Intel’s new Panther Lake processor on the 18A process node offers improved performance and power efficiency for PCs, laptops and edge devices. The company will soon release a developer’s reference board for building edge applications for industrial and business uses.
With Panther Lake, Intel is decidedly taking a bigger step forward in the humanoid robotics world where robots with arms and legs move and think and where motion and mechanical sensors play a key role.
Even though the range of edge applications supported by Panther Lake will be broad, Intel’s new chip will not be combined with built-in sensors, but will certainly support third party sensors and even enhance their capabilities with AI —opening up greater opportunities for sensors, according to Mike Masci, vice president of product management for edge at Intel. The company will continue to support a wide range of sensors from other companies including major suppliers such as Texas Instruments.
In an interview with Fierce, he explained that sensors are complex and often operate in analog, while Intel is focused on a range of functions to serve an entire application at the edge, including AI through various components including GPU cores, as well as CPU and NPU cores and more.
“Sensors in analog are complicated and require a logic analog skills set and that doesn’t necessary relate to a lot of digital logic. But we absolutely work with sensor vendors and MEMS vendors,” he said.
A key market for Intel’s Panther Lake at the edge will be industrial applications, but also satellites and other aerospace products. Some potential customers have even expressed interest in testing how radiation tolerant Panther Lake will be for satellite uses, he said in an interview with several reporters including Fierce.
“We’ve had good interest from military and industrial customers,” Masci said. “Some do testing for radiation tolerance on existing form factors” to see if they might operate more efficiently with Panther Lake.
The number of edge-focused customers interested in testing Panther Lake has exceeded expectations, he added. “Our early access program is…oversubscribed,” and Intel has had to turn people away. Healthcare applications and humanoid robotic use cases are in the mix. Ruggedized design use cases and “everything in between” have been presented by potential customers. Overall, Intel works with 4,000 different companies in edge, including the largest companies in robotics.
Panther Lake for edge will be configurable, down as low as 15 watts to operate, with most in the 45 watt range and also up to 65 watts. But he added, “We’re not getting to sensors at 2 watts.”
Intel showed off applications for robots, restaurants and traffic management at an event in September in Phoenix.
One application allows doctors to type in requests of medical images to be taken to a specific image view to examine a patient’s condition. Eventually that text will be converted to voice as well.
Also, a big opportunity comes for robotics with Panther Lake.
“The robot is the ultimate manifestation of IoT edge,” Masci said. “There’s magic and the peripheral attach (for arms and the head) is hard to do reliably.” He asserted Intel has been the long-time leader in industrial robotics with use of x86 chips for command and control, but added, “we’re not the Roombas and won’t likely be in the future.”
With Panther Lake, AI will take off in robotics, he added. “As humanoids start to emerge, developers will have control and AI processing like they never had before.”
In restaurants, systems are emerging to monitor take-out orders using visual sensors in various forms to detect whether the right products are bagged up for customers. With AI introduced through Panther Lake, a visual sensor will be delivering data that can be interpreted with AI for better accuracy. For instance, if a product label changes color, the AI will read the label for more information, such as the product name. Without AI, a reading would not be as accurate.
Another application is traffic management, where Panther Lake enclosed in a rugged box will work at an intersection to detect traffic flow problems and recommend traffic movement adjustments. The system will incorporate weather conditions and other data.
A new partnership between Nvidia and Intel could enhance AI use in robots especially. Intel’s OpenVino software is an open system that could complement Nvidia’s CUDA software as well as systems from other vendors.
Intel is rolling out an AI suite for developers that will optimize software libraries and make it easy to move from AMD, Qualcomm, Nvidia and others. “We want to make it easy for any developer from any platform,” Masci said.
Intel said a typical industrialized PC or other ruggedized box will have an Intel processor inside and through a variety of I/O protocols, (some wired and some wireless) connect to various sensors. The sensors will be typically low-power and low cost, designed by other vendors, but usually don’t have the processing power to interpret the data. That’s left to the Intel Edge approach.
Intel has previewed on GitHub an Intel Robotics AI Suite of software and related tools to simplify AI deployment. It joins other edge AI suites from Intel that address applications for retail, manufacturing, metro/smart city and media and entertainment.
The reference board Masci showed reporters is separate from the Robotics AI Suite and will be offered separately from the suite with availability and pricing available sometime after the first of the year “around CES timeframe,” according to an Intel spokeswoman. The board will be optimized to run the software, frameworks and benchmarks that are a part of the suite.
The Robotics AI Suite, according to a blog, will allow a single Intel Core Ultra processor (also known as Panther Lake, a codename) to take the place of what has traditionally been two separate processors, sometimes from different companies with different software stacks. Of those two processors, one has taken on real -time controls of a robot, such as the movement of joints to ensure precise timing. The other is the vision and AI perception side that allows a robot to see and understand its environment and how to move through it.
“You’re seeing the latest processors from Intel be able to handle both the real-time control capability and enough AI processing to do the vision and AI perception half, in one processor, “ said Matthew Formica, head of edge product marketing and AI PC/Edge AI software development relations.
That combination makes it simpler for a robot designer to build using one part to lower costs and complexity, also cutting development time, he added.
Intel named three customers who have worked with Intel on robotics applications, including RealSense, Techman Robot and Fourier. Fourier Vice President Willys Zhai said in a statement that the company is anticipating the Robotics AI Suite to “deliver efficient, scalable solutions for real-world deployment …to advance the frontier of embodied AI.”
RELATED: Intel on a roll: client, edge and data center chips on 18A