In May 2020, Arm announced our latest CPU, GPU, and NPU IP for the mobile market. However, what set this launch apart was the Cortex-X Custom (CXC) program. Recently, our partners have announced new SoC chipsets that are redefining mobile devices. At the core of those chipsets is the Cortex-X1, which is the first product of the CXC program. After nearly a year since its launch, I am delighted to give you a brief overview of the program, and the genesis and vision for the Cortex-X class of CPUs.
Arm has the unique distinction of being the world’s leading technology provider of processor IP. With over 180 billion devices shipped by our partners, Arm’s technology powers everything from barely visible sensors to huge datacenters. Our partners and customers expect us to lead the way by constantly innovating and staying ahead of trends and technologies. It is this culture and drive for innovation that led us to create the CXC program.
Arm’s ever-expanding ecosystem of products means that our partners are building devices that must excel under the most demanding applications and workloads, while simultaneously improving battery life. Our planning and discussions with partners revealed that we need a new CPU option built on a scale and scope never seen before. Arm's technical talent responded with not only a novel microarchitecture, but also a culture shift in how we design CPUs. Seamless user experiences and superior performance needs warrant CPUs built and tuned to match use cases, and not the other way round – that is use cases being used around CPU designs. This approach is not only reflected in our CXC products, but also in our Cortex-A CPUs, which are designed to benefit specific use cases. Arm’s Total Compute strategy acts as the blueprint for this shift, from a product evolution focus to use case driven solutions.
A key element of any CPU development is the interaction with our partners. Therefore, we wanted to ensure that our partners had a deep level of engagement in the development process. As part of the CXC program, we collaborate closely and early with partners to hone the key aspects of the IP. This feedback process directly translates to use case customization and reduced time-to-market, which are key competitive advantages.
Cortex-X1 is our most powerful Cortex CPU to date. It is designed to bring ultimate performance to a new class of applications. This performance has a tangible impact on the premium smartphone user experience.
In recent years, smartphones have become the main hub for all consumer technology experiences. Among a host of various applications, it is our primary device of image capture, web surfing and – more recently – binge-watching of TV shows and videos online. The custom-built Cortex-X1 makes the interfaces during these experiences extremely responsive. Applications load in the blink of an eye and web surfing is a completely smooth and seamless. The raw computational performance of Cortex-X1 also serves as a catalyst to the adoption of new camera features. This leads to superior photos and videos, as smartphones continue to add innovative camera features like multiple lenses, image stabilization and larger sensors.
While Cortex-X1 is focused on peak performance, the Arm Cortex-A78 CPU (also announced as part of the Mobile IP launch) serves as the engine for sustained performance. A tri-CPU cluster Cortex-X1, Cortex-A78, and Cortex-A55 built on the foundation of DynamIQ technology delivers a perfect mix of power, performance, and efficiency. Arm's DynamIQ technology combines the big and LITTLE CPUs into a single, fully integrated cluster. In a premium smartphone configuration*, the Cortex-X1 is capable of delivering up to 30 percent performance uplift over previous generation 2019 smartphones.
For the past decade, Arm has led the industry on the highlighting the benefits of the big.LITTLE CPU topology. We strongly believe that such configurations deliver the best and most balanced solutions, so it is pleasing to see the increasing adoption of this philosophy within the tech industry. We are confident that the big.LITTLE clusters bring meaningful improvements to user experiences. In many ways, big.LITTLE is becoming the de-facto standard for segments beyond smartphones.
The reaction from technical press and analysts to the Cortex-X1 and CXC program is very positive. Many are impressed by the performance trajectory of Cortex-X1. Android Authority stated that “Cortex X1 answers calls for an Arm CPU with extreme performance”, while Gizbot wrote that “it is an all-new architecture that offers better performance in every aspect”. Tech media and analysts were also eager to see how Cortex-X1 based silicon would translate to real-world improvements on flagship devices.
Arm IP in the Exynos 2100 chipset
We are now seeing Cortex-X1 in silicon. This includes Samsung LSI's Exynos 2100 SoC and Qualcomm's Snapdragon 888 5G mobile platform. Both partners are adopting the five-nanometer process node in a tri-cluster featuring Cortex-X1, Cortex-A78 and Cortex-A55. The chipsets power the new Samsung Galaxy S21 smartphone launched on 14 January.
The tri-cluster configuration covers all the key performance bases on mobile. Cortex-X1 delivers peak performance, Cortex-A78 delivers sustained performance over a longer time period, and Cortex-A55 supports the performance of key background tasks. Focusing on the Exynos 2100, the tri-cluster CPU configuration delivers enhancements in multiprocessor performance compared to the Exynos predecessor.
Tech reviewers and influencers have been impressed with the CPU performance of the Exynos 2100 in the Samsung Galaxy S21. Marques Brownlee, an influential Tech reviewer, commented “initial testing is already pretty promising for the Exynos 2100”, and reflecting on the battery life he said: “I never once killed it in a single day.” When announcing the S21, Samsung Mobile expressed their confidence in our CPU roadmap – “We have adopted Arm architecture to significantly increase overall chipset speed”. We have loved this initial coverage and in the coming months expect more vendors to adopt chipsets built on Cortex-X1.
The Cortex-X performance trajectory
The CXC program has been a sea change to Arm’s CPU design philosophy. There is now an increased focus on specialized processing, as we build our CPU designs based around requirements and use cases from our partners and the ecosystem. Based on these discussions back in May 2020, we stated our intention to push the boundaries of performance and we continue to be focused on that mission. A high-performance CPU cluster translating into improved performance for more complex and enriching user experiences. We are excited about what the future holds and cannot wait to share more of our latest and revolutionary products. Stay tuned for more announcements in our 2021 IP launch in May.
Learn more about the Cortex-X Custom program
*Comparing Arm single core peak performance at 3.0GHz. Cortex-X1: 1MB priv-L2, 8MB L3 cache vs Cortex-A77: 512KB priv-L2 , 4MB L3 cache, including architectural and process improvements (compared to 2019 devices). Measured estimates on SPECint*_base2006 (SPECspeed* Integer component of SPEC CPU* 2006) Arm single-core performance estimated for mobile platform. Results are measured estimates using specific computer systems, software, components, operations, and functions and changes to any of these factors will cause the results to vary.