Table of Contents
Hardware Specifications
The TetherBox software is compatible with Intel/AMD and ARM-based systems running Linux. This guide helps select appropriate hardware for various deployment sizes.
Quick Summary
Most TetherBox deployments need three things: a CPU with enough power for your camera count, enough RAM, and reliable storage for recordings — and an appropriate graphics processor if you require AI analytics. Here is a quick overview:
| Component | Rule of Thumb | Details |
|---|---|---|
| CPU | ~150 GeekBench 6 points per 4K camera | See CPU Requirements |
| RAM | 2 GB minimum + 1 GB per 4 cameras | See RAM Requirements |
| Storage | ~750 GB per 4K camera for 30 days (50% motion, 4096 kbit average) | See Storage Requirements |
| GPU | Only required for TetherX AI analytics | See Graphics / GPU |
All camera capacity estimates in this document assume 4K recording at 4 Mbit with H.264 High profile and analytics at 720p @ 5fps with H.264 Baseline profile. Using lower analytics resolution (≤480p) can increase capacity by 2-3x beyond these estimates.
Most popular configuration: An Intel N100-based mini PC with 8 GB RAM and a USB SSD handles up to 19 cameras. This covers the majority of deployments.
Raspberry Pi deployments: A Raspberry Pi 5 handles up to 11 cameras and is ideal for space-constrained locations. Add the Raspberry Pi AI Hat 2 with Hailo-10H to enable TetherX AI video and audio analytics on up to 4 cameras without impacting CPU performance.
For detailed specifications, continue to the sections below.
CPU Requirements
Benchmark: GeekBench 6 Multi-Core
We use GeekBench 6 Multi-Core scores as they reflect real-world multi-threaded performance, which is how TetherBox processes camera streams.
Rule of thumb: ~150 GeekBench 6 points per 4K camera
All scores in this document are GeekBench 6 Multi-Core (verified January 2026). GeekBench 5 scores are not comparable and should not be used.
To estimate camera capacity:
- Search for your CPU on GeekBench Browser or CPU Monkey
- Look up the GeekBench 6 Multi-Core score
- Divide by 150 to get estimated 4K camera count
Baseline estimates: Assume 4K cameras recording at 4Mbit with H.264 High profile on the primary stream, and a secondary stream at 720p @ 5fps with H.264 Baseline profile used for analytics (our recommended analytics configuration).
Warning: Always use H.264 Baseline or Main profile for analytics streams. High profile increases CPU load per camera by 30-50% due to more complex decoding requirements. Recording streams can use High profile without issue - only analytics processing is affected.
Warning: Strongly avoid H.265 (HEVC) for analytics streams whenever possible. H.265 decoding requires massively more CPU than H.264 regardless of profile - a single H.265 stream can consume as much CPU as 3-4 H.264 streams. Recording streams can use H.265 without issue - only analytics processing is affected.
Capacity scaling by analytics resolution and framerate:
- Low resolution (≤480p @ 5fps): 2-3× baseline capacity - sufficient for most analytics use cases including motion detection, object detection, and facial recognition
- Standard resolution (720p @ 5fps): Baseline capacity (as per estimates) - recommended default
- High resolution (≥1080p @ 5fps): ~50% of baseline (uses 2× more CPU per camera)
- High resolution (≥1080p @ >5fps): Even lower capacity - CPU usage increases further with framerate
Note: There is no practical benefit to using analytics resolution above 1080p unless you require very high-resolution time-lapse images.
Key insight: Recording and analytics use independent streams from the camera. Configure your camera's primary stream for recording (4K, High profile) and the secondary stream at 720p @ 5fps with Baseline profile for analytics — this is our recommended configuration and what all capacity estimates in this document are based on. If you need to support more cameras than your CPU allows, reduce the analytics secondary stream to ≤480p @ 5fps — field deployments show this can double your effective camera capacity with modest analytics accuracy loss.
CPU optimization options (reduce load without affecting recording quality):
- Analytics profile: CRITICAL - Use H.264 Baseline or Main profile for analytics (recording remains High profile) - High profile can increase load 30-50% per camera
- Analytics resolution: Lower to ≤480p (recording remains at full 4K resolution)
- Analytics framerate: Keep at 5fps (analytics doesn't need 25fps) - higher framerates dramatically increase CPU load
- Analytics bitrate: Limit to 1-2 Mbps for analytics stream (recording stream remains full quality)
Storage optimization (for systems with high IO wait >20%):
- Recording bitrate: Reduce to lower write throughput (e.g., from 4Mbit to 2-3Mbit)
- Recording profile: Switch from H.264 High profile to Main or Baseline profile
- Recording framerate: Reduce from 25fps to 15fps if high-framerate playback not required
- Recording resolution: Lower recording resolution if storage performance is severely limited
Example capacities for Intel N100 (2,840 GeekBench 6 points):
- 19 cameras at 4K recording @ 4Mbit (High profile) with 720p @ 5fps analytics (Baseline profile) - baseline configuration
- 38-57 cameras at 4K recording @ 4Mbit (High profile) with ≤480p @ 5fps analytics (Baseline profile) - field-validated
- 9-10 cameras at 4K recording @ 4Mbit (High profile) with ≥1080p @ 5fps analytics (Baseline profile)
CPU Reference Table
These are common CPUs used in TetherBox deployments. Some systems have been running reliably for over 10 years - we continue to support and push updates to all systems regardless of age.
