What Is High-Performance Computing for Automotive?
High-performance computing (HPC) for automotive is an optimized version of HPC aimed at providing the levels of computing power and software compatibility necessary for the needs of the vehicle manufacturing industry.
The software-enabled precision engineering that goes into manufacturing a modern vehicle requires a considerable amount of compute performance. From the design stage to feature testing to safety simulation, HPC can deliver the required processing power.
There is an increasing focus on software-delivered features within automobiles themselves as well. Cars are transforming into Software-Defined Vehicle (SDVs) with a CASE (connected, autonomous, shared, electric) vision, where the features enabled through code tie together the mechanical capabilities. Developing and maintaining these software features requires constant use of HPC infrastructure.
Examples of High-Performance Computing in Automotive
There are multiple applications for HPC in automotive, ranging from the design and engineering phase to everyday bug fixing and maintenance to developing new features for the future.
- Designing new vehicles typically starts with computer-aided engineering (CAE). While a lot of this takes place on local workstations, vehicle designers are increasingly switching to cloud-based virtual desktop infrastructure (VDI), which now also offers the necessary graphics acceleration.
- Testing new designs used to involve building models and prototypes for physical assessment, but now a lot of this work can be performed using simulations. Aerodynamics can be optimized using Computational Fluid Dynamics (CFD), initial crash testing can be completed without having to destroy physical vehicles, and the interaction of mechanical components can be simulated.
- Connected vehicles can receive Over-the-Air (OTA) updates from a remote central server but can also contribute diagnostic data back to the central server. HPC is then employed to analyze this data, often with the help of AI/ML, to develop improvements that will be delivered in a future OTA update.
- Advanced Driver-Assistance Systems (ADAS) and future autonomous capabilities require constant optimization. Developing safe, effective self-driving systems requires a massive amount of compute power to crunch the data from real-world driving situations and build models from this. Only HPC can deliver the level of performance required.
Benefits of Automotive High-Performance Computing
Harnessing HPC during the design phase of automotive manufacturing can significantly speed up the process, enabling a larger number of iterations to optimize the vehicle and reducing the need for testing physical models and prototypes. This ultimately results in a better final product, delivered more quickly. For example, the combination of CFD and HPC has enabled modern cars to be much more aerodynamic than previous vehicles.
In the era of connected vehicles, HPC also provides an enabling role. As with many products and services that have been through digital transformation, automobiles have become data-centric, from navigation systems with live traffic information to dynamic diagnostics to context-aware ADAS. The application of HPC has allowed these services to improve rapidly and become critical aspects of what gives an automotive brand its value.
Types of High-Performance Computing
There are a few different types of HPC, although they will be based on similar underlying hardware. HPC can be delivered via a dedicated on-premises data center, or it can be delivered via the cloud. The latter can be private (cloud delivered via on-premises infrastructure) or public. A company that uses private cloud infrastructure based on standard implementations can also “burst” into the public cloud seamlessly when extra processing power is required, without having to pay for this to be available on premises permanently, even when not in use.
Either way, the underlying hardware will be high-density servers offering industry-leading CPU core counts and clock frequencies, enabling the most possible computing power per node. On top of that, there will be a focus on CPU, GPU power, or both. Workloads involving AI/ML and CFD can harness GPU power extensively, so HPC aimed at these applications will include considerable GPU provision alongside the CPUs. VDI aimed at remote design workstation applications will also include powerful GPUs.
Check Out Our Other Ultimate Guides
Embedded Systems Security