Open-Source Observability · utkuozdemir

nvidia_gpu_exporter

nvidia_gpu_exporter is a Prometheus exporter that collects GPU metrics from NVIDIA hardware using the nvidia-smi binary. It works on Windows, Linux, and macOS without requiring Docker, Kubernetes, or compiled C bindings, making it accessible for gaming monitoring and multi-platform deployments.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/utkuozdemir/nvidia_gpu_exporter

1.5k

GitHub stars

148

Forks

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	utkuozdemir/nvidia_gpu_exporter
Owner	utkuozdemir
Primary language	Go
License	MIT — OSI-approved
Stars	1.5k
Forks	148
Open issues	23
Latest release	v1.10.0 (2026-07-05)
Last updated	2026-07-08
Source	https://github.com/utkuozdemir/nvidia_gpu_exporter

What nvidia_gpu_exporter is

Written in Go, this exporter parses nvidia-smi output to expose GPU metrics (memory, temperature, utilization, power) in Prometheus format. It supports remote execution of nvidia-smi, optional per-process GPU metrics, and auto-discovery of available metric fields across different driver versions and GPU models.

Quickstart

Get the nvidia_gpu_exporter source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/utkuozdemir/nvidia_gpu_exporter.git2cd nvidia_gpu_exporter3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Multi-platform GPU monitoring for development teams

Developers working on Windows, Linux, or macOS can monitor local GPU performance without Docker dependencies. Ideal for ML engineers, game developers, and researchers tracking training jobs or workload performance across heterogeneous hardware.

Prometheus-native GPU observability at scale

Organizations already invested in Prometheus/Grafana stacks can ingest GPU metrics directly into existing monitoring pipelines. The exporter works with standard Prometheus scraping and integrates seamlessly with alerting rules and dashboards.

Remote GPU metric collection

Execute nvidia-smi remotely (SSH, etc.) to monitor GPUs on machines where you cannot or will not run the exporter directly. Useful for managing GPU clusters or restricted environments.

Implementation considerations

Requires nvidia-smi binary to be installed and accessible on the target machine; verify driver compatibility before deployment.
For remote execution, ensure SSH or equivalent command channel is configured securely; credentials and key management are the deployer's responsibility.
Per-process GPU metrics add overhead; enable only if needed, and monitor exporter CPU/memory footprint in high-cardinality environments.
nvidia-smi output parsing is version and GPU-model dependent; the README explicitly requests hardware capture contributions for coverage.
Metric field names and availability vary across driver versions and GPU types; validate scraped metrics against your expected telemetry schema.

When to avoid it — and what to weigh

You need DCGM or enterprise GPU diagnostics — This exporter relies on nvidia-smi output and does not provide the comprehensive hardware diagnostics, health checks, or tensor-level metrics that NVIDIA Data Center GPU Manager (DCGM) offers.
You require real-time, high-frequency sub-second metrics — nvidia-smi has polling overhead and latency; the exporter cannot match the sub-millisecond precision of kernel-level or CUDA-based monitoring approaches.
You operate in air-gapped environments without nvidia-smi — The exporter is a wrapper around nvidia-smi and will not function if the binary is unavailable, missing, or incompatible with your driver version.
Your primary use case is container/Kubernetes orchestration at massive scale — Although it runs outside Docker, the exporter is simpler than enterprise solutions (DCGM Exporter, Kubernetes device plugins) and may lack advanced scheduling or resource-quota integration.

License & commercial use

Licensed under the MIT License, a permissive OSI-approved license that allows commercial use, modification, and distribution with minimal restrictions.

MIT License permits commercial use in proprietary and closed-source products without requiring disclosure or contributing changes back. However, ensure compliance with NVIDIA's nvidia-smi Terms of Service and driver licensing. Review NVIDIA's legal documentation separately.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Low
DEV.co fit	Good
Assessment confidence	High

Security considerations

No inherent security vulnerabilities claimed or provided in the data. Considerations: the exporter reads nvidia-smi output and exposes metrics on an HTTP endpoint—ensure the endpoint is not publicly accessible without authentication. For remote execution (SSH), secure the command channel and credential storage. Release artifacts are signed (GPG, cosign), supporting integrity verification. No known CVEs mentioned; audit dependencies as part of regular supply-chain review.

Alternatives to consider

NVIDIA Data Center GPU Manager (DCGM) Exporter

Enterprise-grade alternative providing comprehensive GPU diagnostics, health checks, and tensor metrics. Requires DCGM installation and is heavier than nvidia_gpu_exporter but suitable for production datacenters.

Prometheus NVIDIA GPU Exporter (a0s/nvidia-smi-exporter)

Original Python-based predecessor. Less portable (no Windows pre-built), requires Python runtime, and is less actively maintained. nvidia_gpu_exporter forked and rewrote it in Go for better portability.

Kubernetes NVIDIA GPU Device Plugin + kubelet metrics

Native Kubernetes integration for GPU resource scheduling and allocation metrics. Not a Prometheus exporter; designed for container orchestration rather than standalone GPU monitoring.

Software development agency

Build on nvidia_gpu_exporter with DEV.co software developers

Deploy nvidia_gpu_exporter to gain instant visibility into GPU metrics across your development, gaming, or AI workloads. MIT licensed, pre-built binaries ready for Windows, Linux, and macOS.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

snmp_exporter gpustat nvitop

Related on DEV.co

Explore the category and the services that help you build with it.

Open-Source Observability AI Development DevOps & Monitoring Talk to DEV.co

nvidia_gpu_exporter FAQ

Does this work on Windows without Docker?

Yes. The exporter runs as a standalone binary on Windows and queries nvidia-smi.exe directly. This is one of its key design advantages for gaming and local development monitoring.

Can I monitor remote GPUs without installing the exporter on each machine?

Yes, if you configure it to execute nvidia-smi remotely (e.g., via SSH). The exporter itself runs on one machine and queries GPUs on others by invoking the remote command.

What GPU metrics does this exporter expose?

Standard metrics include GPU utilization, memory usage, temperature, power consumption, and (optionally) per-process GPU memory. See METRICS.md for the full list. Available metrics depend on your driver version and GPU model.

Is this suitable for production Kubernetes clusters?

Possible, but not the primary use case. It is simpler than DCGM Exporter or Kubernetes device plugins. For large-scale orchestration, evaluate more feature-rich solutions. For single-node or small-cluster monitoring, it works fine.

Software development & web development with DEV.co

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If nvidia_gpu_exporter is part of your open-source observability roadmap, our team can implement, customize, migrate, and maintain it.

Start monitoring your GPUs with Prometheus today

Deploy nvidia_gpu_exporter to gain instant visibility into GPU metrics across your development, gaming, or AI workloads. MIT licensed, pre-built binaries ready for Windows, Linux, and macOS.

Talk to DEV.co Schedule a Consultation