COMPUTEX 2021

Accelerating Ubiquitous Intelligence: Arm discusses requirements of complex AI and ML workloads

Rene Haas, President of Arm’s IP Products Group (IPG), addressed the topic of “Accelerating Ubiquitous Intelligence” in his keynote at the 2021 COMPUTEX Forum—the annual Taipei computing conference. Haas shared about the unprecedented acceleration of compute trends driven by the growing demands for ubiquitous intelligence across every device on the compute spectrum. He also weighed in on why the complex Artificial Intelligence (AI) and Machine Learning (ML) workloads of today and tomorrow will require more than general-purpose computing.

Arm: The world’s most pervasive compute platform

Arm technology has been changing the way people live and the way businesses operate. Over the past five years, more than 100 billion Arm-based devices have been shipped and reached 70% of the world’s population. Haas expressed great confidence that a 100% of the world’s digitally shared data will soon be securely processed on Arm-based technology at some point of its life: either at the endpoint, in the data networks or the cloud.

Arm’s focus on innovation: Armv9

The high demand for Arm technology across the globe fuels Arm’s motivation for innovation to meet the ever-changing needs of the computing world. With constant upgrades, improvements, and innovation, the Arm architecture is never static. Haas pointed to the extensions and features, such as security advancements, that were added to the Armv8.

Arm’s more recent launch, the Armv9, represents the next major milestone for the Arm architecture for the decade ahead. The new Arm architecture will apply across all CPUs and sports a wide range of new features and extensions including increased performance and efficiency, ML, Digital Signal Processing, as well as improved security to secure the world’s data.

Arm Confidential Compute Architecture

The greatest challenge in the technological world today, Haas expressed, is the issue of how to secure the world’s data. Armv9’s Arm Confidential Compute Architecture is a direct response to this challenge. By performing computation in a hardware-based secure environment, Armv9 shields portions of code and data from access and modification while in use.

Arm Confidential Compute introduces the concept of Realms. Usable by all applications, Realms secures the world’s data and creates regions in a system to separate sensitive and non-sensitive data to protect commercially sensitive data and code from the rest of the system while it is in-use, at rest, and in transit.

Apart from adding new layers of data protection, Haas shared that Arm has taken the vital step in improving the security of the world software. In a collaboration with Google, Arm developed Memory Tagging Extensions, a technology that can uncover vulnerabilities before they can ever be exploited.

Accelerating ubiquitous intelligence: Arm’s SOC revolution

Haas claimed that Armv9 will play a critical role in securing our digital future and increasing our trust in intelligent devices which process our data. In light of the pandemic situation, we have seen the increased use of technology to record and manage personal health data. Arm believes that this is just one of the inevitable compute trends that accelerated in this unprecedented time. Healthcare, along with other industries like retail, technology, as well as industrial manufacturing will (and are) accelerate their use of ML and AI as a response to the pandemic.

For the compute industry, this represents an opportunity to accelerate ubiquitous intelligence across all devices. This entails having our devices run incredibly complex workloads with greater performance and energy efficiency.

In this regard, Haas shared that the one-size-fits-all chip designed for general-purpose compute with big single-threaded processors will no longer suffice. Modern System On a Chip (SOC) with dedicated compute blocks for different tasks and greater efficiency are required. For several years, Arm has been enabling an SOC revolution and designing customized chips for application-optimized and specialized computing.

New era of compute: Specialized Processing as the new normal

Haas shared that we are now entering a new era of compute where AI and Ml will cut across every computing environment and will be enabled by specialized processing. For 30 years, Arm has been enabling partners with flexibility to build specialized processing, effectively equipping them with all they need to enter the new era of compute where specialized processing becomes the new normal for computing. All of these will be built on the foundation of the Armv9 architecture.

Arm’s approach to providing a diverse suite of IP components such as CPUs, NPUs, and GPUs allows partners to develop purpose-built systems they need for ubiquitous intelligence and other emerging challenges. Haas cites examples of use cases of specialized processing in AI video, automotive systems, 8K streaming, VR, and AAA games.

Total Compute: A scalable and flexible approach

In this new era of compute, digital experiences are richer and more dynamic than ever before. With growing challenges brought forth by a combination of 5G and AI, a new approach that is scalable and flexible across entire systems is needed. Haas introduced Total Compute as Arm’s blueprint to address tough challenges around the need for higher performance compute and increasingly complex systems.

From software, tools, and physical IP, Total Compute brings together the powerful assets of the Arm platform to deliver flexible solutions across premium, performance, and efficiency market tiers. This enables Arm’s partners to develop the best products while unlocking new experiences for the entire ecosystem. Consumers, on the other hand, will enjoy stunning content, quicker speeds, and longer battery life on their TVs, smartphones, and laptops.

Arm’s solutions approach of Total Compute does not only apply to client offerings. Its design principles are applied across Arm’s entire IP portfolio. This includes Arm’s Neoverse platform for edge-to-cloud infrastructure solutions across four segments: Hyperscale and cloud computing, 5G, HPC, and infrastructure edge.

A new generation of Arm cores

Just a week ago, Arms launched Cortex-X2, Arm’s first v9 flagship CPU for ultimate performance via the Cortex custom X program. This is coupled with Arm’s Cortex-A710 and Cortex-A510. Taken together, this generation of the new breed of high performance Arm cores promises greater performance, speed,  boost to machine learning, and energy efficiency.

Haas claimed that the potential for performance in this new CPU system will open up many possibilities across different markets and devices. He singled out the potential in the laptop market. Cortex-X2 could deliver 40% more peak single-thread performance and provide all-day battery life.

Alongside Arm’s CPU line-up announcement Haas introduces Arm’s newest Mali GPU line-up with the new Mali-G710 flagship series, the G510 middle-range, and the new ultra-area efficient Mali-G310. These bring extremely large generational performance boosts as well as power efficiency gains over their predecessor.

Building momentum with Arm technology

There is a lot of momentum behind Arm technology for infrastructure. For one, NVIDIA GraceCPU is NVIDIA’s first Arm-based data center CPU delivering 10 times the performance of today’s fastest server while running complex computing workloads. NVIDIA has also collaborated with Marvell to Octeon, the Arm-based family of data processing units from Marvell, to bring GPU acceleration for enterprise computing, 5G infrastructure, and edge applications. Haas also cited the example of a new NVIDIA Arm HPC developer kit.

Bringing Compute Closer to Data

Only 1% to 12% of company data is actually analyzed, according to a new report by Strategy Analytics. The ability to perform advanced localized processing closer to where data is collected results in faster and more accurate responses, allowing companies to maximize data insights. It would also lower dependency on the Cloud, and better manage IoT device data.

Arm is committed to enable on-device ML so as to bring compute closer to data. With a belief that no device should be left behind, even smaller and more cost-sensitive devices can be smarter and more capable.

Arm: accelerating of ubiquitous intelligence

In the age of AI, Haas shared that the combination of Arm’s compute platform and ecosystem and NVIDIA’s expertise in AI will create the world’s leading compute company. Haas concluded with excitement for Arm’s contributions in accelerating ubiquitous intelligence.

Comments (0)

Leave a Reply

Your email address will not be published. Required fields are marked *