The Three Metrics Defined
| Metric | Plain Meaning | Measured As | Good Value | Matters Most For |
|---|---|---|---|---|
| Bandwidth | Maximum capacity of the path | Mbps or Gbps | Depends on plan (100–1000 Mbps typical) | Plan sizing, concurrent users, large downloads |
| Throughput | Actual achieved transfer rate | Mbps or Gbps | 90–98% of bandwidth is excellent | Downloads, uploads, streaming, backups |
| Latency | Round-trip delay | Milliseconds (ms) | <20ms excellent, <50ms good, >100ms poor | Gaming, video calls, remote desktop, web browsing |
| Jitter | Variation in latency over time | Milliseconds (ms) | <5ms excellent, <20ms acceptable | Voice calls, video calls, live streaming |
| Packet loss | Percentage of packets that never arrive | Percent (%) | 0% ideal; >1% causes visible problems | Everything interactive |
The Highway Analogy
Bandwidth is the number of lanes on a highway. Throughput is how many cars actually got through in an hour. Latency is how long it takes one car to drive from one city to another. You can have a 10-lane highway (gigabit bandwidth) and still have a 2-hour drive to the destination (high latency). More lanes do not shorten the drive.
Why High Bandwidth Can Still Feel Slow
A 1 Gbps plan delivers fast bulk transfers but cannot shorten the fundamental round-trip time to a server. Interactive applications — games, video calls, remote desktop — need consistent timing more than raw capacity. Specific examples:
- A game with 300ms ping feels unplayable even on gigabit fiber, because the server processes your input 300ms late every time
- A video call on 500 Mbps with 30ms jitter will have choppy audio despite the bandwidth headroom
- A web page loading from a distant origin server still requires many round trips — high latency makes each one slow regardless of bandwidth
How Throughput Falls Below Bandwidth
Your measured throughput is almost always lower than your plan's advertised bandwidth because of:
- TCP slow start: transfers ramp up gradually, which limits short transfers
- Protocol overhead: headers, acknowledgements, and encryption reduce payload efficiency by 2–10%
- Wi-Fi: wireless introduces variable throughput based on signal quality, interference, and distance
- Server limits: many servers cap individual connection speed below your plan speed
- Shared infrastructure: cable and fixed-wireless connections share capacity with neighbors during peak hours
Latency Targets by Activity
| Activity | Acceptable Latency | Ideal Latency | Impact of High Latency |
|---|---|---|---|
| Competitive gaming (FPS, fighting) | <40ms | <15ms | Input lag, hit registration failures |
| Casual gaming (RPG, strategy) | <80ms | <40ms | Noticeable delay; tolerable |
| Video calls | <100ms | <50ms | Conversation overlap, awkward pauses |
| VoIP/audio calls | <150ms | <50ms | Echo, overlap above 150ms |
| Remote desktop | <80ms | <30ms | Visible input delay, sluggish feel |
| Web browsing | <200ms | <50ms | Slow page loads; DNS and TCP setup each add latency |
| Video streaming | <500ms | Any | Buffering on high jitter; latency alone rarely causes issues |
| Large file downloads | Any | Any | Minimal impact; throughput is what matters |
Bufferbloat: Where Bandwidth and Latency Collide
Bufferbloat is what happens when large buffers inside routers and modems soak up excess bandwidth but add hundreds of milliseconds of latency while doing so. A 500 Mbps plan can show 400ms ping when a large download is running — not because the bandwidth is gone, but because the queue is backed up. This is one of the most common reasons a speed test shows great results but video calls fall apart when someone else downloads a game update. SQM (Smart Queue Management) on routers like those running OpenWrt or pfSense can fix bufferbloat without sacrificing throughput.
Which Metric Should You Fix?
| Symptom | Likely Metric Problem | What to Check |
|---|---|---|
| Downloads are slow | Low throughput | Speed test on Ethernet; Wi-Fi signal; server limits |
| Video calls are choppy | High jitter or packet loss | Upload saturation; Wi-Fi; jitter test |
| Games feel delayed | High latency or packet loss | Ping to game server; route quality; Ethernet vs Wi-Fi |
| Everything slows down when one person downloads | Bandwidth saturation + bufferbloat | SQM / QoS settings; shared plan congestion |
| Speed test is good but experience is bad | Latency, jitter, or routing | Run a ping test alongside the speed test; check Wi-Fi |
| One service is slow, others are fine | Routing or peering | Traceroute to that service; test via VPN |
Frequently Asked Questions
Is bandwidth the same as throughput?
No. Bandwidth is the capacity of the path — the maximum the connection could carry under ideal conditions. Throughput is the actual rate achieved during a real transfer, after accounting for overhead, Wi-Fi, protocol behavior, and server limits. A 500 Mbps bandwidth plan might deliver 480 Mbps throughput in a speed test, but only 150 Mbps when downloading from a slow server.
Is latency more important than bandwidth?
It depends on what you are doing. For gaming, video calls, and remote desktop, latency and jitter often matter more than having a faster plan. For large file downloads, dataset pulls, or 4K streaming, bandwidth and throughput matter more. Many households need both — enough bandwidth for concurrent users, and low enough latency for one person gaming or calling.
Can a 1 Gbps connection still have bad latency?
Yes. Plan speed and latency are separate. A gigabit cable plan can show 150ms ping if the route to a game server is poor, or 300ms under load if bufferbloat is severe. Conversely, a 50 Mbps fiber plan may show 5ms ping because fiber links have lower propagation delay and ISPs with good peering. The plan tier does not guarantee good latency.
What is jitter and why does it matter for calls?
Jitter is the variation in latency from one packet to the next. If packets arrive consistently 30ms apart, audio codecs can compensate. If packets arrive 20ms, then 80ms, then 10ms, the audio buffer cannot smooth out the variation and voices sound robotic or clipped. A jitter buffer on your device helps but has limits — reducing jitter requires fixing Wi-Fi instability, packet loss, and upload saturation at the source.