RAC Cars News


Nvidia supercomputer to power self-driving cars

By raccars Published

Graphics Cars

Graphics card company set to throw its chips into the autonomous car market.

From GPUs to cars; one company’s supercomputer is set to be a driving force in the autonomous vehicle market. The tiny Nvidia supercomputer is to be used to power self-driving cars and has the capability to carry out a mammoth 20 trillion operations every second.

The CEO of Nvidia, Jen-Hsun Huang, announced at the first ever GPU Technology Conference Europe that his firm is developing a compact, simplified supercomputer which can be used to power autonomous cars. The supercomputer has been named Xavier and is an SoC, or system-on-chip, design which features both GPU and CPU on one chip.

Who says bigger is always better?

The company has managed to shrink the silicon in size in order to maximise efficiency whilst minimising space. Yet, Xavier still features a total of eight Nvidia CPUs, or Central Processing Units, along with 512 CUDA Cores. CUDA is the company’s ‘parallel computing architecture’, which is designed to boost computing performance by harnessing the power of GPUs or Graphics Processing Units.

These Cuda Cores ensure the ability of the computer to perform the type of tasks needed to work with the type of complex systems found in self-driving cars. The chip also includes two 8K resolution video processors in order to work through the video feeds provided by autonomous cars’ camera systems.

The Xavier supercomputer has the capacity to perform 20 trillion operations each second and yet its power consumption is just 20 watts. Nvidia’s previous creation, the Drive PX2 car chip, which is used in the Volvo XC90 SUV can also complete 20 trillion operations but is double the size and consumes 250 watts.

The latest Volvo XC90 SUV features the company’s Pilot Assist, a driving system with semi-autonomous features, aimed at boosting comfort levels and overall ease of travel. The earlier version of the Pilot Assist, however, was aimed at improving the traffic jam experience and worked only up to 50km/h. In contrast, the new version has the ability to steer, brake and accelerate up to speeds of 130km/h without the need for a lead vehicle.

There are limits to its capabilities, however, as it is not intended to completely replace the driver. The Pilot Assist system stops if the car detects that the driver’s hands are not on the wheel for a specified time because the car’s cameras cannot recognise every lane marking.

Nvidia's growing history in the automotive sector

Nvidia is best-known in the GPU market but the firm has been experimenting with chips for autonomous and other cars for a while. The head unit in the Tesla Model S is powered by the company’s chips and Nvidia has recently announced that it has joined forces with TomTom, the navigation company, to offer localised, real-time mapping data to be used in cars.

The Xavier supercomputer was teased at the GPU Technology Conference Europe event but it is likely to be at least 12 months before samples begin shipping to research institutions and vehicle manufacturers. The groundwork has begun but it will be a while before self-driving cars powered by the new Nvidia supercomputer will be found on the roads.

More about Nvidia

Nvidia is an American tech company with its base in Santa Clara in California. It is renowned for designing GPUs to be used within the gaming world, in addition to components for the automotive industry and the mobile computing sector.

Since 2014, the vehicle sector has been one of four markets on which the firm has pledged to focus. The other three are data centres, professional visualisation and gaming. The company creates Tegra mobile processors, for example, to be used in tablets and smartphones, as well as in vehicle entertainment and satellite navigation systems.

Image source

Looking to Buy?
Search for cars