What is bufferbloat?
Bufferbloat is excessive latency caused by network equipment holding packets in queues for too long. The delay becomes noticeable when an internet connection handles heavy uploads or downloads and other packets have to wait.
A router uses buffers to hold packets before transmission. This buffering helps manage differences between incoming and outgoing traffic. However, an excessively long queue increases the time packets spend waiting. The Bufferbloat.net FAQ explains how this queue buildup creates unwanted delay at a network bottleneck.
Think of a queue at a checkout counter. When several people arrive together, each person waits for the people ahead of them. Network packets experience a similar delay when they build up behind other packets.
Bufferbloat does not mean that your internet connection has a low download speed. It means that the connection experiences excessive delay while network traffic is active.
What does bufferbloat mean in simple terms?
Bufferbloat means that data packets wait too long before travelling through a congested network link. That waiting time increases latency, even when the connection continues transferring data at a high rate.
For example, a computer can download a large game while another device makes a video call. If the download creates a long queue, packets carrying the video call data can wait behind the download traffic. The call then experiences delayed audio or video.
Why can fast internet still experience bufferbloat?
Download speed and latency measure different aspects of an internet connection. Download speed measures how quickly data arrives, while latency measures how long a packet takes to travel to a destination and receive a response.
A fast connection can still have an overloaded queue at its slowest link. The Technical Introduction to Bufferbloat from Bufferbloat.net explains how excessive buffering increases latency and jitter in shared networks.
For a clearer explanation of these measurements, read latency vs bandwidth.
Why does bufferbloat happen?
Bufferbloat happens when packets reach a network bottleneck faster than the link can transmit them. The excess packets wait in a queue, and that waiting time increases latency.
A network bottleneck is the link that limits how quickly traffic can move through a particular path. For example, a router can receive traffic from a fast local network while sending it through a slower internet connection.
The Bufferbloat.net FAQ explains that bufferbloat can occur wherever a fast link feeds into a slower link.
How does a packet queue create extra delay?
A packet queue stores network packets while they wait for transmission. A basic first-in, first-out queue, commonly called FIFO, sends packets in the order they arrive.
The process works like this:
- A device sends packets through the local network.
- The packets reach a link with limited transmission capacity.
- Incoming traffic arrives faster than the link can transmit it.
- The router or network interface places waiting packets in a queue.
- New packets wait behind packets already in the queue.
- The additional waiting time increases latency.
A short queue helps keep a network link busy. The problem is persistent excessive queueing, not the existence of a queue itself.
Where can bufferbloat occur?
Bufferbloat can occur at different points in a network. Two important locations for home internet users are the router's connection to the ISP and a Wi-Fi interface.
The router's internet-facing link can become a bottleneck when local devices send or receive traffic faster than that link can carry it. A Wi-Fi interface can also become a bottleneck when devices generate packets faster than the wireless connection can transmit them.
Can Wi-Fi cause bufferbloat?
Yes. Wi-Fi equipment also uses packet queues, and wireless capacity can become the limiting link. Bufferbloat.net specifically identifies Wi-Fi interfaces as a possible location for excessive queueing delay.
A weak Wi-Fi signal, interference and wireless congestion can also cause performance problems. These issues are not automatically bufferbloat, so compare Wi-Fi results with a wired Ethernet connection when possible.
What problems does bufferbloat cause?
Bufferbloat increases latency during network activity, making applications that depend on timely packet delivery less responsive. Users commonly notice the problem during gaming, video calls, cloud backups and large downloads.
The IETF's RFC 8033, published in February 2017, describes how high latency and latency variation degrade interactive applications, including voice and real-time video services.
Why do games lag when someone downloads or uploads?
Online games exchange small packets to report player actions and receive updates from the game server. A large upload or download can create a queue that delays those packets.
The result can include delayed responses, sudden increases in ping and controls that feel less responsive. A high-speed internet package does not prevent this queueing behaviour by itself.
For gaming-specific connection guidance, visit internet performance for gaming.
Why do video calls freeze during uploads?
Video calls need packets to arrive in a timely manner. A cloud backup, large file upload or live stream can fill the upstream connection and increase the waiting time for other traffic.
