores with common programming tools and methods. To achieve this, Fine Open uses hundreds of NVIDIA K20 GPUs that are combined in a single time step.
This means that only a few days have been shortened compared to previous methods, which could take up to a week to complete similar calculations. We used the CPU Booster, which enables fast convergence, and we used the 2,000-time steps per GPU or 1.5 times the number of cores. Each time step can be completed in less than a day, including solution letters, which means we can reduce the mass flow from the design point to a stable state in less than two hours, resulting in a total data set the size of 20 terabytes.
AmgX - NVIDIA Collaboration with ANSYS
The performance increase is due to an innovative GPU accelerated solver developed by NVIDIA in collaboration with ANSYS under the name AmgX. GV100 combines unprecedented double precision performance so that users can run simulations throughout the design process and create realistic multi-physics simulations faster than ever before. ANsYS (r) Fluent (r) provides multi-GPU support for increased productivity in CFD simulations.
In this context, the evolution of the Intel (r) processor technology has enormously increased processor performance in CFD applications. While the problem of the performance of individual processes has been partially addressed at the application level with orderly algorithms such as Cuthill and McKee, it can be laborious to fully exploit the performance of the process given the number of calculations involved in the CFD solver. An implementation of Fluent can accelerate the flow rate of a CFD simulation by up to 50% and in some cases up to 100%. Pressure - based coupled flow releases, such as scFLOW and NVIDIA's AmgX.
How Does RAM (Memory Usage) play a role in CFD Simulation
If you have enough RAM, you may run out of memory in the middle of a complex study and end up with no solver. Xeon (r) processors offer more and more cores, memory, and a lot of storage space. Additional memory can speed up your study while ensuring that the memory speed matches that of the CPU.
The effects of ram speed are difficult to measure effectively, making price and performance more difficult to determine. Faster RAM can improve performance, but large amounts of very fast RAM can be expensive. We recommend choosing the fastest RAM that suits your budget, and in case of doubt choose the most expensive RAM available.
Determine whether there is a way to physically occupy the RAM on the board, e.g. with a hard drive or an external drive.
At this point, it should be noted that the improved parallel performance within the node leads to a significant increase in the number of cores per storage unit (MPP) of the given number. Intel Xeon scalable processors have a lower MP P compared to their Xeon processors, as clearly shown by the performance difference between Intel's Xeon Scalable and Xeon E5 - 2650 processors.
In this study, the improvements in cache utilization that the HiFUN solver exploited, leading to super-linear performance, can be attributed to Intel (r). MPI library optimized for powerful memory management and the use of the GPU cache.
The higher core density of the HiFUN solver based on scalable Intel Xeon processors is ultimately due to the Intel (r) MPI library, which is optimized for powerful memory management and the use of GPU cache. Higher core densities improve parallel performance within nodes and should allow users to build more compact clusters with a specific number of processor cores. In summary, we have shown how to improve the performance of a single node of a Hi-FUND solver by increasing the number.
If the available RAM is below the model requirement, the solver must resort to file exchange, which significantly slows down the analysis. We have equipped our machine with 64 GB and use Flex Cloud Solving for larger models as it has a capacity of 120 GB.
SSDs, Improve Disk Drive Performance
Most workstations are designed to balance capacity against performance—at least when it involves disc drive storage. Because engineering files and workloads can have massive file sizes, workstations are normally equipped with large-capacity hard drives, which intentionally may offer less performance than smaller-capacity drives.
Recent improvements in drive technology, namely within the sort of solid-state drives (SSDs), have about eliminated the dimensions vs. performance barriers. SSDs are available large capacities and offer many benefits over traditional spindle-based hard drives.
SSDs can eliminate the bottleneck created by traditional platter-based drives, and thoroughly optimize system performance. High-performance workstations, especially those used for CAD, simulation, or graphics design, are normally equipped with the highest-performance CPU that creates sense relative to budget.
Nevertheless, disk IO performance is often even as important in daily usage to scale back the quantity of your time that's required to finish employment. Because time is literally money for many workstation users, getting the present job finished and on to a subsequent article of labor is important.
