ARM64
Generated by GPT-5-miniExpansion Funnel Raw 87 → Dedup 3 → NER 2 → Enqueued 2
| ARM64 | |
|---|---|
| Name | ARM64 |
| Family | ARM architecture |
| Designer | Arm Holdings |
| First release | 2011 (AArch64) |
| Type | 64-bit RISC |
| Endianness | Little-endian, big-endian (configurable) |
| Registers | 31 general-purpose, 32 SIMD/FP |
| Extensions | NEON, SVE, SVE2, VHE, Pointer Authentication |
ARM64 is the 64-bit execution state of the Arm architecture, introduced to provide wider registers, a larger address space, and modern instruction encodings for servers, desktops, mobile devices, and embedded systems. It evolved through collaborations among companies and institutions to address needs in performance, power efficiency, and ecosystem support across computing platforms. The architecture underpins silicon from multiple vendors and is supported by major software ecosystems, compilers, and operating systems.
Overview
ARM64 emerged from specifications developed by Arm Holdings and partners such as Apple Inc., Qualcomm, Samsung Electronics, NVIDIA, Broadcom, Marvell Technology Group, MediaTek, and Huawei Technologies. Standards bodies and industry consortia including JEDEC, Trusted Computing Group, Linaro, and the OpenPOWER Foundation influenced interoperability and system-level features. Academic and research institutions like Massachusetts Institute of Technology, Stanford University, University of Cambridge, and EPFL contributed microarchitecture research that informed branch prediction, out-of-order execution, and power management features. Major system integrators such as Dell Technologies, Lenovo, HP Inc., Microsoft Corporation, and Amazon (company) adopted ARM64 for servers, desktops, and cloud instances.
Architecture
The ARM64 execution state, defined by Arm architecture specifications, provides 31 general-purpose 64-bit registers and a separate program counter, plus 32 SIMD/FP registers. Implementation choices have been adopted by foundries and design houses like TSMC, GlobalFoundries, Samsung Foundry, Intel Corporation, and GFLOPS Research Lab for process nodes. Architectural features such as memory translation and protection units, trustzone-like security, and support for virtualization are leveraged by hypervisor developers including VMware, Xen Project, KVM, and Hyper-V. System-on-chip integrators such as Apple Inc. (Apple Silicon), Qualcomm (Snapdragon), NVIDIA (Grace), and Ampere Computing tailored the register file and exception levels to meet requirements of cloud providers like Amazon Web Services, Google Cloud Platform, Microsoft Azure and research centers such as CERN and Los Alamos National Laboratory.
Instruction Set and Extensions
The 64-bit instruction set introduced AArch64 encoding and maintained a separate AArch32 state; extensions expanded capabilities with SIMD and vector processing. Vector and HPC-focused features were advanced by collaborative work among Arm Holdings, Cray Research, Intel Corporation, NVIDIA, and national labs such as Oak Ridge National Laboratory. Key named extensions include NEON (SIMD), SVE and SVE2 (Scalable Vector Extension), Pointer Authentication (PAC), and Large System Extensions adopted by enterprise vendors such as Oracle Corporation and IBM. Cryptography and security primitives were influenced by standards organizations like IETF and NIST and implemented by silicon vendors including Broadcom and Marvell Technology Group.
Implementations and Microarchitectures
Silicon implementations span in-house designs and licensed cores: notable commercial cores from Arm Holdings include designs used by Apple Inc. in custom A-series (Apple) and M-series (Apple) chips, while licensees such as Qualcomm produced Kryo cores and Samsung Electronics produced Exynos series. Data center focused microarchitectures were produced by Ampere Computing, Marvell Technology Group, Amazon (company) (Graviton), and NVIDIA (Grace). Academic prototypes and research cores appeared in labs at ETH Zurich, University of Illinois Urbana–Champaign, and Imperial College London. Foundry collaborations involved TSMC and Samsung Foundry for process node scaling to 7 nm, 5 nm, and below; packaging innovations engaged companies such as Intel Foundry Services and ASE Technology Holding.
Software and Operating System Support
Major operating systems added or expanded support: Linux, with distributions from Canonical (company), Red Hat, SUSE, and Debian, supports ARM64 kernels; Android (operating system) targets ARM64 for mobile devices from Samsung Electronics and Xiaomi. Apple transitioned desktop and laptop support with initiatives by Apple Inc.. Server and cloud stacks including OpenStack, Kubernetes, and container runtimes from Docker, Inc. and CoreOS were adapted by cloud providers like Amazon Web Services and Google Cloud Platform. Compiler and toolchain support was provided by projects and vendors such as GNU Project, LLVM Project, GCC, Microsoft Visual Studio, and Clang. Virtualization and orchestration vendors including VMware and Red Hat integrated ARM64 images and tooling.
Performance and Power Efficiency
ARM64 implementations focus on energy proportionality and performance-per-watt, a design goal emphasized by companies and labs such as Apple Inc., ARM Holdings, Intel Corporation, AMD, NVIDIA, Oak Ridge National Laboratory, and Lawrence Berkeley National Laboratory. Benchmarking and performance analysis used suites and organizations like SPEC, Phoronix, Geekbench, and research from Google Research and Microsoft Research. Power management innovations were influenced by mobile device practices from Samsung Electronics and Qualcomm and server optimizations required by hyperscalers such as Meta Platforms and Netflix, Inc..
Market Adoption and Applications
ARM64 is used across mobile devices from Samsung Electronics, Xiaomi, OnePlus, and Oppo; laptops and desktops from Apple Inc., Lenovo, and HP Inc.; servers from Amazon Web Services, Microsoft Azure, Google Cloud Platform, and enterprises including Oracle Corporation and SAP SE. Embedded and IoT deployments come from Texas Instruments, NXP Semiconductors, STMicroelectronics, and Raspberry Pi Foundation. High-performance computing adoption has been driven by collaborations involving Cray Research, HPE, Arm Holdings, and national labs such as Argonne National Laboratory and Sandia National Laboratories. The ecosystem includes software vendors like Canonical (company), Red Hat, Microsoft Corporation, and Oracle Corporation delivering optimized stacks for cloud, edge, and client markets.