ARM Cortex-A72

ARM Cortex-A72
Name	ARM Cortex-A72
Designer	ARM Holdings
Architecture	ARM architecture
Introduced	2015
Cores	1–4 (typical)
Clock	up to 2.5 GHz (implementation dependent)
Ipc	improved over ARM Cortex-A15 and ARM Cortex-A57
Predecessor	ARM Cortex-A57
Successor	ARM Cortex-A73

ARM Cortex-A72 The ARM Cortex-A72 is a 64-bit ARMv8-A processor core developed by ARM Holdings and announced in 2015, targeting high-performance mobile and embedded markets. It was designed to deliver substantial instructions-per-cycle gains relative to prior cores and to compete with contemporaneous designs from Qualcomm, Intel, and Apple Inc.. The core saw widespread adoption in system-on-chip products by vendors such as MediaTek, Samsung Electronics, HiSilicon, Broadcom, and NVIDIA.

Overview

The Cortex-A72 was unveiled as part of ARM's roadmap that included the ARM Cortex-A57 and ARM Cortex-A53 clusters; it aimed to provide a balance between the microarchitectural advances found in server-oriented cores like those from AppliedMicro and consumer-focused designs from Apple A9. ARM positioned the A72 to displace the A57 in flagship system-on-chip designs from companies like Qualcomm Snapdragon, Samsung Exynos, and Huawei Kirin. It emphasized improvements across fetch, decode, and execution stages to increase single-threaded performance for workloads such as web browsing, multimedia playback, and machine-learning inference on devices from HTC and Sony Corporation.

Architecture and Design

The microarchitecture introduced wider instruction windows, deeper out-of-order execution capabilities, and enhanced branch prediction inspired by research from groups associated with Cambridge University and industrial designs from ARM Research. Key structural changes included an enlarged reorder buffer, deeper pipeline optimizations, and improved load–store units influenced by techniques documented by Intel and IBM architects. The A72 implemented the ARMv8-A instruction set and supported features like 64-bit addressing, virtualization extensions used by VMware partners, and cryptographic extensions compatible with standards adopted by NIST. Cache hierarchy choices reflected lessons from designs by AMD and Sun Microsystems (now part of Oracle Corporation), with configurable L1 and L2 sizes and integrated coherency mechanisms for multi-core clusters employed by vendors such as NVIDIA Tegra and Broadcom SoC families.

Performance and Benchmarks

ARM published benchmarks comparing the Cortex-A72 to predecessors and competitors, highlighting improvements against the ARM Cortex-A57, ARM Cortex-A15, and selected Intel Atom parts. Independent evaluations by media outlets and organizations such as AnandTech, Tom's Hardware, TechRadar, and academic benchmarks from SPEC and MobileBench showed notable gains in integer and floating-point throughput, and in real-world application tests like WebKit-based browsers and multimedia codecs from MPEG LA encoders. Performance scaled with process nodes offered by foundries including TSMC, GlobalFoundries, and Samsung Foundry, reflecting differences seen in chips fabricated at 16 nm and 28 nm geometries developed by TSMC Research.

Implementations and SoCs

The Cortex-A72 was licensed and integrated into numerous SoCs: Qualcomm Snapdragon platforms for midrange segments, MediaTek Helio series designs, Samsung Exynos variants combining A72 and A53 cores in big.LITTLE configurations popularized alongside ARM big.LITTLE publications, HiSilicon Kirin chips for Huawei devices, and embedded solutions from Broadcom for networking hardware used by Cisco Systems partners. The A72 also appeared in networking and networking-accelerator appliances from companies like Marvell Technology Group and in single-board computers inspired by Raspberry Pi community projects. Several custom implementations paired the A72 with GPUs from ARM Mali and third-party graphics IP from ARM partner Imagination Technologies.

Power Efficiency and Thermal Management

ARM targeted the Cortex-A72 at improved energy efficiency per instruction, leveraging process node improvements from TSMC and Samsung Electronics as well as microarchitectural power gating strategies discussed in conferences hosted by IEEE and ACM. Implementations used dynamic voltage and frequency scaling techniques similar to those in products from Intel Corporation and thermal management stacks compatible with operating systems from Google and Microsoft Azure services in edge deployments. OEMs integrated thermal throttling policies informed by standards from JEDEC and chassis-level cooling solutions from vendors like Foxconn for smartphones and Dell for laptops.

Software Support and Compatibility

The Cortex-A72 supports operating systems and software stacks including Android (operating system), Linux kernel, Ubuntu (operating system), and virtualization platforms used by KVM and Xen Project developers. Toolchains such as GCC and LLVM provided optimizations targeting the A72, while middleware libraries from OpenSSL, OpenCL runtimes, and machine-learning frameworks like TensorFlow and Caffe were ported to exploit its instruction set and NEON SIMD extensions. Ecosystem support involved silicon partners, OEMs, and cloud services from Amazon Web Services that validated A72-based appliances for edge computing.

Reception and Impact

Industry reception highlighted the Cortex-A72's role in closing the single-threaded performance gap with competitors from Qualcomm Snapdragon and Intel Core mobile lines, influencing design decisions across companies like Samsung Electronics, MediaTek, and Huawei Technologies. Analysts at firms such as Gartner and IDC cited the A72's contribution to the migration of flagship capabilities into mainstream devices, affecting market offerings from LG Electronics and HTC Corporation. Academic and industry citations at venues including HotChips and ISSCC discussed its microarchitectural trade-offs, and its commercial success informed subsequent ARM cores like the ARM Cortex-A73 and ARM's strategy toward server and mobile convergence promoted in collaborations with Canonical and cloud providers.

Category:ARM processors