Tesla supplier Nvidia announces new in-car AI supercomputer
A key supplier to Tesla Motors, Nvidia, is known predominantly for its graphics card business. However, over the past few years, Nvidia has been making a big push into the automotive industry and self-driving vehicle technology. To that end, we covered a fascinating interview earlier this year between Nvidia CEO Jen-Hsun Huang and Tesla CEO Elon Musk at Nvidia’s GPU Technology Conference in which Musk said: “What Nvidia is doing with Tegra is really interesting and really important for self-driving in the future.” The two CEOs have forged a critical alliance as Tesla is currently using Nvidia’s Tegra chips, and, Huang (during the interview) remarked that Nvidia is being pushed to the limit by newer software updates to the Model S.
Above: rendering of Nvidia's new Drive PX 2 capabilities
Although Tesla Motors is not at the Consumer Electronics Show (CES) this week, Nvidia made big news yesterday at CES announcing a new artificial intelligence (AI) supercomputer designed for self-driving cars.
Above: Nvidia CEO Jen-Hsun Huang and Tesla Motors CEO Elon Musk at Nvidia’s GPU Technology Conference earlier this year
As a major supplier to Tesla Motors, this news from Nvidia may prove useful downstream as Tesla expands its self-driving “autopilot” capabilities and preps for its forthcoming, lower-priced Tesla Model 3. And, Huang’s fervent support for Tesla Motors extends beyond simply being a key supplier. It turns out that Huang was an early owner of the Tesla Model X (see his proud tweet below) along with his “stable” of Model S vehicles in his own (yes we're envious) garage.
So what exactly did Huang announce at CES? According to Gizmodo*, "Nvidia has announced its first in-car artificial intelligence supercomputer at CES… called Drive PX 2, [and] is said to be ‘the size of a lunchbox and with the computing capability of 150 MacBook Pros’ in a press release. The second half of that statement is backed by two Tegra processors and two discrete GPUs. It’s claimed it works quickly enough to gobble up data from 12 video cameras, along with lidar, radar and ultrasonic sensors, then processes the streams of information to make sense of the outside world. The result is a computer that can work through 24 trillion deep learning operations per second… in reality that means it can process up to 2,800 images per second using a neural network-based algorithm. That should be enough for a car to orient itself accurately in the world, plan routes and work out what to do in the face of everyday road hazards, whether it’s bad drivers, erratic cyclists, or simply debris.”
Above: Nvidia's new Drive PX 2
To watch highlights from Huang’s announcement at CES, check out the video below…
Source: The Verge
It's important to note that Nvidia works with several different automakers, not just Tesla. And, Tesla's "autopilot" capabilities result from in-house expertise in addition to working with external suppliers like Mobileye and Nvidia. Nevertheless, this recent announcement from Nvidia could prove helpful in the coming years as Tesla refines and optimizes its overall approach to self-driving technology. We'll be sure keep you posted on it.