Intel's Go supercomputer for cars points to a PC-like horsepower race

Intel's new Go computer for self-driving cars will have up to 28 Xeon cores

Intel Go automotive development kit

There's a race to put more computing power in self-driving cars, and it's shaping up to be eerily similar to an earlier battle between Intel and AMD to crank up PC horsepower.

Intel at CES announced the powerful Go computer with up to 28 Xeon chips so self-driving cars can cruise the streets safely. Beyond Xeon, Go will also be available with either next-generation Atom chips or 5G connectivity.

The first 40 self-driving BMW cars based on the Go will hit the streets in tests this year.

Autonomous cars need a lot of computational power under the hood to avoid accidents and make smart driving decisions. That horsepower comes from computers like the Go, which is configured to be faster than gaming PCs and many servers.

Go will compete with Nvidia's Drive PX 2, a water-cooled supercomputer for autonomous cars announced by Nvidia last year. The Drive PX 2 has 12 CPU cores and two Pascal-based GPU cores but no 5G connectivity.

The car chip race between Intel and Nvidia resembles the competition between AMD and Intel in the late 1990s and early 2000s PC heyday to crank up the clock frequency of processors. The chip makers ultimately gave up the battle to focus on power efficiency to increase battery life in laptops. 

Intel believes it has an advantage over its rivals with 5G connectivity, allowing cars to communicate with servers in the cloud to analyze images. The connectivity is an important tool for cars to identify objects on streets.

Both the Go and the Drive PX 2 have the same target: to train computers to be smarter. The computers help cars detect pedestrians, recognize lanes, and stop at signals. Computers make decisions based on algorithms and data collected from cameras and sensors like lidar and radar.

Intel and Nvidia are also in a race to put self-driving cars on streets. Intel is working with Mobileye and BMW on self-driving cars, and Nvidia's hardware is being used in Tesla vehicles with self-driving features.

The biggest benefit to self-driving cars is in safety, but there are other functional and economic benefits, said Kathy Winter, vice president and general manager of the automated driving division at Intel.

Self-driving vehicles could help people make better use of their time. Self-driving cars could bring economic benefits to logistics related to transportation and delivery of products, Winter said.

Uber is already testing self-driving taxis in the U.S., and other technology companies are dabbling in self-driving cars. Companies are working together to establish standards for autonomous vehicles to communicate on traffic and weather conditions.

The Go computers are part of development kits that come with software tools. The kits join a stable of Intel computers that include NUC mini desktops, Compute Sticks, and development boards for smart devices, robots, and drones.

Intel's Go hardware with Xeon is supplemented by two FPGAs (field programmable gate arrays), which can be reprogrammed to do multiple tasks. The FPGAs are to Intel's Go what the GPUs are to Nvidia's Drive PX 2 -- fast chips that will help cars make key decisions instantly. The version of Go with next-generation Atom chips has only one FPGA.

The Go computer with 5G will be demonstrated at the upcoming Mobile World Congress in Barcelona. Intel announced its first 5G modem at CES this week.

Copyright © 2017 IDG Communications, Inc.

7 inconvenient truths about the hybrid work trend
Shop Tech Products at Amazon