How the test works
Idle baseline. We time small HTTP round-trips while the line is quiet — that is your true latency.
Saturated load. We start several concurrent large-file transfers to push the link close to capacity, and keep sampling latency and packet loss while it is loaded.
The gap. Loaded latency minus idle latency is your bufferbloat. The bigger and more consistent the jump, the more your router buffers are flooding interactive traffic with a large queue.
All computation happens in your browser; no speedtest server signup, no plugin. Results depend on your current network state — run it when the line is otherwise quiet for a clean reading.
What your score means
| Added latency | Reading | You will feel it |
|---|---|---|
| ≤ 20 ms | Good | Barely — SQM/CAKE is likely already active or the link is well-provisioned. |
| 20–50 ms | Fair | Noticeable in games and calls during a download. |
| 50–150 ms | High | Clear bufferbloat — background traffic stalls interactive ones. |
| > 150 ms | Severe | Classic "bufferbloat" — enable SQM and cap the queue. |
Bufferbloat test results explained
The tool reports four numbers. Idle latency is your baseline round-trip time (RTT) while the line is quiet — this is the “ping” most people know. Loaded latency is the same RTT measured while the connection is saturated with concurrent downloads: that is when oversized router buffers fill up. Bufferbloat is simply the difference between the two — the lag your connection only shows under load, and the number the score card above judges. Packet loss counts probes that timed out entirely during the loaded phase, which points at congestion severe enough that buffers overflow.
One method note: this test measures HTTP round trips from your browser, not ICMP ping from a console. Absolute numbers land a few milliseconds higher than ICMP, but the idle-to-loaded difference — which is what bufferbloat grading is about — is unaffected. For jitter (packet-to-packet variation), run the dedicated jitter test.
What counts as a bad bufferbloat score
The judge is the added-latency delta, never your absolute ping. Under 20 ms of added latency you will barely feel background traffic at all. Between 20 and 50 ms, competitive gamers notice input lag and video callers see a quality dip whenever a download starts. From 50 to 150 ms the effects are public: calls stutter, web pages hesitate on the second click, games rubber-band. Beyond 150 ms your line effectively collapses under load — the classic slideshow moment where a single OS update makes the whole household shout.
Counterintuitively, gigabit lines can score worse than 100 Mbit lines. Big pipes are paired with big buffers, and a saturated upload — tiny on most consumer plans — queues ACK packets that stall even the download direction. That is why the fix is almost never “more bandwidth” but SQM (CAKE) queue management that caps the queue slightly below your real rate.
Third-party graders such as Waveform use similar bands and an A+–F letter scale; if two tools disagree, trust the one whose loaded phase actually saturates your line for long enough.
How this compares to Waveform and Speedtest
Different tools talk to different servers, so absolute numbers are not directly comparable between them — the idle-to-loaded trend is. What you can compare is methodology and what each tool optimizes to show you.
Waveform is a free, donation-backed community tool with the friendliest bufferbloat grading (A+–F). Speedtest.net optimizes for throughput against a nearby dedicated server; its loaded-latency figure exists but hides in the detailed results. DSLReports popularized the loaded-latency test in the first place, and Cloudflare’s speed test measures quality metrics against its own anycast edge.
This test differs in three ways: it runs purely over HTTP in any browser with no plugin or account, it puts idle and loaded latency on one screen with the delta graded, and its method is fully open — the same measurement engine backs the NetSteer router app, so you can verify browser numbers against what your router actually does with SQM enabled.
How to fix bufferbloat
The standard fix is queue management (SQM with the cake plugin): it keeps the router’s queue near-empty by pacing traffic to your real bandwidth, so a background download stops monopolising the line. Three things matter:
- Set SQM to your measured upstream speed (not the advertised number).
- Use the slow path for latency-critical traffic (games, calls, DNS).
- Enable it on the router, not per-device — one router buffers for everyone.
On OpenWrt, NetSteer does all of this in one toggle: it auto-configures cake for your bandwidth, spreads packet receive across all cores, and gives you a Fullcone NAT for P2P. See our OpenWrt SQM guide. What bufferbloat is and how to fix it walks through pfSense, DD-WRT and consumer routers.
Frequently asked
What is bufferbloat?
Bufferbloat is high latency that appears only when a connection is near capacity, caused by large router buffers. A line that is fast when idle can add 100–500 ms of delay for a game or video call while a background download is running. The fix is queue management (SQM/CAKE) that keeps buffers small.
How does this test work without installing anything?
It runs in your browser. We measure idle latency with small round-trip timings, then fire several concurrent large-file transfers to saturate the link, sampling latency and packet loss while it is loaded. The gap between idle and loaded latency is the bufferbloat signal.
Is a result of 100 ms bufferbloat bad?
Yes. Anything above ~30–50 ms of added full-load latency will make games laggy and calls drop during a download. Good modern links (with CAKE/SQM) stay under ~20 ms. The tool scores your result against these bands.
How do I fix bufferbloat?
Enable SQM with the cake plugin on your router (OpenWrt, pfSense, DD-WRT), set it to your measured upstream bandwidth, and use the slow path for latency-sensitive traffic. NetSteer does this in one toggle on OpenWrt. See our OpenWrt SQM guide.
Why we open-source
The bufferbloat fix is only half the job — NetSteer runs on your router. Star the project on GitHub and follow the releases. ★ Star NetSteer on GitHub
More free tools from this site: ping test · jitter test · packet loss test · DNS leak test · DNS speed test · comparisons with Speedtest, DSLReports and Cloudflare.