Frigate NVR hardware sizing
Best Hardware for Frigate NVR: Builds by Camera Count
Frigate is not sized well by one specification such as CPU model or camera count alone. Video decoding, simultaneous motion, object detection, enrichments, retention, storage writes and other services all place different demands on the host. This guide starts with the workload, shows live hardware immediately, and then sizes the complete NVR instead of recommending the same mini PC to every installation.
Quick answer
What this Frigate workload needs
For most home installations, start with a modern x86 host that supports AVX2, hardware video decoding, at least 8GB RAM, and a supported detector path. Frigate's current planning guidance says 16GB is highly recommended for many-camera or enrichment-heavy systems. Camera activity matters: six quiet cameras can be easier than four cameras pointed at a busy road.
Live Amazon hardware
Current products that fit this decision
These listings come from this sprint's dedicated Amazon catalogue. Barebones systems, accessories, camera kits, enclosures, replacement batteries, unrelated product classes and obvious multipacks are excluded by the normalizer.
Buying decision
Match the hardware to the bottleneck
Choose the decoder and detector architecture before choosing the box. A low-power Intel mini PC can be excellent for a smaller system when its iGPU handles decoding and OpenVINO or a dedicated accelerator handles detection. Larger or highly active systems may need more detector throughput, more memory, local recording capacity, higher PoE budget and better cooling.
Interactive decision tool
Frigate Hardware & Storage Sizer
Use the workload inputs to size a practical Frigate hardware tier before comparing live products. Results are planning guidance; verify current Frigate support, camera bitrates and the exact hardware configuration before deployment.
Compatibility checklist
Four checks before you purchase
Use hardware video decoding
A detector does not decode camera streams. Verify Intel, AMD or NVIDIA hardware acceleration separately.
Size by simultaneous activity
Busy overlapping cameras can require more detector throughput than a larger but quiet camera count.
Use supported detector hardware
Compare Hailo, OpenVINO, supported GPUs and other current Frigate detector paths before buying.
Calculate storage from bitrate
Retention depends on recording bitrate, duty cycle and days kept—not a fixed TB-per-camera rule.
Camera count is only the first sizing input
Frigate sends regions for object detection when motion is present, so simultaneous activity is often more important than the raw number of camera feeds. A quiet interior and a busy driveway do not create equal detector pressure. Frigate's planning guide separates low, moderate and high simultaneous activity and recommends stronger or multiple accelerators as activity increases. When buying hardware, count the cameras, then estimate how many may be active together during the busiest part of the day.
Hardware video decoding protects CPU headroom
Every camera stream still has to be decoded for features such as motion detection. Frigate strongly recommends integrated or discrete GPU acceleration for video decoding because an AI accelerator does not perform that job. Intel Quick Sync, AMD VCN and supported discrete GPUs can offload decode work that would otherwise consume CPU. A good detector paired with software-only decoding can still produce high CPU usage, so the host's media engine is part of the core design, not a bonus feature.
Detector choice changes the whole build
Frigate supports Hailo, Intel OpenVINO, NVIDIA and AMD GPU paths, Coral and other integrations. The current recommended-hardware guidance no longer treats Coral as the default choice for new installations, although it can still be useful in low-power or compatibility-constrained deployments. Hailo-8 and Hailo-8L are supported on compatible platforms, while modern GPUs can serve larger or more feature-heavy systems. Buy the host and detector as one compatibility decision.
RAM should include enrichments and growth
Frigate describes 4GB as a basic minimum for a very small accelerator-backed setup, 8GB as the minimum when enrichments are used, and 16GB as the recommended amount for most users with many cameras or heavier enrichment use. Those numbers are system planning points rather than a reason to buy the smallest possible memory configuration. Leave room for the operating system, Docker, Home Assistant, semantic search, face recognition, LPR and future camera additions.
Separate detection streams from recording streams
Frigate recommends configuring camera streams directly for the resolutions and frame rates you actually want. The detect stream can be lower resolution while the record stream stays high quality. The documentation recommends about 5fps for detection for most users and generally treats 10fps as the upper recommended range for fast-moving objects. Buying a larger CPU to decode needlessly high detection streams is usually worse than configuring the cameras correctly before spending more money.
Recording capacity must be calculated, not guessed
Storage demand depends on camera count, bitrate, recording mode and retention. Continuous 4K recording can consume many times the space of event-focused recording from lower-bitrate 1080p cameras. Cloudzat's calculator estimates raw video volume and then adds planning headroom. HDDs remain economical for long retention, while SSDs can be useful for smaller installations and high responsiveness. Surveillance-class HDDs such as SkyHawk and WD Purple are designed for sustained NVR workloads.
Local recording storage is the safest baseline
Frigate supports network storage, but its planning guidance generally favors local recordings for performance and reliability. NAS recording introduces network latency and another dependency between the camera system and its footage. If a NAS is part of the design, use robust wired networking and verify that the array can absorb sustained writes. A common architecture is local Frigate recording storage plus a separate backup or archive workflow rather than making live recording dependent on a remote share.
