Now owned by Intel, Mobileye believes its software/hardware pairing and focus on cameras will be competitive in terms of price and efficiency.
As Nvidia steals the show for self-driving car tech at CES, Mobileye has argued that its own autonomous car platform will be more affordable, efficient and simple to implement.
The Israeli startup was acquired last year by Intel for more than $15 billion. Engineers from both firms apparently began working together immediately to create an advanced autonomous driving product that combines Intel's chip prowess with Mobileye's expertise in autonomy software.
Intel has now released a white paper (PDF) claiming the Mobileye EyeQ5 platform offers more than double the efficiency for deep learning operations than Nvidia's Xavier hardware. Designed for Level 4/5 self-driving cars, EyeQ5 is said to benefit from a "smarter approach" to AI in which different types of operations are handled in the most efficient manner.
Speaking at CES, Mobileye communications chief Dan Galves admitted that Nvidia is competitive in terms of raw hardware capabilities but automakers must still write their own software. The custom coding model is said to be less efficient than matching hardware specifically for a predetermined software configuration.
"If you only have hardware ... it's more difficult to create an efficient piece of hardware because you don't know what software is going to be on it," he said, as quoted by ZDNet. "We bring the hardware and software together."
Mobileye also takes a different approach to sensor hierarchy, relying primarily on camera input to reduce processing overhead.
Galves' comments appear to ignore Nvidia's extensive software development kits that provide flexibility for automakers to implement different sensor configurations, object detection algorithms and other critical elements.
Notably, Tesla and Mobileye parted ways long before the Intel acquisition due to a disagreement over how the latter company's semi-autonomous system would be implemented in Autopilot. Tesla consequently gravitated toward Nvidia and is now rumored to be designing its own AI-focused processing chips in-house.