Buying an autonomous car may be in your future, but make sure it has a capable processor installed so the vehicle doesn’t get in trouble with traffic police.
Autonomous cars could make timely and more accurate decisions with Nvidia’s new Tegra chip code-named Parker. The chip has the computing power to allow autonomous cars to recognize a wide range of signs, objects, signals, and lanes.
Parker can also deliver 4K video to in-car entertainment systems. Details of the chip were presented for the first time at the Hot Chips conference this week in Cupertino, California.
The Parker chip is to vehicles what Intel’s Core chip is to PCs. It delivers a lot of computing horsepower to autonomous vehicles, considered a new form of “mobile” computing by chipmakers.
The announcement also raises questions about whether Nvidia will put Tegra chips in mobile devices and board computers like the Jetson. In a blog entry, Nvidia called Parker a “mobile processor that will power the next generation of autonomous vehicles.”
Parker, like its predecessor Tegra X1, is designed for low-power and battery operation, according to an Nvidia spokesman. The company didn’t immediately share details on other uses for the chip.
Nvidia has gotten out of the smartphone market, and the future of Tegra in tablets is in question. Last year’s Tegra X1 was used in Google’s Pixel C tablet, and Nvidia’s outdated Shield Tablet uses the Tegra K1, the predecessor to X1.
The Parker chip is already used in Nvidia’s Drive PX 2, a computer that sits in car trunks. Drive PX 2 will be used by Volvo in road tests of its XC90 sports utility vehicle next year. Car companies and researchers are also developing autonomous cars with Drive PX 2, Nvidia said.
Drive PX 2 does the heavy calculations so autonomous vehicles can do their job, but the computer also needs to consult cloud computing services for training and object recognition.
The computer supports deep-learning models from companies like Google, Baidu, Microsoft, and Facebook. The Parker chip has been optimized to process data related to a variety of deep-learning models.
Parker has four 64-bit Cortex-A57 cores based on the ARM architecture, and two homegrown Denver 2.0 CPU cores. It has security layers to prevent the chip from being hacked, according to slides presented at Hot Chips.
Nvidia says Parker outperforms other chips, a claim other ARM-based chipmakers like Qualcomm and Apple may contest. However, its CPUs are quickly going out of fashion. The Cortex-A57 cores are being replaced by ARM’s faster and more power-efficient Cortex-A72 cores in chips.
Parker’s GPU blows away the competition. It has a 256-core GPU based on the latest Pascal graphics architecture, and the performance is important for visual computing. Cameras capture images and then work with the GPU and algorithms to recognize objects, signs, and signals. The chip supports up to 12 cameras.
The GPU supports hardware-based encode and decode for 4K video based on the H.265 and VP9 codecs. Nvidia claims it can support playback on three 4K displays at 60 frames per second.
Gamers would rejoice if Nvidia decides to release a new Shield tablet with the Parker chip.
Parker supports LPDDR4 video and is made using a 16-nanometer manufacturing process.