PoE capacity is part of the NVR budget
A Frigate host can be perfectly sized while the switch is not. Each PoE camera consumes power, and switches advertise both port count and a total PoE budget. Eight ports do not guarantee enough power for eight higher-draw cameras, especially when PTZ, heaters, IR illumination or Wi-Fi bridges are involved. Buy with spare ports and spare wattage so a future camera does not force a second switch. Gigabit per-port speed is normally ample for cameras, but uplink design still matters.
UPS sizing should include the network path
Keeping only the server on battery does not preserve recordings if the PoE switch and router immediately shut down. Size the UPS around the Frigate host, recording drives, PoE switch, router and any critical network equipment. Runtime needs vary: a few minutes may be enough for graceful shutdown, while security-focused installations may want substantially longer. Use actual measured wattage where possible and treat Amazon VA ratings as a starting point rather than a guaranteed runtime.
Home Assistant sharing changes available headroom
Frigate is commonly paired with Home Assistant, but whether both services share one machine affects sizing. A lightly loaded Home Assistant instance may add little pressure; additional add-ons, databases and automation workloads can make memory and storage contention more important. If the host also runs Plex, Immich, databases or VMs, size for the combined peak rather than Frigate in isolation. The calculator includes an enrichment and workload margin, but production monitoring should guide future upgrades.
Thermals matter in an always-on vision workload
NVR workloads are sustained. Compact systems that benchmark well for a few minutes may behave differently after hours of decoding, detection and storage activity. Keep air intakes clear, avoid stacking hot mini PCs and switches, and consider the heat from internal or nearby drives. Discrete GPUs and multiple HDDs can move a build beyond what a fanless appliance should handle. Stable temperatures and predictable fan behavior are often more valuable than a slightly faster burst benchmark.
Build for the next camera, not only today's list
The best value point is rarely the smallest configuration that works on day one. Leave spare memory, storage bays, PoE ports and detector headroom if the installation is likely to grow. A four-camera household planning eight cameras next year should not optimize around four quiet streams today. Cloudzat's recommendations deliberately show multiple hardware classes so the buyer can choose between a low-power appliance and a more expandable system instead of receiving one universal winner.
Questions people ask
Best Hardware for Frigate NVR questions
How much RAM does Frigate need?
Frigate currently describes 4GB as a basic minimum, 8GB as a minimum for enrichments, and recommends 16GB for many-camera or enrichment-heavy systems. Leave extra headroom for the OS and other services.
Does Frigate need a GPU?
A GPU or integrated graphics device is strongly recommended for video decoding. Object detection can use a separate accelerator such as Hailo, OpenVINO-capable Intel hardware, or a supported GPU path.
Is Google Coral still recommended for new Frigate builds?
Frigate currently says Coral is no longer recommended for most new installations, except particularly low-power deployments or hardware that cannot use newer detector options.
What FPS should I use for Frigate detection?
Frigate recommends about 5fps for detection in most cases and says up to about 10fps may be useful for very fast-moving objects. Configure the camera stream directly when possible.
Can Frigate record to a NAS?
Yes, but Frigate generally recommends local recording storage for best performance and reliability. A NAS adds network and storage dependencies that should be engineered carefully.
Is an Intel N100 enough for Frigate?
It can be a good small-system host when hardware decoding and a suitable detector are used. Camera activity, enrichments and other services determine whether more CPU and RAM headroom is needed.
How do I size Frigate storage?
Use camera count, average recording bitrate, recording duty cycle and retention days. Add free-space headroom rather than filling the recording volume to its theoretical maximum.
Do I need a PoE switch for Frigate?
Not for every camera, but wired PoE is common because it combines network and power. Check both the number of PoE ports and total switch power budget.
Should Frigate and Home Assistant share one mini PC?
They can for modest workloads. If the system also runs enrichments, databases or many other services, size memory, CPU and storage for the combined peak.
What is the best Frigate detector?
There is no universal detector. Match Hailo, OpenVINO, GPU or other supported hardware to camera activity, model needs, power target, interface availability and budget.
Official references and methodology
Verify current Frigate support before deployment
Cloudzat sizes Frigate hardware from current official guidance on camera activity, decoding, detector support, memory and storage. Live products come only from this sprint's dedicated Amazon catalogue and are filtered by product class. Prices and seller claims are not used to override Frigate's compatibility documentation.
- Frigate - Planning a New Installation
- Frigate - Recommended Hardware
- Frigate - Video Decoding
- Frigate - Object Detectors
- Frigate - Camera Setup
- Frigate - Enrichment Hardware Acceleration
- Frigate - Installation
As an Amazon Associate, Cloudzat may earn from qualifying purchases. Prices, seller terms, exact configurations, camera firmware and Frigate support can change.