Bufferbloat & Latency Reference 2026
An unloaded ping under 30 ms is excellent; under 30 ms of added latency under load (bufferbloat) earns an A; jitter under 5 ms and 0% packet loss are ideal. Those four numbers, not your megabits, are what actually decide whether calls, games, and browsing feel fast. A 300 Mbps connection with 150 ms of bufferbloat feels worse than a 50 Mbps one with 10 ms. Here is the full threshold table for judging a connection in 2026.
What is a good ping, jitter, and bufferbloat?
Most speed tests give you one big download number and stop. But the metrics that decide how a connection feels are latency-based. Below is how Y2KDASH grades each one, using thresholds consistent with the bufferbloat.net project and Waveform's bufferbloat methodology.
| Metric | Excellent | Good | Acceptable | Poor |
|---|---|---|---|---|
| Unloaded ping | < 30 ms | 30–60 ms | 60–100 ms | > 100 ms |
| Bufferbloat (added latency under load) | < 30 ms | 30–60 ms | 60–100 ms | > 100 ms |
| Jitter | < 5 ms | 5–15 ms | 15–30 ms | > 30 ms |
| Packet loss | 0% | < 0.5% | 0.5–1% | > 1% |
Why does bufferbloat matter more than bandwidth?
Bufferbloat is the latency your connection adds when a buffer somewhere (usually your router or modem) fills up under load. Measured as loaded ping minus unloaded ping, it is the single best predictor of whether a video call breaks up or a game lags — and it is completely invisible to a one-shot download test. This is why a connection can advertise 500 Mbps and still feel broken: the bandwidth is fine, the responsiveness under load is not.
How do I fix bad bufferbloat?
Bufferbloat is fixable in software, for free. Enable Smart Queue Management (SQM) with the CAKE or FQ-Codel algorithm on your router — OpenWrt, and the stock firmware on many modern routers, support it. Set your SQM download/upload rates to about 85-90% of your measured line speed so the queue lives in your router (where CAKE can manage it) rather than in your ISP's oversized buffer. Most connections drop from triple-digit bufferbloat into single digits after this one change.
Why does my fast connection still feel slow?
Because you are measuring the wrong thing. Bandwidth is capacity; latency under load is responsiveness. Continuous monitoring — sampling the connection while you actually use it — is the only way to catch the intermittent bufferbloat, jitter spikes, and packet-loss bursts that a click-to-run test, running only when the line is quiet, will always miss.
Y2KDASH samples your latency, jitter, bufferbloat, and packet loss continuously against Cloudflare's edge. Leave it open and watch the real story.
> LAUNCH Y2KDASH →