LLMpediaThe first transparent, open encyclopedia generated by LLMs

ARM Cortex-A72

Generated by DeepSeek V3.2
Note: This article was automatically generated by a large language model (LLM) from purely parametric knowledge (no retrieval). It may contain inaccuracies or hallucinations. This encyclopedia is part of a research project currently under review.
Article Genealogy
Parent: Graviton (processor) Hop 4
Expansion Funnel Raw 40 → Dedup 0 → NER 0 → Enqueued 0
1. Extracted40
2. After dedup0 (None)
3. After NER0 ()
4. Enqueued0 ()
ARM Cortex-A72
NameARM Cortex-A72
DesignerARM Holdings
Bits64-bit
ArchARMv8-A
MicroarchARM Cortex-A72
Cores1–4 per cluster
L1cache48–64 KB I$, 32–64 KB D$ per core
L2cache512 KB – 4 MB shared
L3cacheOptional
PredecessorARM Cortex-A57
SuccessorARM Cortex-A73

ARM Cortex-A72. The ARM Cortex-A72 is a high-performance 64-bit ARM architecture processor core designed for mobile computing and embedded applications. It was announced in early 2015 as the successor to the ARM Cortex-A57, offering significant improvements in performance and power efficiency. The core is widely implemented in systems on a chip (SoCs) from major semiconductor companies.

Overview

The Cortex-A72 was unveiled by ARM Holdings as part of its premium-tier big.LITTLE processing strategy, intended to be paired with smaller cores like the ARM Cortex-A53. It was a central component of ARM's 2015–2017 roadmap for flagship mobile devices, competing directly with offerings from Qualcomm and Apple. The design focused on delivering a generational leap over its predecessor, the ARM Cortex-A57, particularly in terms of sustained performance and thermal management. Its introduction coincided with the broader industry transition to advanced FinFET manufacturing processes, which enabled its efficiency gains.

Architecture

The Cortex-A72 implements the ARMv8-A instruction set architecture, supporting both the AArch64 and AArch32 execution states. Its microarchitecture features a 15-stage integer pipeline, an improved branch predictor, and an enhanced prefetch unit compared to the ARM Cortex-A57. The core includes an advanced floating-point unit and NEON media processing engine for accelerated multimedia workloads. Key memory system improvements consist of a more efficient cache coherence protocol and support for larger, shared L2 cache configurations. The design also incorporated specific optimizations for better performance in virtualized environments managed by hypervisors like KVM.

Performance

ARM claimed the Cortex-A72 delivered up to a 3.5x performance improvement over the ARM Cortex-A15 core from 2012 when manufactured on a comparable process node. In typical mobile SoC configurations, it offered approximately 1.8x the performance-per-watt of the ARM Cortex-A57. These gains were realized through both microarchitectural refinements and the shift to newer semiconductor fabrication technologies from partners like TSMC and Samsung. Independent benchmarks, such as those from AnandTech, showed significant gains in integer and floating-point workloads, making it highly competitive in devices like the Raspberry Pi 4 Compute Module. Its performance profile made it suitable for demanding applications in premium smartphones, tablets, and single-board computers.

Implementations

The Cortex-A72 was licensed and implemented in numerous commercial SoCs. Qualcomm utilized it in a custom quad-core configuration as part of its Snapdragon 650, 652, and 653 platforms. MediaTek integrated it into the Helio X20 and X25 deca-core processors using its Tri-Cluster architecture. Broadcom featured the core in the BCM2711 SoC for the Raspberry Pi 4. Other notable implementations include the HiSilicon Kirin 950 and 955 from Huawei, the NXP i.MX 8 series for automotive and industrial markets, and the Marvell Armada 3700 for networking equipment. These chips were fabricated on processes ranging from TSMC's 16nm to Samsung's 14nm FinFET.

Comparison with other ARM cores

Compared to its direct predecessor, the ARM Cortex-A57, the Cortex-A72 provided superior energy efficiency and higher peak clock speeds within similar thermal envelopes. Against its successor, the ARM Cortex-A73, it typically offered higher maximum performance but with greater power consumption, as the A73 focused on area and efficiency for thinner devices. In the broader market, its performance was often compared to custom cores like the Apple A9's Twister core and Qualcomm's partially custom Kryo cores in the Snapdragon 820. While it lacked the specialized machine learning accelerators found in later cores like the ARM Cortex-A75, its robust general-purpose performance ensured a long lifecycle in embedded and cost-sensitive applications. Category:ARM microarchitectures Category:2015 introductions