Bufferbloat Test
Find out whether your connection lags when it is busy. The test measures your ping at rest, then again while it saturates your download and upload, and grades how much the latency rises.
Last updated:
Press Start to measure how much your latency rises when your connection is busy. Takes about 25 seconds.
- 1 Idle latency
- 2 Latency during download
- 3 Latency during upload
Idle → download → upload. Close other downloads and streams first for a clean result.
Load is created with Cloudflare's public speed-test servers (speed.cloudflare.com); latency is timed with small HTTPS requests during each phase. Each load phase transfers as much data as your line can carry (about 12 MB per second at 100 Mbps), so avoid running it on a metered connection. Nothing else is sent.
How to use the Bufferbloat Test
- Pause other internet use. Stop downloads, cloud backups and streams on all devices so they do not skew the result.
- Press Start bufferbloat test. The first 5 seconds measure your idle latency.
- Wait for the download and upload phases. The test fills your connection in each direction while it keeps measuring latency.
- Read your grade. It is based on the largest median latency increase in either direction.
- Fix and re-test. If you get C or worse, enable SQM or QoS on your router and run the test again to compare.
How this bufferbloat test works
A normal speed test tells you how much data your line can carry. A bufferbloat test asks a different question: how responsive does the connection stay when it is full? This test runs three phases and sends small latency probes about five times per second throughout:
- Idle (5 s): latency with no load. The median is your baseline, and jitter is the average change between consecutive probes.
- Download load: several parallel downloads from Cloudflare fill your downstream bandwidth while probes keep running.
- Upload load: parallel uploads fill your upstream bandwidth while probes keep running.
Latency increase = median loaded latency − median idle latency. The first second of each load phase is excluded (shown as hollow dots) because the transfers are still ramping up. Probes that time out after 4 seconds count as 4,000 ms, since a lost probe under load usually means a very long queue.
Bufferbloat grades
| Grade | Latency increase under load | Real-world effect |
|---|---|---|
| A+ | under 5 ms | No difference while the line is busy |
| A | 5 to 30 ms | Fine for competitive gaming and video calls during downloads |
| B | 30 to 60 ms | Slight delay under load; most users will not notice |
| C | 60 to 200 ms | Game lag and talk-over in calls when someone uploads or downloads |
| D | 200 to 400 ms | Web pages and calls feel sluggish during transfers |
| F | 400 ms or more | The connection is nearly unusable for real-time apps while busy |
Fixing bufferbloat on your router
- Enable SQM / Smart QoS. Look for SQM, Smart Queue, Adaptive QoS, Dynamic QoS or a gaming mode. Algorithms named fq_codel or CAKE are designed specifically to keep queues short.
- Set limits slightly below your real speed. Start at about 90% of the download and upload speeds measured here. If the grade does not improve, lower them in 5% steps; the queue must form in the router, not in the modem.
- Prioritise upload first. Upload bandwidth is small on most cable, DSL and 4G/5G lines, so cloud backups and video uploads often cause the worst lag.
- Re-test on Ethernet and Wi-Fi. If the cable result is good but Wi-Fi is bad, look at Wi-Fi settings, channel congestion or a mesh system's backhaul.
- Turn off old "bandwidth limiters" that use plain QoS rules if your router supports SQM; they usually do little for latency.
Frequently asked questions
What is bufferbloat?
Bufferbloat is extra delay caused by oversized buffers in your modem, router or the provider's equipment. When your connection is fully used, packets wait in a long queue instead of being dropped or paced, so every other packet, including game updates and voice data, waits behind them. The result is ping that jumps from tens to hundreds of milliseconds whenever someone uploads or downloads something big.
How is the bufferbloat grade calculated?
The test takes the median latency while idle and the median latency during the download and upload phases, ignoring the first second of each load phase while the connection ramps up. The grade uses the larger of the two increases: under 5 ms is A+, under 30 ms A, under 60 ms B, under 200 ms C, under 400 ms D, and 400 ms or more F.
How do I fix bufferbloat?
The most effective fix is Smart Queue Management (SQM) on your router, using an algorithm such as fq_codel or CAKE. Many routers call it SQM, Smart QoS, Adaptive QoS or a gaming mode. Set the bandwidth limits to about 85 to 95% of your measured speeds so the queue builds in your router, where it can be managed, instead of in the modem. Upload bufferbloat is usually the worst on cable and DSL lines.
Why is my speed lower than in a normal speed test?
This test focuses on latency, so it uses a moderate number of parallel streams and short phases. On very fast connections the measured speed may be lower than a dedicated speed test. Choose 10 parallel streams for fibre lines; the latency grade matters more than the exact Mbps.
Does Wi-Fi affect the bufferbloat result?
Yes. Wi-Fi adds its own queues and retransmissions, so the result shows the combined effect of your Wi-Fi and your internet line. Test on an Ethernet cable to see the line alone, then on Wi-Fi to see how much the wireless link adds.
How much data does the bufferbloat test use?
Each load phase uses your full connection speed for the selected time. At 100 Mbps, a 10 second download phase moves about 125 MB, and the upload phase moves as much as your upload speed allows. Avoid running it on capped mobile data plans.
Which servers does the test use?
Load is generated with Cloudflare's public speed-test servers at speed.cloudflare.com, and latency is measured with tiny HTTPS requests to Cloudflare's nearest location. No results or personal data are sent anywhere by this page.
Are all the tools on FreeToolCabin tested for quality and accuracy?
Yes! Every tool on our site undergoes automated Q/A testing in real browser environments to ensure high precision, correct output, and zero errors. If you happen to spot a bug or have a suggestion, please contact us at [email protected] and our team will review it right away.