What Is Bufferbloat? The Hidden Cause of Lag

You pay for fast internet. Your speed test says you have fast internet. But the moment someone in your house starts downloading a game update, your entire network turns into molasses. Web pages crawl. Video calls stutter. Online games become unplayable. You restart the router. You call your ISP. They tell you everything looks fine on their end. Because technically, it is fine -- your speed is exactly what you are paying for. The problem is something your ISP will never test for, and something most speed tests completely ignore. The problem is bufferbloat.

Bufferbloat is arguably the most widespread and least understood problem in consumer networking. It affects hundreds of millions of internet connections worldwide, and the irony is painful: it was caused by well-intentioned engineers trying to make the internet work better. In this guide, we will explain exactly what bufferbloat is, why it exists, how it wrecks your daily internet experience, and most importantly, how to test for it and fix it.

What Is Bufferbloat? The Simple Explanation

Bufferbloat is excessive latency caused by oversized network buffers. That is the one-sentence definition. Now let us actually explain what that means in human terms.

Every piece of networking equipment in your home -- your router, your modem, your ISP's gear -- has a buffer. A buffer is just a small area of memory where data packets wait in line before being sent to the next hop on their journey across the internet. Think of it like a waiting room at a doctor's office.

In a well-designed network, the buffer is small. Packets arrive, wait a fraction of a millisecond, and get sent on their way. Your base latency might be 15 milliseconds, and even when the network is busy, it stays around 15-20 milliseconds. The waiting room has three chairs, and people move through quickly.

But here is the problem: most consumer routers ship with enormously oversized buffers. Instead of a waiting room with 3 chairs, they have a waiting room with 500 chairs. When your connection is idle, this does not matter -- packets still breeze through. But the moment your connection gets saturated (a large download, a system update, someone streaming 4K, a backup running), the buffer fills up completely. New packets enter the back of a very, very long line.

The result? Your latency skyrockets from 15 milliseconds to 200, 500, or even 1000+ milliseconds. Every packet -- whether it is a video call frame, a game command, or a tiny DNS lookup -- has to wait behind hundreds of queued packets. Your internet is not slow. It is delayed. And that delay makes everything feel broken.

The Highway Analogy: Why Bigger Buffers Are Not Better

The best way to understand bufferbloat is with a highway analogy. Imagine a two-lane highway (your internet connection) with a merge ramp where cars (data packets) enter the highway from a side road.

In a well-designed system, the merge ramp is short. If the highway is full, cars on the ramp see the congestion immediately and can adjust -- slow down, take an alternate route, or wait a brief moment for a gap. The feedback is fast. The system self-regulates.

Now imagine someone "improves" the system by extending the merge ramp to be 5 miles long. Cars entering the ramp cannot see the highway at all. They drive onto the ramp at full speed, not knowing that there is a massive traffic jam ahead. By the time they reach the merge point, they have been sitting on the ramp for 20 minutes. Even worse, the cars behind them keep piling up on the 5-mile ramp, completely unaware of the delay.

That absurdly long merge ramp is your router's oversized buffer. The cars are your data packets. The 20-minute wait is your 500ms latency. The drivers who cannot see the traffic jam are the applications on your devices, which keep sending data at full speed because they do not know the buffer is full. The entire system becomes inefficient because the feedback loop is broken.

How We Got Here: The Accidental History of Bufferbloat

Bufferbloat was not always a problem. In the early days of the internet, memory was expensive. Network equipment had tiny buffers because RAM cost real money. If a link got congested, packets would be dropped almost immediately, and TCP's congestion control algorithms would detect the loss and slow down. The system worked as designed.

Then memory got cheap. Really cheap. And networking engineers, observing that packet loss caused retransmissions and reduced throughput, made what seemed like an obvious improvement: make the buffers bigger. If packets are getting dropped during traffic spikes, just add more buffer space so they can wait in line instead. Problem solved, right?

Wrong. The problem was that TCP's congestion control relies on packet loss (or increased delay, in newer variants) as a signal to slow down. With enormous buffers, packets almost never get dropped. TCP never gets the signal to back off. Applications keep sending data at full speed, the buffer fills up to capacity, and every single packet now experiences hundreds of milliseconds of queueing delay. The buffers absorbed the symptom (packet loss) while creating a far worse disease (latency).

The problem went largely unrecognized until 2011, when Jim Gettys, a Bell Labs researcher and former editor of the X Window System, published a series of blog posts and a paper titled "Bufferbloat: Dark Buffers in the Internet." Gettys had been investigating why his home internet connection felt so terrible despite having respectable speeds. What he found was that oversized buffers were endemic throughout the internet -- not just in consumer routers, but in cable modems, DSLAMs, switches, and even operating system network stacks.

