Siemens Simcenter Nvidia GPU

Simcenter STAR-CCM+ gets Nvidia GPU enabled

509 0

Siemens Digital Industries Software has been busy collaborating with Nvidia in order to add CUDA-enabled GPU acceleration to its Simcenter STAR-CCM+ 2022.1 software release. It includes single and multiple GPU support on Linux operating systems.

Siemens says that the move will bring faster turnaround times for CFD simulation on is Xcelerator platform while at the same time lowering hardware investment costs for within design and engineering organisations of every size.

The partnership explained that while CPU-based high-performance computing in conjunction with highly scalable CFD parallelisation allows for massive speed-ups in absolute times to solution, it comes at the cost of high hardware or cloud computing investments.

In turn they describe the GPU technology as a “game changer”. Siemens’ engineers have run a set of external vehicle aerodynamics simulations, which it claims that by using the Nvidia GPUs could reduce required hardware compute investments by “up to 40%” and power consumption “down to 10% of the CPU equivalent”, while maintaining identical simulation turnaround times.

Simcenter Nvidia GPU CFD
Comparison of GPU- and CPU-based equivalent simulations: GPU run time and number of CPU cores required to reach almost identical turn-around times for the same vehicle aerodynamics CFD on GPUs and CPUs. GPUs see legend. Compute-optimised CPU instance: dual-socket Intel Xeon Gold (40 cores per node)

“Siemens and Nvidia are opening the door to a whole new era of CFD simulation acceleration techniques in Simcenter STAR-CCM+. With GPU-enabled technology, users can achieve faster turnaround time of CFD simulations at significantly lower per-simulation cost​,” said Stamatina Petropoulou, technical product management, Siemens Digital Industries Software.

“This technology will enable engineers working on external aerodynamics, aerospace, building and infrastructure and civil engineering applications, etc. to massively improve their simulation throughput at equivalent hardware investments with an increased per-dollar performance of GPUs compared to CPUs​.”

Nvidia developer relations manager Niveditha Krishnamoorthy, said: “Siemens’ Simcenter STAR-CCM+ is giving an incredible boost to CFD simulations by using Nvidia GPU technology via the CUDA platform and accelerated libraries.

Siemens_Simcenter_STAR-CCM+_NVIDIA_01 copy
Siemens and Nvidia say GPU enabled acceleration will deliver faster CFD turnaround times at lower hardware investment costs

“Leveraging Nvidia GPU architecture, Simcenter STAR-CCM+ users can now run more simulations, faster, and can gain critical insights for their design and operation workflows without compromising on accuracy.”

Simcenter STAR-CCM+ 2022.1 joins a growing number of simulation tools that can be accelerated by Nvidia CUDA. These include Altair AcuSolve, Altair nanoFluid, Ansys Fluent, MSC Apex Generative Design and others.

In the past GPU-accelerated simulation has been limited by the amount of GPU memory. While the largest, most complex simulations will still likely be best run on CPU servers and clusters, the new Nvidia A100 GPU features up to 80 GB of memory.

Siemens says Simcenter STAR-CCM+ is only a first milestone and it plans to extend GPU-enabled acceleration across ‘all relevant core solvers’ .

Leave a comment