That delay can cause conversation interruptions, delayed audio or inconsistent video. A connection with sufficient upload throughput can still experience this problem when its upstream queue becomes excessively long.
Why does browsing become unresponsive when the network is busy?
Browsing involves small requests and responses that benefit from low latency. When those requests wait behind a large transfer, pages and interactive services can feel slow to respond.
Bufferbloat is one possible explanation for this behaviour. DNS problems, wireless interference, packet loss and remote-server delays can also affect browsing.
Can uploads and downloads both cause bufferbloat?
Yes. Uploads and downloads can both create bufferbloat when either direction reaches a congested network bottleneck. The direction that causes the larger latency increase helps identify where to investigate.
A bufferbloat test should therefore compare latency during upload traffic and download traffic instead of treating all network load as one condition.
How can uploads increase ping?
An upload sends data from your device or local network towards the internet. Large file transfers, cloud backups and live streaming can saturate the upstream link.
When packets wait in an upstream queue, other traffic travelling in that direction experiences additional delay. This can affect online games, calls and other interactive applications.
How can downloads increase ping?
A download brings data from the internet into your local network. Large game downloads, operating-system updates and high-volume file transfers can load the downstream path.
When the download creates excessive queueing, other packets can experience additional delay. The effect depends on where the bottleneck and packet queues exist.
| Network activity | Possible latency effect |
|---|---|
| Large upload | Increased latency while sending data |
| Large download | Increased latency while receiving data |
| Simultaneous upload and download | Added latency in one or both directions |
The Waveform Bufferbloat Test also compares idle latency with measurements during active upload and download traffic.
How do you test for bufferbloat?
A bufferbloat test compares baseline latency with latency measured while the connection handles active network traffic. The comparison shows whether uploads or downloads add delay to your connection.
Bufferbloat.net recommends using tests that measure latency both while the line is idle and during upload or download activity. Its testing guidance describes this measurement approach.
What should you do before running a bufferbloat test?
Use consistent conditions when measuring your connection. Avoid unnecessary background transfers and use the same device when comparing repeated results.
A wired Ethernet connection provides a useful comparison with Wi-Fi. Keep the test destination and other conditions consistent where possible.
How do you run a bufferbloat test on this website?
Use the bufferbloat test to compare idle and saturated latency in both traffic directions.
Follow these steps:
- Open the bufferbloat test page.
- Allow the test to establish its baseline latency measurement.
- Review the latency measured during download activity.
- Review the latency measured during upload activity.
- Compare the loaded measurements with the baseline.
- Read the displayed result and identify which direction adds more delay.
The test provides an A–F grade as a summary of its result. Read the accompanying measurements as well, because the letter alone does not show the exact latency increase or identify the physical device responsible.
What does a bufferbloat test measure?
A bufferbloat test measures how latency changes when network traffic is active. It compares the baseline result with results during loaded conditions.
The difference between loaded latency and baseline latency is called added latency. A larger increase means packets experience more delay while the connection is busy.
Advanced testing tools use different methods. For example, Flent's Realtime Response Under Load (RRUL) test measures round-trip time while running eight TCP streams, with four downloads and four uploads. That is an advanced testing method, not a description of the browser test on this website.
What does a bufferbloat test result mean?
A bufferbloat result shows how much response delay increases when your connection handles network traffic. Compare the baseline, upload-loaded and download-loaded measurements before deciding what to investigate.
Do not judge a connection by its download speed or idle ping alone. The additional latency under load is the measurement that helps identify bufferbloat.
What is the difference between idle latency and loaded latency?
Idle latency is the baseline response time measured without the test's intentional traffic load. Loaded latency is the response time measured while upload or download traffic is active.
Added latency is the difference between the loaded measurement and the baseline measurement. This difference helps show how network activity affects responsiveness.
What does a good or bad bufferbloat grade mean?
A grade summarises the result using the grading rules of the specific test provider. Different tests can use different grading scales and thresholds.
On this website, read the A–F grade together with the displayed latency measurements. Do not assume that a letter grade from another provider represents the same latency range.
What does it mean when only upload or download latency increases?
An increase during upload activity points towards an upstream queueing problem. An increase during download activity points towards a downstream queueing problem.
