Jellyfin Power Consumption Guide 2026: Real Measurements
Your Jellyfin server runs 24/7/365. That is 8,760 hours per year. Even a small difference in power draw adds up to real money over time. An old desktop pulling 80W costs $105/year in electricity. An Intel N100 mini PC pulling 10W costs $13/year. That is $92/year in savings, which pays for the mini PC itself in 18 months.
This guide provides real-world power measurements across common Jellyfin hardware, measured with a Kill-A-Watt meter under actual Jellyfin workloads.
How We Measured
- Tool: P3 Kill-A-Watt P4400 electricity meter
- Methodology: Measured at the wall outlet (includes PSU efficiency losses)
- States tested: Idle (server running, no streams), 1x 1080p transcode, 1x 4K transcode, 2x simultaneous transcodes
- Software: Jellyfin 10.11.7, Docker, Ubuntu Server 24.04
- Electricity cost: $0.15/kWh (US average 2026)
Power Measurements by Hardware
Intel N100 Mini PC (Beelink Mini S12 Pro)
| State | Power draw | Notes |
|---|---|---|
| Off (standby) | 0.8W | Wake-on-LAN enabled |
| Idle (Jellyfin running, no streams) | 7-9W | Typical: 8W |
| 1x Direct Play (1080p) | 8-10W | Almost no CPU increase |
| 1x Direct Play (4K HEVC) | 9-11W | Minimal CPU for muxing |
| 1x HW Transcode (4K to 1080p, QSV) | 12-16W | GPU active |
| 2x HW Transcode (4K to 1080p, QSV) | 15-20W | GPU at ~60% |
| 3x HW Transcode (4K to 1080p, QSV) | 18-24W | GPU near max |
| Library scan (full) | 14-18W | CPU + disk active |
| Tdarr encoding (QSV) | 18-25W | Sustained GPU load |
Annual cost at idle: 8W x 8,760h x $0.15/kWh = $10.51/year Annual cost typical use: ~12W average = $15.77/year
Intel N305 Mini PC (Beelink EQ13)
| State | Power draw | Notes |
|---|---|---|
| Idle | 10-13W | 8 cores, slightly higher baseline |
| 1x Direct Play | 11-14W | |
| 1x HW Transcode (QSV) | 16-22W | |
| 3x HW Transcode (QSV) | 22-30W | More headroom than N100 |
| Full load (Tdarr + streams) | 28-35W |
Annual cost at idle: 12W average = $15.77/year Annual cost typical use: ~18W average = $23.65/year
Synology DS224+ (NAS, Intel J4125)
| State | Power draw | Notes |
|---|---|---|
| Idle (drives spinning) | 15-18W | 2x HDD spinning |
| Idle (drives sleeping) | 8-10W | HDD hibernation enabled |
| 1x Direct Play | 16-20W | |
| 1x HW Transcode (QSV) | 20-25W | J4125 iGPU |
| Library scan | 22-28W | Drives active |
Annual cost: 17W average = $22.34/year
Note: NAS power includes the hard drives. A diskless mini PC + external USB drive would be comparable.
Raspberry Pi 5 (8GB)
| State | Power draw | Notes |
|---|---|---|
| Idle | 4-5W | Minimal baseline |
| 1x Direct Play | 5-7W | |
| 1x Software Transcode (1080p) | 12-15W | CPU at 100%, throttling |
| 4K Software Transcode | Not viable | Thermal throttle, unwatchable |
Annual cost: 5W average = $6.57/year
Cheapest to run, but no hardware transcoding. Direct Play only.
Old Laptop (2018 Intel i5-8250U)
| State | Power draw | Notes |
|---|---|---|
| Idle (lid closed) | 12-18W | Display off, Wi-Fi on |
| 1x Direct Play | 14-20W | |
| 1x HW Transcode (QSV) | 22-30W | 8th gen QSV |
| 2x HW Transcode | 28-38W |
Annual cost: 16W average = $21.02/year
Surprisingly efficient. Old laptops are underrated Jellyfin servers.
Old Desktop PC (2016 Intel i5-6500, no dGPU)
| State | Power draw | Notes |
|---|---|---|
| Idle | 45-65W | PSU inefficiency at low load |
| 1x Direct Play | 50-70W | |
| 1x HW Transcode (QSV) | 60-80W | 6th gen QSV |
| 1x Software Transcode (4K) | 90-120W | CPU at 100% |
Annual cost: 60W average = $78.84/year
This is where the money bleeds. An old desktop costs 5-7x more to run than an N100 mini PC.
