Bufferbloat Test: Check Ping Under Load & Fix It (2026)
Bufferbloat is latency caused by oversized network buffers: when a download or upload saturates your link, your router queues packets instead of dropping them, so ping jumps from around 20 ms to 200-500 ms even though your Mbps looks fine.
Test it by comparing idle ping with ping under load — a rise of 100 ms or more confirms it. The fix is Smart Queue Management (SQM, using cake or fq_codel) on your router, which usually brings ping under load back below 30 ms at the cost of about 5-10% of raw throughput.
Bufferbloat is when network equipment buffers too many packets, causing latency to balloon during heavy use. Idle ping is fine, but the moment anyone starts a download or video call, ping triples and Zoom freezes. The fix is one router setting.
On this page
- Where bufferbloat actually comes from
- How to run a bufferbloat test (two minutes, no software)
- The fix: Smart Queue Management (SQM)
- A few SQM settings worth getting right
- What a healthy result looks like
- What is bufferbloat in simple terms?
- How do I test for bufferbloat?
- What is a good bufferbloat test result?
- Does Smart Queue Management (SQM) really fix bufferbloat?
- Will SQM slow down my internet?
Routers, modems, and even DSLAMs ship with oversized packet buffers. When the link is saturated, those buffers fill, and every new packet waits in line behind the queue. A 200-packet queue at a 50 Mbps line rate adds about 60 ms of latency. Stacked across modem, router, and ISP equipment, you can see latency jump from 15 ms idle to 300 ms loaded.
A bufferbloat test is simply a comparison of your latency when the line is idle against your latency when the line is saturated. You can do it here:
How to read the difference: under about 30 ms of increase is healthy, 30-100 ms is mild bufferbloat you will notice on calls, and anything above 100 ms — the classic jump from 20 ms to 200 ms and up — is severe bufferbloat. Rising jitter tells the same story: a queue that keeps filling and draining makes latency erratic, not just high.
One honest limitation: this test reports ping and jitter from its own latency phase, so the comparison above is manual rather than automatic. If you want a single-click loaded-latency grade, dedicated bufferbloat tests such as Waveform's measure latency continuously during the transfer.
SQM uses an algorithm called CAKE (or fq_codel) to keep queues short by dropping packets early when they would otherwise pile up. Routers that support it well:
Set the SQM download/upload caps to about 90% of your measured speed . The 10% headroom is what gives the queue room to drain. Setting them too high (above your real link rate) defeats the purpose.
With SQM tuned correctly, loaded ping should be within 5-15 ms of idle ping. Video calls stop freezing, game ping stays flat during a Steam download, and the entire household feels faster, even though the raw Mbps number is slightly lower on paper.
Key points
- Get your idle baseline. Run the speed test with nothing else using the connection and write down the ping and jitter values. These are measured before the transfer phases, so they represent your unloaded latency.
- Saturate the line. Start a large download or a cloud backup upload and leave it running.
- Measure again under load. Run the speed test a second time while that transfer is still going, and compare the new ping and jitter with your baseline.
- Eero Pro 6 / 6E, automatic.
- OpenWrt and DD-WRT routers, manual SQM-CAKE.
- GL.iNet routers, built-in SQM toggle.
- Asus AX/AXE, under "Adaptive QoS".
- Most ISP-supplied routers, usually no support; bridge them and add your own router.
Frequently asked questions
- What is bufferbloat in simple terms?: Bufferbloat is when an oversized network buffer (usually in your router) holds packets in a long queue during heavy use, causing your ping to spike from ~20 ms to 200–500 ms even though raw download speed looks fine.
- How do I test for bufferbloat?: Compare your latency when the line is idle with your latency when it is saturated. Run a speed test with nothing else running and note the ping and jitter, then start a large download or upload and run the test again. An increase under 30 ms is healthy, 30-100 ms is mild bufferbloat, and more than 100 ms is severe. Dedicated tools such as Waveform's bufferbloat test automate the same comparison during the transfer.
- What is a good bufferbloat test result?: Ping under load should stay within roughly 5-30 ms of your idle ping, with jitter staying in single or low double digits. A jump from 20 ms idle to 200 ms or more under load is severe bufferbloat and is why video calls freeze and games lag while someone downloads.
- Does Smart Queue Management (SQM) really fix bufferbloat?: Yes. SQM (cake or fq_codel) caps the queue depth and fairly shares bandwidth across flows. Most users see ping under load drop from 200 ms to under 30 ms after enabling it.
- Will SQM slow down my internet?: It trims raw throughput by roughly 5–10% in exchange for keeping latency low and stable. For gaming, video calls and web browsing the trade-off is overwhelmingly worth it.
Steps
- Test your bufferbloat baseline
- Run a bufferbloat-aware speed test and note your idle ping vs ping under load. A jump of 100 ms+ means bufferbloat.
- Check if your router supports SQM
- Open your router admin. Look for QoS, Smart Queue, fq_codel or cake. OpenWrt, ASUS Merlin, eero, IQrouter and Ubiquiti EdgeRouter all support it natively.
- Set bandwidth limits
- Enter 90% of your real download and upload speeds (from your speed test). The router needs headroom to manage the queue.
- Pick the cake algorithm if available
- Cake is the modern default. If only fq_codel is offered, that's also fine.
- Re-test under load
- Run the same bufferbloat test again. Ping under load should now stay within 20 ms of idle.
More from FastSpeedTest