A result in both directions indicates that loading either direction increases response delay. These patterns guide further troubleshooting, but a browser measurement does not identify the exact network hop or physical device responsible.
How can you compare Ethernet and Wi-Fi results?
Run the same test over Wi-Fi and then over Ethernet using comparable conditions. A noticeably worse Wi-Fi result points towards the wireless path as a place to investigate.
An Ethernet comparison does not prove that every wireless problem is bufferbloat. Signal strength, interference and wireless airtime contention also affect Wi-Fi performance.
Use your result to choose the next action
| Result pattern | What it suggests | What to investigate next |
|---|---|---|
| Little added latency in either direction | No substantial load-related delay appeared in this test | Compare again during normal household activity |
| Higher latency during upload | Upstream queueing may be contributing | Cloud backups, uploads and upstream traffic management |
| Higher latency during download | Downstream queueing may be contributing | Large downloads and downstream queue management |
| Higher latency in both directions | Both traffic directions are affected by load | Router queue management and link capacity |
| Wi-Fi result is worse than Ethernet | The wireless path is contributing to the difference | Access point, signal and wireless conditions |
| Latency is high even before loading | The issue is not explained by added load alone | Baseline ping, routing, Wi-Fi or ISP path |
Use these results as diagnostic clues, not proof of a specific faulty device. Repeat the test after changing one condition so you can compare the outcome.
How do you fix or reduce bufferbloat?
Smart Queue Management (SQM) is a practical way to reduce excessive queueing delay on supported routers. Before changing hardware, identify the affected traffic direction and check which traffic-management features your current router supports.
Bufferbloat.net recommends measuring latency under load before selecting a solution. Its practical bufferbloat guide describes checking existing SQM support and adjusting traffic settings while monitoring both latency and throughput.
Can you reduce bufferbloat without buying a new router?
Start by identifying the traffic that creates the latency increase. Pause unnecessary uploads or downloads when possible, compare Ethernet and Wi-Fi, and check the router's existing traffic-management settings.
A router with suitable SQM support can manage queues without requiring a new device. The result depends on the router's features and processing capacity.
Can router settings reduce bufferbloat?
Yes. Suitable queue-management settings can reduce excessive waiting at the bottleneck. SQM combines traffic shaping, queue management and packet scheduling to manage how traffic moves through a constrained link.
OpenWrt explains that SQM includes per-packet or per-flow scheduling, Active Queue Management (AQM), traffic shaping, rate limiting and Quality of Service (QoS) prioritisation. Its SQM documentation also explains that SQM runs on the router CPU.
Basic QoS prioritisation and SQM are not identical. A router's gaming mode or priority setting does not automatically provide the same queue-management behaviour.
When should you consider a different router?
Consider different hardware when your existing router lacks suitable queue-management features or cannot process traffic at your connection's required rate.
Check for documented support for SQM, CAKE or FQ-CoDel. Also check processing capability and compatibility with your ISP connection before purchasing equipment.
How do you check whether the fix worked?
Repeat the bufferbloat test under comparable conditions after changing the router settings. Compare baseline latency, upload-loaded latency and download-loaded latency.
Also check download and upload throughput. A successful configuration should reduce unnecessary latency under load without imposing an unacceptable reduction in usable speed.
What is SQM, and how do CAKE and FQ-CoDel help?
SQM is a system that combines queue management and traffic shaping to control delay at a network bottleneck. CAKE and FQ-CoDel are queue-management options used in systems designed to reduce bufferbloat.
The IETF's RFC 7567, published in July 2015, recommends testing and deployment of Active Queue Management in network devices. Later IETF documents describe specific approaches, including PIE, CoDel and FQ-CoDel.
What is the difference between AQM and SQM?
Active Queue Management (AQM) manages packet queues to control congestion and excessive delay. Smart Queue Management (SQM) combines AQM with other functions, including traffic shaping and scheduling.
Basic QoS prioritisation assigns different treatment to network traffic. SQM addresses queue behaviour and traffic rates as part of a broader system.
What is the difference between CAKE and FQ-CoDel?