Gettys sounded the alarm, and a community of researchers and developers rallied around the problem. The result was a wave of new queue management algorithms designed to keep buffers small and latency low: CoDel (Controlled Delay) in 2012, fq_codel (Fair Queuing with CoDel) shortly after, and CAKE (Common Applications Kept Enhanced) in 2018. These algorithms are the cure for bufferbloat, and they are freely available in open-source router firmware. But fifteen years later, most consumer routers still ship with the same oversized, unmanaged buffers that caused the problem in the first place.

How Bufferbloat Destroys Your Everyday Internet

Bufferbloat does not just affect gamers and power users. It hits everyone in a household, often without anyone realizing what is actually going on. Here are the most common scenarios where bufferbloat makes your internet feel broken:

The Download Dilemma

Someone starts downloading a large file -- a game update, a system update, a movie for offline viewing. Within seconds, everyone else in the house notices their internet has become sluggish. Web pages take 5 seconds to load instead of 1 second. You assume the download is "using all the bandwidth," but that is only partially true. The real problem is that the download has filled the buffer, and now every other packet has to wait in a 200-500ms queue. Your DNS lookups, your web page requests, your video call packets -- they are all stuck behind the download traffic.

The Video Call Collapse

You are on a Zoom call when someone else starts streaming Netflix. Your video freezes, your audio cuts out, and your colleagues ask if you are still there. This is classic bufferbloat. The streaming traffic saturates the upload or download link, filling the buffer, and your real-time video call packets get trapped behind hundreds of milliseconds of queued data. For a deeper dive into how this specifically affects video calls, check out our guide on bufferbloat and Zoom calls.

The Gaming Nightmare

You are playing an online game and your ping is a steady 25ms. Then your cloud backup starts syncing, or someone opens Instagram and starts loading images. Your ping jumps to 250ms. Your character teleports. Your shots do not register. You rage-quit and blame the game's servers, but the problem is entirely local. Your router's bloated buffer is holding your game packets hostage.

The Smart Home Slowdown

Your smart home devices -- thermostats, cameras, light switches -- all need to communicate with their cloud servers. With bufferbloat, even small command packets get stuck in the queue during heavy usage. Your smart lights take 3 seconds to respond instead of instantly turning on. Your security camera's live feed lags. Your voice assistant takes forever to respond because its tiny request packet is stuck behind megabytes of buffered download traffic.

Bufferbloat Grades Explained: A Through F

When you test your connection on pong.com, you receive a bufferbloat grade from A+ to F. This grade is based on how much your latency increases when your connection is under maximum load. Here is what each grade means and what it tells you about your connection's real-world performance:

A surprising number of connections that show excellent speed test results will score a D or F on bufferbloat. This is the great disconnect in internet quality: speed and quality are not the same thing. A 50 Mbps connection with an A grade will provide a dramatically better daily experience than a 500 Mbps connection with an F grade for any household with more than one person or device.

How to Test for Bufferbloat

Testing for bufferbloat requires a fundamentally different approach than a standard speed test. A regular speed test measures your throughput on an idle connection. A bufferbloat test measures your latency while the connection is fully loaded. This is the critical distinction.

Here is how pong.com tests for bufferbloat:

1. Step 1: Measure idle latency. We send a series of pings to our test servers to establish your baseline latency when the connection is not under load. This is your "best case" latency.
2. Step 2: Saturate the connection. We initiate multiple parallel download streams to push your connection to its maximum throughput, filling up any buffers in the path.
3. Step 3: Measure loaded latency. While the download is running at full speed, we continue sending ping measurements. The difference between your idle latency and your loaded latency is the bufferbloat measurement.
4. Step 4: Repeat for upload. We do the same thing in the upload direction, since bufferbloat can affect uploads and downloads independently.
5. Step 5: Calculate and grade. We compute the latency increase and assign a letter grade based on the scale above.

The entire test takes about 30 seconds. You can run it for free at pong.com, no account required. We recommend testing over Ethernet for the most accurate results, since Wi-Fi introduces its own latency variables that can confuse the picture.

