Intel® Performance Counter Monitor (Intel® PCM) is a BSD-3-Clause licensed C++ API and command-line toolset from Intel for monitoring the performance and energy metrics of Intel® Core™, Xeon®, Atom™ and Xeon Phi™ processors.

What it is

Intel PCM lives in the Infrastructure & Operations / Monitoring & Observability space, and specifically in the hardware-level performance counter ecosystem of Intel processors. It provides both an application programming interface (API) and a set of tools built on that API, allowing software to read performance and energy metrics directly from the processor rather than inferring them from the operating system or from application-level instrumentation.

The concrete problem it solves is fragmented, per-platform visibility into processor behaviour. The same API and the same utilities apply across Linux, Windows, FreeBSD, DragonFlyBSD and ChromeOS, instead of each platform requiring its own separate means of gathering core frequency, memory bandwidth, cache or energy data. The topic list reflects this scope directly: cpu, energy, intel, monitor-performance, performance-counters, power, processor, processor-architecture, xeon and performance-visualization.

Key capabilities

• pcm, the basic processor monitoring utility, reports instructions per cycle, core frequency including Intel® Turbo Boost Technology, memory and Intel® Quick Path Interconnect bandwidth, local and remote memory bandwidth, cache misses, core and CPU package sleep C-state residency, core and CPU package thermal headroom, cache utilization, and CPU and memory energy consumption.
• pcm-memory monitors memory bandwidth per channel and per DRAM DIMM rank.
• pcm-accel monitors the Intel® In-Memory Analytics Accelerator.
• pcm-sensor-server acts as a collector that exposes metrics over HTTP in JSON or Prometheus (exporter text based) format, and is also distributed as a Docker container.
• A C++ API underpins all tools, with published API documentation, environment variables and custom compilation options.
• Dedicated documentation files cover integration: doc/PCM-EXPORTER.md, doc/PCM-SENSOR-SERVER-README.md and doc/DOCKER_README.md.
• Continuous integration covers Linux (linux_make.yml), FreeBSD (freebsd_build.yml), Docker (docker.yml) and Windows via AppVeyor, alongside CodeQL, OpenSSF Scorecard and OpenSSF Best Practices checks.

Who uses it and how

• Performance engineers profiling Intel Core, Xeon, Atom and Xeon Phi servers, reading instruction-level and cache-level counters from the command line.
• Capacity and energy analysts tracking CPU and memory energy consumption and package C-state residency on Intel hardware.
• Monitoring teams running pcm-sensor-server to scrape PCM metrics into Prometheus, typically for Grafana-style dashboards.
• Container-oriented deployments, where pcm-sensor-server runs as a prebuilt Docker image rather than being compiled on the host.
• Mixed-OS fleets, since the same toolset is used on Linux, Windows, FreeBSD, DragonFlyBSD and ChromeOS machines.

Getting started

Build the tools from source as described in the README's "Building PCM Tools" section, or download a pre-compiled PCM build from the "Downloading Pre-Compiled PCM Tools" section; pcm-sensor-server can alternatively be run as the Docker container documented in doc/DOCKER_README.md.

How it compares

This registry entry provides no list of paid products that Intel PCM replaces, and the facts name no directly comparable tools. It therefore stands alone in this registry as the vendor-authored, BSD-3-Clause implementation of Intel processor performance and energy monitoring.

When to use it — and when not to

Choose Intel PCM when the target hardware is an Intel Core, Xeon, Atom or Xeon Phi processor and the goal is direct hardware-level performance or energy measurement on Linux, Windows, FreeBSD, DragonFlyBSD or ChromeOS. It is the wrong tool for non-Intel CPUs, and it is not a general application-performance-monitoring product, so teams wanting language-level tracing or service-level dashboards should look elsewhere. A prospective user must also be prepared to build and install the toolset, or run the sensor server container, rather than expecting a single hosted service; the repository currently carries 69 open issues, so some fix-up work or workaround may be required.

git clone --recursive https://github.com/intel/pcm
cd pcm

git submodule update --init --recursive

cd build
cmake ..
cmake --build .

cmake --build . --parallel

cmake --build . --config Release