CAKE and FQ-CoDel are different queue disciplines. Both are used to manage network traffic, but their designs and available features differ.
| Technology | Main purpose |
|---|---|
| CoDel | Controls excessive persistent queueing delay |
| FQ-CoDel | Combines flow queueing with CoDel's delay management |
| CAKE | Combines traffic shaping, AQM and flow isolation |
| PIE | Uses a control algorithm to manage average queueing latency |
The IETF's RFC 8289, published in January 2018, describes CoDel as a framework for controlling bufferbloat-generated excess delay. RFC 8290, also published in January 2018, specifies FQ-CoDel as a hybrid packet scheduler and AQM algorithm.
The IETF published RFC 8033 in February 2017 to describe PIE, an AQM approach for controlling average queueing latency. The Linux CAKE manual documents CAKE's traffic shaping, AQM, flow isolation and overhead-compensation features.
No queue discipline is universally best for every router and internet connection. Hardware, firmware support, link speed and configuration all matter.
Does SQM have hardware requirements?
SQM uses router processing resources. A router with limited CPU capacity can struggle to process traffic at the full speed of a high-bandwidth connection.
OpenWrt's SQM guidance explains this processing limitation and provides configuration information for the WAN bottleneck. Check the router's documented capabilities before enabling advanced queue management.
What if your internet still lags after fixing bufferbloat?
Bufferbloat is not the only cause of high latency, jitter or an unresponsive connection. If loaded latency improves but the problem remains, investigate baseline latency, Wi-Fi conditions, packet loss and the network route.
Could high idle ping be the problem instead?
High idle latency exists before the test introduces additional network load. It can relate to the distance to the test server, routing, wireless conditions or the ISP path.
Bufferbloat describes additional delay created by queueing under load. A good loaded-latency result does not mean that every other latency problem has been resolved.
Could Wi-Fi interference or a weak signal cause lag?
Yes. Wireless interference, weak signal and airtime contention can affect connection responsiveness. Compare results over Ethernet to check whether the wireless path contributes to the problem.
Use the Wi-Fi speed test to investigate wireless performance.
Could packet loss or routing cause the problem?
Packet loss, routing changes and remote-server performance can cause symptoms that resemble bufferbloat. These issues require separate checks.
A persistent problem after loaded latency improves should not automatically be blamed on bufferbloat. Use the ping test, packet loss test and general connection troubleshooting guide to investigate other causes.
Can fixing bufferbloat have any drawbacks?
SQM can improve responsiveness under load while reducing the maximum throughput achievable on a router with limited processing capacity. The trade-off depends on the router, connection and traffic-shaping configuration.
Traffic shaping deliberately controls the rate at which packets leave a network interface. This gives the queue-management system control over the bottleneck, but the configured rate and router processing capability affect peak throughput.
The OpenWrt SQM documentation warns that slower router devices can struggle to keep up with full internet bandwidth when SQM is enabled.
A lower speed-test result does not automatically mean that SQM failed. Compare throughput with loaded latency to decide whether the configuration provides a useful balance between speed and responsiveness.
Related reading: how internet speed tests work.
Sources and references
- Bufferbloat.net — Bufferbloat FAQs. Supports definition, packet queues, bottlenecks and Wi-Fi queueing.
- Bufferbloat.net — Technical Introduction. Supports how excessive buffers increase latency and jitter.
- Bufferbloat.net — What Can I Do About Bufferbloat?. Supports idle and loaded testing, measuring both directions and practical SQM guidance.
- IETF — RFC 7567 (July 2015). Supports recommendations concerning Active Queue Management.
- IETF — RFC 8033 (February 2017). Supports PIE queue management and control of average queueing latency.
- IETF — RFC 8289 (January 2018). Supports CoDel and control of bufferbloat-generated delay.
- IETF — RFC 8290 (January 2018). Supports FQ-CoDel, flow isolation and packet scheduling.
- OpenWrt — SQM. Supports SQM components and router CPU limitations.
- OpenWrt — SQM Details. Supports bottleneck shaping and link-layer adaptation.
- Linux tc-cake manual. Supports CAKE shaping, AQM, flow isolation and overhead compensation.
- Flent — Supplied Tests. Supports RRUL testing with eight TCP streams and RTT measurements.
- Waveform — Bufferbloat Test. Supports idle, download-loaded and upload-loaded latency comparison.