Sorted by performance (highest to lowest). All scores are GeekBench 6 Multi-Core verified from official sources.
| CPU Model | Cores/Threads | Year | GeekBench 6 Multi | Est. Cameras |
|---|---|---|---|---|
| Apple M4 Max | 14C/32T | 2024 | 25,616 | 171 |
| AMD Ryzen Threadripper 7980X | 64C/128T | 2023 | 25,211 | 168 |
| Intel Core Ultra 9 285K | 24C/24T | 2024 | 22,580 | 150 |
| AMD Ryzen 9 9950X | 16C/32T | 2024 | 21,440 | 143 |
| Apple M3 Max | 14C/30T | 2023 | 20,933 | 140 |
| AMD Ryzen 9 9900X | 12C/24T | 2024 | 19,786 | 132 |
| AMD EPYC 9654 | 96C/192T | 2022 | 19,181 | 128 |
| Intel Xeon w9-3495X | 56C/112T | 2023 | 18,656 | 124 |
| Intel Core i5-14600K | 14C/20T | 2023 | 16,065 | 107 |
| Intel Core i5-13600K | 14C/20T | 2022 | 14,997 | 100 |
| AMD Ryzen 7 7700 | 8C/16T | 2023 | 14,828 | 99 |
| Apple M4 | 10C/10T | 2024 | 14,673 | 98 |
| Intel Core i7-13650HX | 14C/20T | 2023 | 13,939 | 93 |
| AMD Ryzen Threadripper 3960X | 24C/48T | 2019 | 13,167 | 88 |
| Apple M3 | 8C/8T | 2023 | 11,679 | 78 |
| AMD Ryzen 7 6800H | 8C/16T | 2022 | 8,860 | 59 |
| Apple M1 | 8C/8T | 2020 | 8,344 | 56 |
| AMD Ryzen Threadripper 1950X | 16C/32T | 2017 | 7,988 | 53 |
| Intel Xeon E-2336 | 6C/12T | 2021 | 7,772 | 52 |
| AMD Ryzen 5 5600G | 6C/12T | 2021 | 7,671 | 51 |
| AMD Ryzen 5 5500 | 6C/12T | 2022 | 7,652 | 51 |
| AMD Ryzen 5 PRO 5650G | 6C/12T | 2021 | 7,256 | 48 |
| Intel Xeon Silver 4309Y | 8C/16T | 2021 | 7,051 | 47 |
| AMD Ryzen 5 PRO 4650G | 6C/12T | 2020 | 6,106 | 41 |
| Intel Xeon E5-2670 v3 | 12C/24T | 2014 | 5,959 | 40 |
| Intel Core i5-8500 | 6C/6T | 2018 | 4,867 | 32 |
| Intel Xeon E-2224 | 4C/4T | 2019 | 4,163 | 28 |
| Intel Core i3-9100 | 4C/4T | 2019 | 3,653 | 24 |
| Intel N100 (most popular) | 4C/4T | 2023 | 2,840 | 19 |
| Intel Core i3-7100T | 2C/4T | 2017 | 2,228 | 15 |
| Raspberry Pi 5 B | 4C/4T | 2023 | 1,600 | 11 |
| Intel Pentium Silver N6005 | 4C/4T | 2021 | 1,435 | 10 |
| Intel Core i3-5010U | 2C/4T | 2015 | 1,336 | 9 |
| Intel Celeron N5105 | 4C/4T | 2021 | 1,264 | 8 |
| Intel Core i3-4010U | 2C/4T | 2013 | 1,062 | 7 |
| Intel Celeron J4105 | 4C/4T | 2017 | 928 | 6 |
| Intel Celeron J3455E | 4C/4T | 2017 | 796 | 5 |
| Intel Celeron N4505 | 2C/2T | 2021 | 763 | 5 |
| Intel Celeron J3455 | 4C/4T | 2016 | 745 | 5 |
| Raspberry Pi 4 B | 4C/4T | 2019 | 640 | 4 |
| Intel Celeron J1900 | 4C/4T | 2013 | 523 | 3 |
| Intel Celeron J3160 | 4C/4T | 2016 | 503 | 3 |
| Intel Celeron N2830 | 2C/2T | 2014 | 276 | 2 |
| Intel Celeron N3050 | 2C/2T | 2015 | 266 | 2 |
Tip: The Intel N100 (19 cameras) is our baseline reference at 2,840 points (150 points/camera). All scores verified from GeekBench 6 official databases. Actual performance varies based on camera resolution, motion percentage, and enabled features.
