Bufferbloat Test

See how your connection responds while it is busy. Run a manual download and upload measurement, then compare idle HTTP latency with latency under each load. You get the measured change in delay, without an unexplained letter grade.

Measure your connection

Before you start: this test deliberately transfers data to and from Cloudflare. A full run can use up to about 1.25 GB of test payloads, plus network overhead. It can take up to three minutes and may affect other activity on this connection. Start only when ready; you can stop at any time. No camera, microphone or location access is requested. How the measurement works.

Loading the measurement engine. No bandwidth test has started.

Final measured results

Only a completed, verified run fills these fields. Added delay is the loaded median minus the idle median; these are HTTP timings to Cloudflare, rather than a service’s actual media or game-server path.

Download capacity
Unavailable
Mbps
Upload capacity
Unavailable
Mbps
Idle HTTP latency
Unavailable
ms
Latency during download
Unavailable
ms
Latency during upload
Unavailable
ms
Idle HTTP timing variation
Unavailable
ms
Download added delay
Unavailable
ms
Upload added delay
Unavailable
ms

A complete run is required. Missing loaded latency stays unavailable.

How to use this tool

  1. Pause other speed tests, downloads and uploads. Keep this device and browser tab in view.
  2. Read the test-data note, then press Start Measurement. The test measures idle latency, download capacity and upload capacity in separate stages.
  3. Compare the final idle, download-loaded and upload-loaded readings. A field stays Unavailable if there were no usable measurements.
  4. Repeat under similar conditions. If possible, compare Wi-Fi with an Ethernet connection before changing router settings.

What this bufferbloat test measures

Bufferbloat describes delay associated with excessive queuing when a connection is busy. A large download or upload can fill a network bottleneck while a small interactive request waits. That helps explain why a call or game can feel slow even when a connection has plenty of Mbps.

This page compares successful HTTP request timings to Cloudflare before and during this browser’s download and upload traffic. It uses the same adaptive capacity measurement core as our internet speed test. Download and upload are measured separately, and completed throughput requires sustained, verified transfer evidence. This is an HTTP connection snapshot, rather than an ICMP ping, router queue inspection or direct measurement of a game server.

How to read added delay

Download added delay equals the download-loaded median minus the idle median. Upload added delay uses the upload-loaded median instead. A larger repeatable increase means that these HTTP probes took longer during that traffic stage. It does not identify which router, Wi-Fi hop, ISP link or remote endpoint caused the extra time.

A negative change can occur because the idle and loaded probes were taken at different moments. It is not negative queuing. Check the successful sample counts and repeat the run before treating a small difference as meaningful. This tool deliberately reports the measured values instead of applying an unvalidated A–F score.

What to try when latency rises under load

First compare a quiet run with the behavior you notice during a real upload, backup or download. An Ethernet comparison can help separate a wireless problem from the wider connection. Keep the same device, VPN setting and time window so that you change one condition at a time.

If the increase is repeatable, your router may offer smart queue management or active queue management. Read its documentation before changing settings, record the original configuration, then compare another run. The IETF’s FQ-CoDel description explains one approach to controlling queue delay; support and configuration vary by router. More Mbps alone does not establish whether queue management or Wi-Fi contention is the actual problem.

Limits and useful next checks

These measurements include browser, HTTP and endpoint behavior. The chosen route to Cloudflare can differ from a meeting service or game, and other devices may add traffic while the test runs. Loaded latency may be unavailable even when a complete bandwidth result is available.

For a meeting problem, use our video-call capacity comparison and the meeting application’s own statistics. For uneven performance around your home, use the room-by-room Wi-Fi comparison. See how we test for the difference between measured bandwidth, HTTP latency and game-server telemetry.

Frequently asked questions

Why is my download speed fast but calls still lag?

Download throughput measures data volume per second. Calls also depend on latency, variation, packet delivery, the device and the service route. A repeatable rise in loaded HTTP latency is useful context, but it does not diagnose every call problem.

Does this tool measure packet loss or give a bufferbloat grade?

No. It shows final HTTP latency readings and their signed differences. It does not measure UDP packet loss, inspect queues or award a validated bufferbloat grade.

Why did a loaded latency field stay unavailable?

The bandwidth run may have finished without a usable loaded-latency sample for that stage. Missing values are kept unavailable rather than replaced by idle latency or a made-up result.

Primary sources and further reading

Explore all connection tools · Read our test methodology · Run the main speed test