Bufferbloat — excessive buffering in network equipment that dramatically increases latency without reducing packet loss — is the most common cause of the phenomenon where a connection that tests at 100 Mbps still feels laggy during video calls and gaming, and it's almost entirely invisible to standard ping tests
The previous articles on this site covered ping latency basics, the network diagnostic workflow, ICMP blocking, the speed-of-light floor, and what ping can and cannot diagnose. This article addresses bufferbloat — the specific networking problem where large buffers in routers and modems cause latency spikes that standard speed tests and basic pings miss entirely.
What bufferbloat is and why it exists
Network buffers exist in every piece of networking equipment — routers, switches, modems, cable headend equipment. Their purpose: hold packets briefly when a link is temporarily congested, smoothing out brief bursts rather than dropping packets immediately.
The problem: equipment manufacturers, seeking to minimise packet loss complaints, started installing very large buffers. A packet that would have been dropped (triggering TCP's congestion control to slow down) instead waits in a queue that may hold hundreds or thousands of packets.
The latency consequence: large buffers fill up during saturation. Each packet must wait behind all the packets already in the queue. A queue of 500 packets on a 10 Mbps link (each packet ~1500 bytes) represents:
500 × 1500 bytes × 8 bits = 6,000,000 bits ÷ 10,000,000 bits/second = 600 milliseconds of queuing delay
A 600ms ping when the link is saturated — while the same link shows 10ms ping when idle. This is bufferbloat.
Why standard tests miss bufferbloat
A standard ping test (pinging Google.com, 4 packets, idle connection) shows the idle latency — typically excellent. It doesn't reveal what happens when the connection is under load.
A speed test measures throughput, not latency under load. A bufferbloated connection may achieve full rated speed while adding 500ms of latency to every packet during the download — the speed test celebrates a fast connection; the user complains that everything else stops working while downloading.
The Waveform Bufferbloat Test (waveform.com/tools/bufferbloat) specifically measures latency while simultaneously saturating the connection — revealing the difference between idle latency and loaded latency. A connection with good queue management (Active Queue Management, AQM) shows similar latency under load and at idle. A bufferbloated connection shows idle latency of 10ms and loaded latency of 300-800ms.
Active Queue Management: the solution
AQM (Active Queue Management) algorithms drop or mark packets proactively before the buffer fills, preventing the queue from growing to latency-inducing sizes:
CoDel (Controlled Delay): drops packets when the queuing delay exceeds a target (typically 5ms) for a sustained period. Keeps latency low by preventing buffer buildup rather than waiting until the buffer is full.
FQ-CoDel (Fair Queuing + CoDel): adds per-flow queuing — each TCP or UDP flow gets its own small queue. Short-latency traffic (DNS, interactive keystrokes) is prioritised over bulk transfers (large downloads). This is why FQ-CoDel dramatically improves the subjective experience of using a connection during large downloads.
CAKE (Common Applications Kept Enhanced): a more recent AQM algorithm combining FQ-CoDel-like fairness with integrated traffic shaping. Often considered superior to FQ-CoDel for home routers.
Where to find AQM: OpenWrt (open-source router firmware), pfSense, and newer consumer routers from some manufacturers (Ubiquiti, select ASUS models with SQM support) implement CoDel or CAKE. Most ISP-supplied routers do not.
How ping reveals bufferbloat (when used correctly)
To detect bufferbloat with ping:
- Start a continuous ping to a nearby server:
ping -t google.com(Windows) orping google.com(Linux/Mac) - Start a large file download that saturates your connection
- Observe the ping responses during the download
Without bufferbloat: ping latency during download is similar to idle latency (perhaps 2-3× higher — 10ms → 20-30ms)
With bufferbloat: ping latency during download spikes dramatically (10ms → 300-800ms or more) and becomes extremely variable
The pattern to look for: the standard deviation of ping times under load is a better indicator than the peak — highly variable latency (jitter) indicates severe bufferbloat even if the peak isn't extreme.
ISP-level bufferbloat and DOCSIS/cable modems
Cable internet (DOCSIS) has historically been particularly prone to bufferbloat because the upstream direction (upload) has a more limited bandwidth and larger buffers in the cable modem and cable headend:
Upload-direction bufferbloat: a large upload (cloud backup, video upload) saturates the upload link. The cable modem's buffer fills. Every packet in every direction — including acknowledgement packets for your incoming downloads — must wait behind the upload queue. Both upload and download slow dramatically and latency spikes.
DOCSIS 3.1 and Low Latency DOCSIS: newer DOCSIS specifications include AQM and latency-reduction provisions specifically to address cable modem bufferbloat. The improvement requires the ISP to deploy updated headend equipment and enable these features.
How to use the Ping tool on sadiqbd.com
- Idle baseline: run the ping tool on a completely idle connection — record the latency as your baseline
- Loaded test: start a large download (a Linux ISO, a Steam game update) and run the ping tool again during the download — compare against your baseline
- Bufferbloat indicator: if the loaded ping is 3-5× higher than idle ping, you have moderate bufferbloat; if it's 10-50× higher, you have severe bufferbloat that likely explains your connection feeling slow despite good speed test results
Frequently Asked Questions
If I have bufferbloat, can I fix it without replacing my router? Sometimes — depending on the cause. If your router firmware has SQM (Smart Queue Management) settings, enabling CoDel or CAKE with the correct rate limit can dramatically reduce bufferbloat. If your ISP-supplied router doesn't support AQM, adding a secondary router behind it with AQM can help (though the upstream cable modem's buffers may still contribute). For DOCSIS cable connections, flashing a DOCSIS modem that supports AQM or requesting your ISP enable Low Latency DOCSIS addresses the modem-level bufferbloat. The most effective fix is a router running OpenWrt or pfSense with properly configured CAKE or FQ-CoDel — this eliminates the router-level bufferbloat regardless of what ISP equipment is upstream.
Is the Ping tool free? Yes — completely free, no sign-up required.
Try the Ping tool free at sadiqbd.com — measure round-trip latency to any host instantly.