What Causes Bufferbloat? The Usual Suspects

Bufferbloat can exist at multiple points in your network path. Understanding where the bloat lives is essential for fixing it. Here are the most common culprits:

1. Your Router

The number one source of bufferbloat in most homes is the router itself. Consumer routers from major brands -- Netgear, TP-Link, Asus, Linksys -- typically ship with large, unmanaged buffers and no Smart Queue Management enabled by default. The router firmware uses simple FIFO (First In, First Out) queuing, which means it processes packets in the order they arrive without any intelligence about priority or fairness. When the outgoing link is saturated, the FIFO buffer fills up and every packet pays the price.

2. Your Cable Modem or ONT

Cable modems (for DOCSIS connections) and ONTs (for fiber) also contain buffers. DOCSIS cable modems in particular have been historically bad about bufferbloat, partly because the DOCSIS protocol itself introduces queueing at multiple levels. Some newer DOCSIS 3.1 modems have improved buffer management, but older models are notorious offenders.

3. ISP Equipment

Your ISP's network equipment -- the CMTS for cable, the BNG for DSL, or the OLT for fiber -- also has buffers. ISP-side bufferbloat is harder to diagnose and impossible for you to fix directly. However, some ISPs have started deploying better queue management on their equipment. If your bufferbloat persists even after fixing your router, ISP-side buffers may be the remaining cause.

4. Your Operating System

Even your computer's operating system has network buffers. Linux, macOS, and Windows all have configurable transmit queue lengths and TCP buffer sizes. While OS-level bufferbloat is less common than router-level bufferbloat, it can contribute to the overall problem, especially on high-speed connections where the OS may buffer large amounts of data before sending.

How to Fix Bufferbloat: SQM, fq_codel, and CAKE

The fix for bufferbloat is surprisingly elegant: instead of blindly queuing packets in a giant FIFO buffer, use an intelligent queue management algorithm that keeps the buffer small and prioritizes packets fairly. These algorithms are collectively called Smart Queue Management (SQM), and the three you need to know about are CoDel, fq_codel, and CAKE.

CoDel (Controlled Delay)

Developed by Kathleen Nichols and Van Jacobson in 2012, CoDel monitors the time packets spend in the queue. If packets are consistently sitting in the buffer for longer than 5 milliseconds (the "target" delay), CoDel starts dropping packets to signal to TCP that the link is congested. This restores the feedback loop that oversized buffers break. CoDel is elegant because it adapts automatically -- no manual configuration needed.

fq_codel (Fair Queuing + CoDel)

fq_codel combines CoDel's delay management with fair queuing, which creates separate virtual queues for each network flow (roughly, each device-to-server connection). This means a massive download gets its own queue, your video call gets its own queue, and your web browsing gets its own queue. A single heavy flow cannot starve the others. fq_codel is the default queue discipline in Linux since 2012 and is the most widely deployed SQM algorithm.

CAKE (Common Applications Kept Enhanced)

CAKE is the newest and most sophisticated option, developed by the bufferbloat.net community. It includes everything fq_codel does, plus per-host fairness (so one device cannot hog bandwidth even across multiple flows), better handling of variable-rate links (common on DSL and LTE), and built-in traffic shaping. If your router supports CAKE, it is the best option available. It is available in OpenWrt and some commercial router firmware.

The improvement from enabling SQM is often described as transformative. Users routinely report that it feels like getting a completely new internet connection. And the best part? These algorithms are free, open-source, and available right now on many routers.

Fixing Bufferbloat on Your Router: A Quick Overview

The specific steps to fix bufferbloat depend on your router. Here is a quick overview of the options by platform:

The critical configuration step for any SQM setup: you must set your upload and download bandwidth limits to about 85-90% of your actual measured speeds. SQM works by taking control of the queue before your modem or ISP equipment does. If you set the limits higher than your actual speed, the bottleneck moves to a buffer you do not control, and SQM cannot help.

For detailed, step-by-step router configuration guides, check out our article on how to fix bufferbloat which covers the most popular routers with screenshots and specific settings.

Which ISPs Have the Worst Bufferbloat?

Not all ISPs are created equal when it comes to bufferbloat. The type of connection technology, the age of the infrastructure, and whether the ISP has deployed modern queue management all play a role.

Cable (DOCSIS) -- High Risk

Cable internet connections using DOCSIS technology are historically the worst offenders for bufferbloat. The DOCSIS protocol introduces multiple layers of buffering, and many cable modems ship with oversized buffers. DOCSIS 3.1 with Low Latency DOCSIS (LLD) improvements helps, but older DOCSIS 3.0 modems and infrastructure remain problematic. Major cable ISPs like Comcast, Spectrum, and Cox often show higher bufferbloat rates in user testing data.

