2026-09-19 · IPV4 Scanner
Internet packages are sold on one number, and it is usually the least interesting one. Plenty of people pay for 500 Mbps to do things that work perfectly on 25, and are then baffled when video calls still stutter.
Here is what each activity actually needs, and which numbers matter more than the headline.
| Activity | Realistic need |
|---|---|
| Web browsing, email, messaging | 5 Mbps |
| Music streaming | 1–2 Mbps |
| HD video (1080p) | 5–8 Mbps |
| 4K video | 15–25 Mbps |
| Video call, one participant | 2–4 Mbps each way |
| Online gaming | 3–6 Mbps, but latency is what matters |
| Large file uploads, backups | as much upload as you can get |
These add up rather than compete, so size for what happens simultaneously. A household streaming two 4K shows while someone is on a video call and a console is downloading a game is using perhaps 70 Mbps at peak. A 100 Mbps connection handles that comfortably. A 1 Gbps connection is not doing anything the 100 could not.
Most consumer connections are asymmetric — a large download figure and a much smaller upload one. Cable packages advertising 300 Mbps down frequently offer 10 or 20 up, and DSL can be far worse.
That was a sensible design when people mostly consumed. It is a poor fit now. Video calls, cloud backups, uploading footage and working from home all lean on upload, and it is nearly always the upload figure that causes the trouble people blame on their download speed.
If calls break up while everything else seems fine, look at upload before anything else. A single cloud backup running quietly in the background can saturate a 10 Mbps upload and ruin every call in the house.
Bandwidth is how much fits through the pipe. Latency is how long a round trip takes. They are independent, and for anything interactive, latency is what you feel.
Gaming is the clearest case. A 50 Mbps connection at 20 ms latency is a better gaming connection than a 1 Gbps connection at 150 ms, and no amount of extra bandwidth improves the second. Satellite is the extreme example: plenty of bandwidth, latency dictated by the distance to orbit.
Video calls care about latency and jitter — the variation between packets. Steady 80 ms is fine. Bouncing between 20 and 200 is what makes people talk over each other.
If everything is fine until somebody starts a big download, and then calls break up and pages crawl, that is almost certainly bufferbloat. Oversized buffers in your router or your provider's equipment queue packets rather than dropping them, and latency under load goes from 20 ms to several hundred.
The headline speed is unaffected, which is why a speed test says everything is fine while the connection is plainly unusable. Test latency while something is downloading and compare it with idle latency. A large gap is the diagnosis.
The fix is usually router-side queue management rather than a faster package. Upgrading your speed does not help and often makes it marginally worse.
Several reasons, most of them unremarkable. Wi-Fi is usually the limiting factor rather than the line, especially on 2.4 GHz or through walls. Older devices cap out well below modern connection speeds. Anything above roughly 100 Mbps needs gigabit Ethernet end to end, and one old cable or switch caps the lot. And a single connection rarely saturates a fast line — you need parallel transfers to see the full figure.
Test over Ethernet before concluding anything about your provider. The speed test reports upload and latency alongside download, which is where the useful answers usually are.
Is 100 Mbps enough for a family?
For nearly all households, comfortably. Multiple 4K streams, video calls and downloads fit inside it. Upload capacity and Wi-Fi coverage are far more likely to be your constraint.
Do I need gigabit internet?
Rarely. It helps if you routinely move very large files or have many simultaneous heavy users. For normal use it mostly buys a bigger number, and your Wi-Fi and devices will often not deliver it anyway.
Why do my speed test results differ every time?
Because you are measuring a changing system: server choice, Wi-Fi conditions, other traffic in the house and time of day all move the figure. Test several times over Ethernet and look at the pattern, not one result.
What is a good latency?
Under 30 ms is excellent, under 60 is fine for everything including gaming, over 150 is noticeable in anything interactive. Consistency matters as much as the number.
Will upgrading my package fix buffering?
Usually not. Buffering is more often Wi-Fi, an overloaded router, or bufferbloat. If a wired test already shows well above what the stream needs, extra bandwidth changes nothing.