Traditionally, performance workstations have used small computing system interface (SCSI) drives with very high rotational speeds (15,000 rpm) to beat the difficulty of disk IO latency. Combining the best-performing drives during a redundant array of independent disks (RAID) 1—or for more performance, a RAID 0 Stripe—creates an efficient, albeit pricey solution to the performance challenges inherent with traditional mechanical drives.
SSDs aren't hampered by mechanic latency, making that an irrelevant benchmark to point out the performance differences between SSDs and traditional drives. Performance is best judged with IOs per second, and SSDs trump physical drives therein category.
That performance advantage is compounded even further when SSDs are utilized during a RAID. Generally, RAID 1 or 0 is the norm for workstation use, though in some cases RAID 5 or other configurations could also be warranted.
SSDs adding capacity via RAID doesn't increase latency. If anything, performance is improved as more drives are added—thanks to striping and mirroring technologies that are normally performance-hobbled by physical drives, which don't suffer mechanical latency issues when SSDs are used.
As SSDs still fall in price, the technology is sensible for those looking to maximize simulation performance, where large disk writes and reads are the norms.
How GPU has stepped in CAE Simulation
Nothing affects the visual performance of a simulation quite a graphics card. While tons of calculation and data movement goes on behind the scenes during a simulation, it's the visual representation of the simulation that has a lasting impact on the observer. Whether it's a simulation of flow, perspective or stress points isn’t really the important part here; it all comes right down to how quickly, smoothly and accurate the representation of that simulated event is.
Workstation graphics cards aren’t exactly cheap and convince be a big investment when upgrading a workstation. However, understanding the worth of that investment may be a little more complex—it is entirely possible to overinvest in graphics card technologies.
The trick is to correlate needed performance against overall graphics card capabilities, which may be determined with the assistance of the simulation software vendor. Most simulation software vendors offer recommendations on graphics hardware and truly do a reasonably good job of correlating simulation performance to a given hardware platform.
That said, with some experimentation and in-depth understanding of graphics cards or GPUs, users could also be ready to pick better-performing products than what's suggested by the seller. It all comes right down to what to seem for during a graphics card.
Once again, the simulation software used dictates the simplest choice. for instance, software that depends upon 3D capabilities or other advanced imaging needs may dictate what sort of card to settle on. Luckily, there's a mess on the market.
Simply put, a GPU manages how your special effects process and display. because of multiprocessing, it’s typically more efficient than a CPU on software that's designed to require advantage of it. The GPUs that are best optimized for professional graphics-intensive applications, like design visualization and analysis, are found in workstation-caliber AMD FirePro and NVIDIA Quadro graphics cards.
Professional 2D cards can manage some 3D processing but aren't optimized for normal 3D applications. they typically aren’t compatible with engineering. For professional-level simulation work, a Quadro or FirePro 3D add-in card is perhaps a requirement. Each of those product lines includes approximately half-a-dozen models that fall under four product categories, like entry-level, mid-range, high-end, and ultra-high-end. (Desktop Engineering will review a variety of graphics cards in an upcoming issue.)
There are always exceptions, but most buyers will want to match the performance and capabilities of the GPU with the remainder of the system—that is, an entry-caliber card for an entry-caliber workstation. Achieving good balance, where each component hits a performance level that's supported by the remainder of the system, is that the best thanks to maximizing ROI for your workstation purchase and optimize your productivity.
- https://www.nvidia.com/en-us/data-center/gpu-accelerated-applications/ansys/ 0
- https://www.engineering.com/DesignSoftware/DesignSoftwareArticles/ArticleID/18406/NUMECA-Brings-Massive-Boosts-to-CFD-Simulation-Speeds.aspx 1
- https://knowledge.autodesk.com/support/cfd/learn-explore/caas/sfdcarticles/sfdcarticles/Hardware-recommendation-for-Simulation-CFD-2016.html 2
- https://www.flow3d.com/hardware-selection-for-flow-3d-products/ 3
- https://www.digitalengineering247.com/article/optimize-workstations-for-faster-simulations/ 4
- https://www.javelin-tech.com/blog/2016/09/solidworks-simulation-hardware-recommendations/ 5
- https://techdecoded.intel.io/resources/parallel-cfd-with-the-hifun-solver-on-the-intel-xeon-scalable-processor/ 6