DSL -- Moderate Risk

DSL connections have moderate bufferbloat risk. The slower speeds of DSL mean that buffers fill up faster (there is less throughput to absorb), so any bufferbloat that exists tends to be more noticeable. However, DSL modems are often simpler devices with less aggressive buffering than cable modems.

Fiber (GPON/XGS-PON) -- Lower Risk

Fiber connections generally have lower bufferbloat because the higher speeds (1 Gbps+) mean buffers drain faster, and many fiber ISPs have deployed more modern equipment. However, fiber does not guarantee zero bufferbloat -- the ONT and your router can still introduce it. ISPs like Google Fiber and AT&T Fiber tend to score well, but results vary by region and equipment.

Starlink and Fixed Wireless -- Variable

Satellite and fixed wireless connections face unique bufferbloat challenges due to the high base latency and variable link conditions. Starlink has made significant progress with its network management, but users still report bufferbloat during peak congestion periods. Fixed wireless ISPs vary widely.

For a more detailed breakdown of ISP performance including bufferbloat scores, see our ISP Speed Report.

The Future: Is Bufferbloat Finally Going Away?

The short answer: slowly, yes. The long answer: it is complicated.

On the positive side, awareness of bufferbloat has grown enormously since Jim Gettys first named the problem. Major developments include:

• Linux adopted fq_codel as its default queue discipline in 2012, meaning any Linux-based router (including OpenWrt) gets basic bufferbloat protection out of the box.
• Apple integrated queue management into macOS and iOS, reducing OS-level bufferbloat on Apple devices.
• DOCSIS 3.1 Low Latency DOCSIS (LLD) specification includes mandatory Active Queue Management, which should reduce bufferbloat on future cable modem deployments.
• Wi-Fi 6 introduced BSS Coloring and OFDMA which reduce airtime contention and wireless bufferbloat.
• L4S (Low Latency, Low Loss, Scalable Throughput) is an emerging standard that provides explicit congestion signals, potentially eliminating the need for large buffers entirely.
• More router manufacturers are including SQM-like features in stock firmware, though often under different marketing names ("QoS optimization," "lag reduction," "gaming mode").

On the negative side, the installed base of bufferbloated equipment is enormous. Millions of routers and modems with oversized, unmanaged buffers are still in active use, and ISPs are slow to upgrade their infrastructure. Most consumers do not know what bufferbloat is, so there is limited market pressure on manufacturers to fix it. The problem is solvable, but it will take years of gradual equipment turnover and continued advocacy from the networking community.

Bufferbloat in 60 Seconds: The TL;DR

• What: Bufferbloat is excessive latency caused by oversized network buffers in your router, modem, or ISP equipment.
• Why it exists: Manufacturers made buffers too big to prevent packet loss, accidentally breaking the feedback mechanisms that keep networks responsive.
• How it feels: Your internet works great when idle but becomes laggy and unresponsive the moment anyone in your house does something bandwidth-intensive.
• How to detect it: Run a speed test on pong.com that measures latency under load, not just throughput on an idle connection.
• How to grade it: A/B = good, C = mediocre, D/F = your internet experience is being significantly degraded by bufferbloat.
• How to fix it: Enable SQM (Smart Queue Management) on your router using fq_codel or CAKE algorithms. Set bandwidth limits to 85% of your measured speeds.
• If you cannot fix your router: Consider upgrading to a router that supports SQM, or flash OpenWrt firmware on your existing router.
• The payoff: Fixing bufferbloat often feels like getting a brand-new internet connection. Latency drops from hundreds of milliseconds to single digits under load.

Frequently Asked Questions

Stop Blaming Your Speed. Start Testing Your Bufferbloat.

Bufferbloat is the internet's dirty secret. It hides behind speed test numbers that look perfectly fine. It makes you blame your ISP, your router, your devices -- everything except the actual cause. And it has been silently degrading hundreds of millions of internet connections for over a decade.

The good news is that it is completely fixable. A free test on pong.com takes 30 seconds and will tell you exactly how bad your bufferbloat is. If your grade is C or worse, enabling SQM on your router (or upgrading to a router that supports it) will likely be the single most impactful thing you can do to improve your internet experience. More impactful than upgrading your speed plan. More impactful than buying a mesh Wi-Fi system. More impactful than switching ISPs.

Test your connection. Check your grade. Fix the bloat. Your video calls, your gaming sessions, and everyone else in your household will thank you.

For more on related topics, explore our guides on bufferbloat and Zoom calls, how to fix bufferbloat, Wi-Fi vs Ethernet testing, and how speed tests work.