Desktop PC with NVIDIA RTX 3060
| State | Power draw | Notes |
|---|---|---|
| Idle | 55-75W | GPU idle power + system |
| 1x Direct Play | 58-78W | |
| 1x HW Transcode (NVENC) | 70-95W | GPU video engine active |
| 5x HW Transcode (NVENC) | 90-130W | GPU at moderate load |
| Gaming + Transcoding | 250-350W | Full GPU + CPU |
Annual cost: 70W average = $91.98/year
Powerful but expensive to run 24/7. Best as a dual-purpose gaming/media PC that sleeps when not in use.
Full Rack Server (Dell PowerEdge R730, Dual Xeon)
| State | Power draw | Notes |
|---|---|---|
| Idle | 120-180W | Fans, dual PSU, ECC RAM |
| Active (multiple transcodes) | 200-350W |
Annual cost: 150W average = $197.10/year
Overkill for Jellyfin. Only justified if running 20+ other services alongside.
Summary Table
| Hardware | Idle watts | Typical watts | Annual cost | Best for |
|---|---|---|---|---|
| Raspberry Pi 5 | 5W | 6W | $7/year | Direct Play only, ultra-budget |
| Intel N100 Mini PC | 8W | 12W | $16/year | Best value (recommended) |
| Intel N305 Mini PC | 12W | 18W | $24/year | Multi-user, more headroom |
| Old Laptop (2018) | 15W | 20W | $26/year | Free hardware, decent efficiency |
| Synology NAS | 16W | 20W | $26/year | Already own it, dual-purpose |
| Old Desktop (2016) | 55W | 65W | $85/year | Wasteful, upgrade recommended |
| Desktop + RTX 3060 | 65W | 80W | $105/year | Dual-purpose only |
| Rack Server | 150W | 180W | $237/year | Overkill, enterprise only |
The Upgrade Break-Even Calculation
Should you buy an N100 mini PC to replace your old desktop?
| Factor | Old Desktop | Intel N100 Mini PC |
|---|---|---|
| Purchase cost | $0 (already own) | $130 |
| Annual electricity | $85 | $16 |
| Annual savings | - | $69/year |
| Break-even | - | 22 months |
After 22 months, the N100 has paid for itself in electricity savings. After 3 years, you have saved $77 net. After 5 years, $215 net.
If your old desktop draws 80W+ at idle, the upgrade pays for itself in under 2 years.
Tips to Reduce Power Consumption
1. Use hardware transcoding
Software transcoding uses 4-8x more power than hardware transcoding for the same task. Always enable Intel QSV or NVIDIA NVENC.
2. Optimize for Direct Play
Direct Play uses near-zero additional power. Encode your library in H.265 with AAC/EAC3 audio so clients Direct Play without transcoding.
3. Enable HDD sleep on NAS
If your media drives are not being accessed, let them spin down. Saves 5-8W per drive.
4. Schedule heavy tasks overnight
Tdarr encoding, library scans, and Intro Skipper analysis spike power draw. Schedule them during off-peak electricity hours if your utility offers time-of-use pricing.
5. Use Wake-on-LAN (advanced)
If your server is only used a few hours per day, configure Wake-on-LAN. The server sleeps at 1W and wakes when a client connects. Saves 50-80% of annual electricity for light-use servers.
Environmental Impact
| Hardware | Annual kWh | CO2 equivalent (US grid avg) |
|---|---|---|
| Intel N100 | 105 kWh | 44 kg CO2 |
| Old Desktop | 570 kWh | 239 kg CO2 |
| Rack Server | 1,314 kWh | 552 kg CO2 |
Switching from an old desktop to an N100 saves ~195 kg of CO2 per year. That is equivalent to driving 480 fewer miles.
FAQ
Does Jellyfin itself use much power? No. Jellyfin at idle uses <1% CPU. The power draw is dominated by the hardware platform (PSU efficiency, fans, drives), not the software.
Does the number of library items affect power? Minimally. A 10,000-item library uses the same idle power as a 100-item library. Power only increases during active scans or transcoding.
Is an N100 powerful enough despite using so little power? Yes. The N100 handles 3-4 simultaneous 4K transcodes via QSV while drawing only 20W. Power efficiency and performance are not mutually exclusive with modern hardware.
Should I turn off my server at night? For most users, no. The convenience of 24/7 availability outweighs the $1-2/month in overnight electricity. If you only use it 2-3 hours/day, Wake-on-LAN is a good compromise.
Know what your server is doing (and costing you). Monitor from your phone. Download JellyWatch on Google Play - CPU usage, active transcodes, and server health for Jellyfin.
On Emby? Download EmbyWatch on Google Play
Comments
No comments yet. Be the first to share your thoughts.
Leave a comment