Complexity is driving new technology paths as engineers ponder how to design a global network of AI, IoT and 5G-driven devices from chip to cloud. Dealing with that complexity requires fundamental changes in technology approaches and how people across the sector collaborated.
“There was a simple truth guiding us: hardware designers and software developers have to become even closer allies,” Simon Segars, Arm CEO, says. “As is often said: Silicon is just sand without software, and software is a movie with nothing to play on without hardware.”
It is fair to say that the recent acquisition of ARM by NVIDIA has drawn mixed reaction from media and analysts but that is a sentiment that Segars does not agree with. “For me, the negative comments by some commentators (dissipating as we explain the deal further) were anchored in misconceptions,” he says. “I truly believe the combination of Arm and NVIDIA underscores perfectly the concept: A way to accelerate connections and build an even broader, more pervasive AI-enhanced technology platform for the world’s hardware engineers and software designers.”
The three Ps
The reasoning behind that feeling formed the backbone of Segars keynote speech at the ARM DevSummit, held virtually this year. He focused on what he called the three fundamental elements that govern why Arm has become the most widely deployed advanced architecture ever: power, platform, and pervasiveness.
For power his key message was more performance per watt, from the world’s fastest supercomputer to devices that harvest energy from the air. “Power efficiency has long been part of Arm’s DNA, but what began as a drive to maximize battery life in early smartphones is now a moral imperative to reduce the carbon footprint inside every device Arm’s an integral part of,” he explains. “Take the world’s fastest supercomputer – Fugaku. Despite it having a reported 7.3 million Arm processors drawing over 28,000 kilowatts, Fugaku does not only top the supercomputer performance list, it also ranks within the top five most power-efficient too. It is already being used in the fight to find a COVID-19 vaccine.
“At the opposite end of the spectrum, we have been working with DARPA on the M0N0 project. M0N0 is an Arm Cortex-M33 based system-on-chip (SoC) drawing a mere 10nW at standby and 10µW while active. For context: at 10nW, a single 30mAHr button cell battery could power the M0N0 SoC for 340 years.
“M0N0 is smart too; paired with DARPA’s N-ZERO acoustic sensors, a system could run machine learning (ML). When combined with other N-ZERO sensors, M0N0 might be sensing and processing infrared data, radio frequency (RF) or even chemical signals. Outside of N-ZERO, M0N0’s use cases could include adding voice activation into devices such as car key-fobs or monitoring for anomalies in inaccessible gas pipelines.”
A robust and accessible platform
When it comes to platform, Segars continues to toe the traditional ARM line that partners need a robust accessible and standardised platform of technology to enable them to dream big and deliver in a highly efficient way. “Our Partners range from the most recognizable names in the industry to pre-launch start-ups working out of university labs and garages, he explains. “Atmosic’s Ali Bukhari wrote an Arm Blueprint article in June discussing how the freedom to experiment with a range of our IP afforded by Arm’s Flexible Access model had been crucial in bringing the company’s energy-harvesting M3 chip product to market quickly. The Atmosic team was able to audition various CPUs and run simulations without losing development time, paying an up-front NRE (non-recurring engineering costs) or getting locked-in to a particular CPU.”
The way Arm IP can be licensed continues to be expanded so the next step is thinking about how to ease the path to deployment. To this end in his keynote he announced the launch of Arm SystemReady. “SystemReady is a new program bringing a level of consistency across a broad range of Arm-based devices in the cloud, in the network and in high-performance IoT (HPIoT) endpoints,” he continues. “It will tackle the common software stack as well as system architecture. NVIDIA’s AI compute platform is a perfect fit with Arm’s broad processor base and neutral ecosystem-driven business model. Together we can provide a new kind of innovation acceleration that couples Arm’s platform reach and ties in with NVIDIA’s leading AI compute capability.
A pervasive technology
It is hard to argue that Arm technology is pervasive throughout the world, from sensor to smartphone, self-driving car to supercomputer. 180 billion chips in, Arm technology deployment continues to grow year on year. What will the ultimate deployment look like? “No-one can predict the final quantum, but I think Arm-based ‘things’ will eventually be counted in their trillions,” Segars concludes. “By delivering an ever-broader platform of technology with which developers can bring their visions to life, we stand to ensure that the future for all of us is unlimited.”[/et_pb_blurb][/et_pb_column][/et_pb_row][/et_pb_section]