Realize car cost minimization and safety optimization

Author: Bryce Johnstone, Director of Automotive Business, Imagination Technologies

People say that you can’t price safety, but when we look at the wide range of prices from low to high in the entire automotive industry, as well as the safety levels corresponding to different price points, this statement is not necessarily true.

This is not to say that any car manufacturer is making unsafe cars to sell to the public, but that in order to obtain the best and most comprehensive modern safety features, corresponding costs need to be paid. However, the cost/benefit analysis of automobile design and production is constantly optimized, so how can we use the existing technologies in automobiles to greatly improve safety without drastically increasing prices?

Security issues are at hand

When we think about car safety, most people immediately think of hardware devices: airbags, anti-lock brakes, and anti-collision buffers are all used in accidents to ensure the safety of people in the car. But safety is not just about surviving a collision, it also requires installing the system in place to avoid collisions and malfunctions.

The European Union New Car Safety Evaluation Association (Euro NCAP) defines the highest five-star safety rating as: “The overall performance of collision protection is excellent, equipped with comprehensive and powerful anti-collision technology.” In terms of avoiding collisions, we can not only use the latest advanced driver assistance system (ADAS) functions to achieve automatic braking and guide the vehicle to drive in its lane, but we can also seek a simpler solution-avoiding hardware failures.

Automobile mechanical failure is the main cause of accidents, so a simple solution is to remove the mechanical parts as much as possible where it is possible to remove the mechanical parts. Whenever you drive, the area in front of you (dashboard) happens to be the typical situation involved in this scheme.

For a safe driving experience, the dashboard is a vital part, responsible for indicating speed, gear and displaying all vehicle warning lights. If there is a failure here, it may cause catastrophic consequences. This means that traditional mechanical instrument panels must be designed according to strict tolerances and carefully assembled, resulting in high costs for such seemingly small vehicle components. The solution adopted by all priced vehicles is to use digital Displays instead of mechanical dashboards, but this may also cause other related problems.

Modern car monitors need to meet the ASIL-B certification of ISO 26262. This means that various appropriate measures need to be taken to support the system to check its own calculations, thereby greatly extending the mean time between failures. This kind of repeated checking will generate a huge computational burden, which means that everything needs to be done twice in an instant. This also causes cost and benefit issues.

In order to achieve safety,CPU（CPU) Is a good way to perform fast repetitive calculations, but it is impossible for you to render rich display effects by using only the CPU. This will cause the display to be monotonous and non-interactive, and it is also installed in a fixed location. Although this is affordable, it does not provide the best content for the driver in front of him and in the vehicle’s control position.Traditionally, another option is to use graphicsprocessor（GPU), but in this way, each frame requires two rendering channels, which means that although the rich three-dimensional (3D) rendering effects will amaze the driver, a large amount of chip area and cost are needed to facilitate its realization.

So, what we need is the amazing user experience brought by traditional GPU, as well as the fast computing power and low cost brought by CPU.

Achieve scalability in terms of performance and cost

If there is no new hybrid computing solution that can do everything well, what can be done is to optimize the existing solution and customize it to meet the needs of the car, especially considering the deployment of high-end security functions to The car manufacturer’s lowest value model is medium time. With this in mind, a successful automotive GPU for rendering and display needs to be scalable, efficient and functionally safe.

When discussing scalability, people almost always hope that there is no limit and unlimited computing potential. For certain situations, such as data centers and home computing, this can be great. However, few people will talk about the architecture can also be reduced, in fact, in this case, can provide the same high level of computing power and functionality, but the cost and chip area are only a small part of the original.

The GPU for multiple 4K displays used in modern luxury cars is the same in architecture as the GPU used for simple high-definition displays used in small city cars. This will not only reduce chip costs, but also reduce software costs, and design engineering time. Reduced because the underlying structure and principles remain the same.

Open tool acquisition and rapid prototyping

As general-purpose GPU solutions cover the entire group of automakers, developing software and systems that can run efficiently is the next area where we hope to cut costs without compromising quality. To achieve this goal, it is necessary to have easy-to-access and use tools to optimize the development experience and to prototype functions and systems as quickly as possible. Imagination’s recently released working group-protected OpenCL extension kit is such a system that can support automakers to build working models in a closed environment and iterate quickly before large-scale deployment.

GPU is not the end of the road, but it is a good start

The use of digital methods can save costs without affecting all important NCAP ratings. If a suitable GPU is used, the cost savings will be even greater. In addition to using the GPU as a preferred method to achieve automotive digital safety, modern ADAS and autonomous driving systems require specific neural network accelerators (NNA) to perform their own calculations.

By adopting a comprehensive approach to its entire value chain, and at the same time find a solution that covers everything from the smallest single core to the multi-core solution for fully autonomous vehicles (such as Imagination supports single-core and multi-core BXS GPU and Series4 NNA), automakers can save time and cost, and can win customers with a high-quality experience on models at any price.