Arm Blog

Learn how Arm Neural Frame Rate Upscaling works with Neural Super Sampling to improve motion, reduce rendering work, and support Unreal Engine games today

This blog post provides a brief overview of the MRS as discussed in a white paper. The white paper provides an introduction to the data structures, and potential use cases for working with the MRS dat…

Discover the winners of the Arm Create: AI Optimization Challenge and how their projects explore AI optimization across Cloud, Mobile, and Physical AI

Explore how ExecuTorch can partition and optimize one PyTorch model across Arm CPUs, GPUs, and NPUs using five target-specific execution paths

New Mali timeline and Dispatch graph views connect Vulkan submission, NX execution, shader activity, and system performance in one analysis workflow

How to find, optimize, and deploy AI models on Arm with pre-optimized models, performance insights, and agent-ready development workflows

Explore how Llama-Mobile uses S3D8 and quantization-aware training to reduce model size while retaining accuracy for multimodal AI on Arm mobile CPUs

Learn how to find and fix hidden graphics performance problems in mobile games using Arm Performance Studio.

Moving beyond a local-model demo requires more than an inference endpoint. Developers need an operational local-first runtime with explicit data boundaries

Discover how JDK 21 and JDK 25 improve AArch64 performance with enhanced C2 auto-vectorization, NEON, SVE, and new JVM intrinsics

First Contact explores an Android-native architecture for running on-device LLMs in mobile games, enabling real-time NPC dialogue

Learn how Elasticsearch on AWS Graviton5 delivers a strong balance of query throughput, median latency, tail latency and indexing performance.

Arm and Google Cloud partnership establishes the foundation for cloud-native software development. The collaboration enables a continuous shift-left workflow long before production hardware is availab…

Dynamic Insights connects Arm Performix run data to coding agents through MCP. Here is how we test that the generated performance advice is not just plausible, but grounded in profile evidence

Explore the latest Arm Performix features, from AI-assisted profiling to system analysis, code insights, and streamlined performance investigations

Learn how Harbor gained upstream arm64 support through native builds, multi-architecture releases, and upstream collaboration for Arm-based cloud platforms

Discover support.arm.com, Arm's new platform for product documentation, downloads, training, technical support, and customer services

Discover how the Arm SDS Framework enables deterministic testing, reproducible AI workflows, and faster Physical AI development on embedded devices

How the on-device Smart Gallery enables more natural, private user experiences on Android.

How an Agent Skill Turned Arm Performix into an AI Performance Optimization Workflow

Persistent AI is about more than local inference. Learn why CPU orchestration, memory, and runtime design are key to agent systems

Learn how Web AI brings local AI inference to the browser, improving responsiveness, privacy, and portability across Arm platforms

Learn how AMBA CHI C2C enables chiplet interoperability and scales for AI workloads through property negotiation for greater flexibility

Learn how Frame Advisor uses Vulkan debug labels to organize API calls, name graphics objects, and make complex frame captures easier to understand

How the Ambassador experience shaped my view of developer enablement

Discover how TPL plugins extend AMBA Viz with custom hardware signal, bus, and event analysis for more flexible RTL debug and performance analysis

Learn how hierarchical all-reduce and RDMA accelerate distributed LLM inference on Arm Neoverse, delivering up to 7.45× higher decode throughput