Download Speed Test: What Mbps You Need and What the Numbers Really Mean
Download speed test decoded: Mbps vs MB/s math, requirement tables by activity, how tests can lie with bursts, and how to read sustained vs peak honestly.
Quick answer: A download speed test measures how fast data flows to your device in megabits per second. 25 Mbps handles one 4K stream; 100 Mbps handles a household; 500+ only matters for heavy households and giant downloads. What actually matters is sustained throughput, not the first burst.
Mbps, MB/s and the confusion every store page exploits
Plans are sold in megabits per second (Mbps, small b). Browsers and Steam show megabytes per second (MB/s, capital B) — 8 bits to a byte. The quick math: divide Mbps by 8. "300 Mbps fiber" is at best ~37 MB/s in the download queue — and after protocol overhead, 33-35 MB/s is a perfect result.
What download speed do you actually need?
| Use case | Comfortable | Bare minimum |
|---|---|---|
| Email, browsing, messaging | 10-25 Mbps | 2-5 Mbps |
| HD (1080p) streaming per stream | 5-8 Mbps | 3.5 Mbps |
| 4K streaming per stream | 20-30 Mbps | 15 Mbps |
| Game downloads (patient) | 25-50 Mbps | 10 Mbps |
| Big household / multi-4K / WFH | 200-500 Mbps | 100 Mbps |
| Multi-gigabit media servers | 1-2.5 Gbps | 500 Mbps |
Notice the pattern: past the 100-300 Mbps range, download speed stops being the experience bottleneck for most homes. When something feels slow at 300+ Mbps, the cause is usually ping, jitter, Wi-Fi or bufferbloat — the topics covered in the latency guide and the bufferbloat guide.
How download testing works — and how it can lie to you
A client downloads test data, measures bytes over time, and averages. The lies hide in the details: a short test can ride a burst (cache, TCP slow-start, turbo windows some ISPs grant to test traffic) and report 2-3x sustained reality; a single-connection test on a high-latency line under-reports because throughput is capped by the bandwidth-delay product. A trustworthy download test ramps up, uses several parallel connections, and averages the sustained window — exactly the protocol Toolfyra Internet Speed Test follows, with a live graph so you can see the difference between burst and steady state yourself.
Reading a download result honestly
- Spike then collapse: the first seconds ride fast-start; trust the plateau. If the graph decays steadily, suspect ISP congestion or a data-cap overage policy.
- Plan 500, test 380 on Wi-Fi: normal — check the wired number before writing to your ISP (Wi-Fi guide).
- Plan 500, test 120 on ethernet: real evidence. Document three runs at different hours.
- Gigabit plan, test 940: textbook — near-plan results on ethernet are achievable on modern PON/fiber plants; see the gigabit fiber guide.
Downloads that matter more than the number
Latency while downloading is the underrated metric: a 400 Mbps connection that stalls every video call while updating a game is worse than a 150 Mbps line that stays smooth. This "speed under load" is exactly what bufferbloat measures. Test your second number: run a download, then check ping. If it triples, your router needs SQM/QoS more than your plan needs upgrading.
FAQ
Why does Steam/PS5 download slower than my speed test? Server-side throttling, peak-hour CDN load, disk write speed and MB/s-vs-Mbps confusion. A PS5 pulling 30 MB/s is downloading at ~240 Mbps — often the console’s realistic ceiling.
Is 300 Mbps overkill? For 1-2 people, mostly yes. For a family with simultaneous 4K, cloud backups and gaming, it is a sensible default.
Why do download speeds drop in the evening? Neighborhood congestion on shared infrastructure. Three tests at 9 am, 6 pm and 10 pm document it precisely.