Bufferbloat Test
This bufferbloat test measures your ping three times: with the line quiet, while it downloads and while it uploads. You get an A to F grade for how much the ping rises under load, and router fixes if it rises too much.
Bufferbloat test tool
- Pick a test size. Data saver suits mobile data; Standard keeps a fast line busy for longer.
- Press Start test. It takes about 10–30 seconds.
- Read the grade and how much your ping rose while downloading and uploading.
- Changed a router setting? Test again and compare with your earlier runs below.
This test will use up to 37 MB of data.
Data saver uses up to 37 MB. Standard uses up to 217 MB and collects more samples on a fast line. On lines faster than about 60 Mbps, Data saver reaches its cap within a few seconds (about a second at 200 Mbps) and times fewer pings under load, so choose Standard for a firmer grade. A test never starts on its own.
Ready. Press Start test to begin.
Bufferbloat grade Our own A to F scale for how much your ping rises while the line is busy: A under 30 ms extra, B under 60, C under 150, D under 400, F 400 ms or more.
Your grade and a one-line verdict appear here after the test.
- Idle latency The median round trip of about 20 small HTTPS requests to the test server while nothing else runs in the test. This is your baseline. ms
- During download The median round trip of small requests sent every 300 ms while the test downloads as fast as your line allows. ms
- During upload The median round trip of small requests sent every 300 ms while the test uploads as fast as your line allows. ms
- Worst rise The larger of the two increases over idle latency (download or upload). The grade is based on this number. ms
Each box shows the middle half of the samples, the thick line is the median and the whiskers reach the lowest and highest sample. The table below lists the same numbers.
| Phase | Samples | Median (ms) | Middle half (ms) | Lowest – highest (ms) | Jitter (ms) | Rise vs idle (ms) |
|---|---|---|---|---|---|---|
| Idle | – | – | – | – | – | – |
| During download | – | – | – | – | – | – |
| During upload | – | – | – | – | – | – |
- Download speed
- Upload speed
- Test server
- Data used
- Test size
- Tested at
Your recent bufferbloat tests
| Date | Grade | Idle (ms) | Download (ms) | Upload (ms) | Size |
|---|
No tests yet. Your results stay in this browser, so you can compare before and after a router change.
How to read your bufferbloat result
Look at the rise, not the idle number. Idle latency mostly reflects distance to the server. The rise shows what your router does when the line is full, which is what freezes a video call while someone else downloads a film.
The grade is our own A to F scale, based on the larger of the two rises. The bufferbloat project itself calls a rise below 15–25 ms “under control”, so our A is a little looser than that.
| Grade | Ping rise under load | What you are likely to notice |
|---|---|---|
| A | under 30 ms | Calls and games stay steady while others download. |
| B | 30–59 ms | Small extra lag; fast games may feel it. |
| C | 60–149 ms | Clear lag in games and calls when the line is busy. |
| D | 150–399 ms | Calls freeze or people talk over each other. |
| F | 400 ms or more | Everything feels stuck while anyone uploads or downloads. |
Check which direction causes the rise. A high number only during upload points at uploads such as cloud photo backup, CCTV or your own camera on a call. A high number during download shows up when someone streams or downloads a large update.
What causes bufferbloat
Bufferbloat is the extra delay you get when your router or modem holds too much data in a queue. OpenWrt's SQM guide says it is most evident when a connection is heavily loaded with downloads or uploads, and that it hurts voice calls, video chat and online gaming.
The bufferbloat project's advice page points at the router, not the line: “link speed isn't the problem - it's your router buffering more data than necessary.” Without queue management, data piles up in the modem, the router or at the provider whenever the line is full, and the setting that manages it is rarely offered on provider-supplied routers. That is why a line can pass a speed test and still lag.
How to fix bufferbloat on your router
The fix is SQM (smart queue management). SQM makes the router send data slightly below your line speed, so the queue builds in the router, where it is managed. CAKE and fq_codel are the two common methods: OpenWrt prefers CAKE, and fq_codel uses less CPU. SQM runs on the router's processor, so a slow router may not keep up with a fast plan.
OpenWrt's guide describes these steps:
- Run a speed test with nothing else using the line.
- In the router's SQM settings, set download and upload to about 90% of those results.
- Choose CAKE (on OpenWrt, with the piece_of_cake.qos script) and turn off hardware flow offloading, which OpenWrt says does not work with SQM.
- Save and run this test again. If the rise is still high, lower the numbers a little; if it is fine, try raising them.
The bufferbloat project lists SQM or fq_codel in OpenWrt 22.03 or newer, GL.iNet routers, Asuswrt-Merlin (Adaptive QoS > QoS), eero, Ubiquiti, pfSense and OPNsense. If your provider's router has no SQM, a second router with SQM placed after it can fix the problem. Plain QoS settings are weaker: the project says they “only help a bit”.
Our general advice, not from those sources: until the router is fixed, pause cloud photo backup during important calls and use a LAN cable for the device where lag matters most.
How to retest after a change
Keep the device, the room and the test size the same, with nothing else using the line, and run the test two or three times. Your earlier runs stay in the table above, in this browser only. To check under real conditions, start a large download on another device and test again. For hours of data, use the internet stability test.
How this test works
The test uses the nearest Cloudflare speed test server, named in the result. It times about 20 small HTTPS requests with the line quiet, then downloads with one or two connections (up to four in Standard) while sending a timing request every 300 ms, and then does the same while uploading. The rise is the loaded median minus the idle median; the grade uses the larger rise and is our scale, not a standard.
Latency is the round trip of HTTP requests to Cloudflare, not ICMP ping. It covers your whole path, Wi-Fi included, and other traffic on your line also counts. Data saver stops at 25 MB down and 10 MB up (about 35 MB measured on a fast line); Standard at 150 MB and 60 MB. A very short load may not fill the router's queue, so a good Data saver grade on fast fibre is less certain. Cloudflare sees your IP address; results stay in this browser. Read the full method.
Questions about bufferbloat
Is a B grade good enough for online gaming?
A B grade means your ping rose by 30–59 ms under load in our test. You will mainly feel it when someone else downloads or uploads during a match, and fast shooters feel it most. The gaming ping test shows your base ping to cloud regions in India and nearby.
Can I fix bufferbloat on a JioFiber or Airtel Xstream router?
Look in the router settings for SQM, smart queue or a QoS option. We could not find a Jio or Airtel page confirming that their routers offer SQM, so do not count on it. A second router with SQM, placed after the provider's router, is the common workaround.
Will a faster internet plan fix bufferbloat?
A faster plan rarely fixes bufferbloat, because the delay comes from a queue in the router, not from the speed of the line. The bufferbloat project says this delay “can never be cured by faster transmission rates”.
Why is the bufferbloat test not working for me?
The bufferbloat test needs to reach speed.cloudflare.com. An ad blocker, a work firewall or some VPNs block it, and the test then shows an error instead of a grade. Allow the site for this page and press Start test again.
Why do other bufferbloat tests give me a different grade?
Each bufferbloat test uses its own servers, load time and grade scale. A longer or heavier load fills the router's queue more, so the rise can be larger. Compare results from the same test, test size and device.