Key Considerations
Analytics Resolution: High resolution (≥1080p) analytics uses approximately 1.9× more CPU per camera than low resolution (≤480p) analytics. This is the most significant factor affecting camera capacity.
Storage Performance: If a TetherBox shows high CPU load despite being within camera recommendations, check the IO wait percentage on the dashboard. Values exceeding 20% indicate storage bottlenecks rather than CPU limitations. Storage issues can manifest as apparent CPU overload but are resolved by replacing storage hardware, not upgrading the CPU.
Recommendations by Deployment Size
| Cameras | Minimum Score | Example CPUs |
|---|---|---|
| 1-2 | 350+ | Intel Celeron N2830, Intel Celeron J3160 |
| 3-4 | 700+ | Raspberry Pi 4 (4 cam), Intel Celeron J3455 (5 cam) |
| 5-8 | 1,250+ | Intel Celeron J4105 (6 cam), Intel Celeron N5105 (8 cam) |
| 9-16 | 1,600+ | Raspberry Pi 5 (11 cam), Intel Core i3-7100T (15 cam), Intel N100 (19 cam) |
| 17-32 | 5,600+ | AMD Ryzen 5 5600G (51 cam), Intel Xeon E-2336 (52 cam) |
| 33-64 | 11,200+ | Apple M3 (78 cam), AMD Ryzen Threadripper 3960X (88 cam) |
| 65-128 | 22,400+ | AMD Ryzen 9 9950X (143 cam), Apple M4 Max (171 cam) |
| 129+ | 45,000+ | AMD Threadripper 7980X (168 cam) |
Raspberry Pi
Raspberry Pi units are ideal when you have power and/or space constraints. They can be placed on sides of lamp posts, in vehicles, hidden on sides of racks, or in other locations where traditional hardware cannot fit.
Key considerations:
Cooling is critical - Ensure the enclosure has good passive or ideally active (fan) cooling. Raspberry Pi units can last 7+ years, but only if cooling is adequate. Units that run hot will fail prematurely.
Storage for video recording - If using a Raspberry Pi to record video, you MUST use an M.2 HAT with an appropriate drive or adequate external storage. SD cards and USB memory sticks are not suitable for continuous video recording. See Recommended Storage Hardware for drive recommendations.
Raspberry Pi AI Hat 2
The Raspberry Pi AI Hat 2 is an M.2 HAT+ add-on board featuring the Hailo-10H AI accelerator with 40 TOPS of neural network inference performance. When installed on a Raspberry Pi 5, this enables TetherX AI video and audio analytics on up to 4 cameras — with all AI processing offloaded to the Hailo chip so the Raspberry Pi's CPU remains free for recording and other tasks.
What it enables:
- TetherX AI — AI-powered smart motion detection using object recognition to detect and classify people, vehicles, and animals, track their trajectory across frames, and make recordings searchable by what was seen
- TetherX Audio AI — Real-time audio event detection that classifies security-relevant sounds such as glass breaking, screaming, gunshots, smoke alarms, dog barking, and vehicle horns
Key details:
- Supports up to 4 cameras for AI analytics
- AI processing runs entirely on the Hailo-10H — no CPU impact on the Raspberry Pi
- Fits in the same M.2 HAT+ slot — compatible with Raspberry Pi 5 enclosures that support the AI Hat 2 form factor
- TetherBox automatically detects the Hailo accelerator and enables AI analytics when available
Note: The AI Hat 2 uses the M.2 slot, so if you also need local storage for recording, use USB storage (SSD recommended) or an additional HAT for the M.2 drive. See Recommended Storage Hardware for options.
RAM Requirements
Formula: 2GB minimum, add 1GB per 4 cameras for standard analytics
Standard Analytics (≤720p)
Most deployments use standard analytics resolution for optimal performance and capacity.
| Cameras | Minimum RAM | Example Systems | Notes |
|---|---|---|---|
| 1-4 | 2 GB | Portable, solar powered | Minimal deployments |
| 5-8 | 4 GB | MiniPC, embedded systems | Standard deployments |
| 9-16 | 4 GB | N100 fanless | Most common configuration |
| 17-24 | 8 GB | Rack units | Server-grade hardware |
| 25-32 | 8 GB | High-density rack | Multi-drive systems |
| 33-64 | 16 GB | Enterprise deployments | High camera count |
| 65-128 | 32 GB | Large installations | Purpose-built systems |
| 129+ | 64 GB | Massive deployments | Bespoke configurations |
High Analytics (>720p)
High-resolution analytics requires approximately 1.5× more RAM than standard analytics.
| Cameras | Minimum RAM | Example Systems | Notes |
|---|---|---|---|
| 1-4 | 4 GB | High-res portable | 30-40% extra headroom |
| 5-8 | 6 GB | High-res MiniPC | Increased requirements |
| 9-16 | 6 GB | N100 high-res analytics | Most common |
| 17-24 | 12 GB | High-res rack units | Server-grade |
| 25-32 | 12 GB | High-density high-res | Multi-drive |
| 33-64 | 24 GB | Enterprise high-res | High camera count |
| 65-128 | 48 GB | Large high-res | Purpose-built |
| 129+ | 64 GB | Massive high-res | Maximum capacity |
Warning: Analytics resolution significantly impacts RAM requirements. High analytics (>720p) uses approximately 1.5× more RAM than standard analytics (≤720p). Field validation shows systems with high-resolution analytics require 30-40% additional RAM headroom for stable operation.
Danger: Do not under-provision RAM to save costs. Systems with insufficient RAM (>85% usage) experience recording failures and instability. The recommendations above provide appropriate headroom for reliable operation.
Storage Requirements
Storage depends on resolution, motion activity, and retention period. Below are estimates for 30 days retention per camera including video recording and a 720p snapshot every 5 seconds (24/7):
| Resolution | Avg Bitrate | Motion % | Storage/Camera (30 days) |
|---|---|---|---|
| 720p | 1024 kbit | 50% | 250 GB |
| 1080p | 2048 kbit | 50% | 400 GB |
| 4K | 4096 kbit | 50% | 750 GB |
| 4K | 4096 kbit | 100% | 1.4 TB |
Tip: Multiply by camera count. For local and cloud storage capacity planning, see TetherBox Recording Capacity.
Recommended Storage Hardware
USB Storage Options (in Order of Preference)
- USB SSD (recommended) - Plug-and-play, high speed, reliable, excellent endurance
- USB HDD (24/7 rated) - Suitable for lower camera counts, must be surveillance/NAS-class
- USB to SATA/NVMe adapter - Flexible, use quality adapters from reputable manufacturers
Hard Drive Selection
| Cameras | Recommended Drive Type |
|---|---|
| 1-3 | 4800+ RPM HDD acceptable |
| 4-5 | 5400 RPM HDD minimum |
| 6-30 | 7200 RPM NAS/Surveillance HDD |
| 30+ | Multiple drives or RAID array |
Recommended HDD Models
Choose NAS or Surveillance-class drives designed for 24/7 high-write workloads:
- Seagate: IronWolf (NAS) or SkyHawk (Surveillance)
- Western Digital: Red (NAS) or Purple (Surveillance)
- Toshiba: N300 (NAS) or S300 (Surveillance)
Storage Warnings
- Never use USB memory sticks - Not designed for continuous writes, will fail quickly
- Avoid "basic/value" HDDs - Often fail after 1-2 years of 24/7 use
- Use quality USB adapters - Cheap adapters cause misreported failures or system freezes
- Avoid manual power switches - Enclosures requiring manual intervention after power loss are problematic
Diagnosing Storage Performance Issues
If a TetherBox shows high CPU load despite being within camera recommendations, check IO wait percentage on the dashboard:
- >20% IO wait: Storage bottleneck detected - investigate immediately
- Common causes:
- Failing USB storage devices (end of life, bad sectors)
- Poor quality USB controllers or adapters
- Slow or degraded RAID arrays
- Dying HDDs showing early failure signs
- USB 2.0 connections (should use USB 3.0+)
Storage bottlenecks can manifest as CPU overload but are resolved by replacing storage hardware, not upgrading the CPU. The TetherBox dashboard displays IO wait percentage under system metrics.
Graphics / GPU
Standard deployments: No dedicated GPU required. TetherBox uses onboard graphics or CPU for processing.
Hardware acceleration: TetherBox automatically utilises Nvidia, AMD, or Intel dedicated graphics when available.
Analytics: Out of the box, TetherBox leverages any analytics built into your Cameras or recorder. For edge analytics requirements (local AI/ML processing), please contact TetherX support to discuss GPU requirements for your specific use case.
Operating System Installation
For complete installation instructions including partitioning and configuration, see Installing Operating System.
Cloud Backup (optional)
For off-site redundancy, enable TetherX Cloud Backup to protect critical footage against theft, damage, or local disasters.
References
- CPU Monkey - Compare CPU performance across generations
- GeekBench Browser - Search specific CPU benchmarks
Referenced in: