What the numbers say about the network
Analysis of connection speed, latency and stability around the world, drawn from tests that real people ran themselves.
Average speed rankings
Illustrative figures for this build; the real product refreshes them monthly.
| # | Country | Download (Mbps) | Upload (Mbps) | Ping (ms) |
|---|---|---|---|---|
| 1 | Singapore | 324.6 | 297.1 | 4 |
| 2 | Hong Kong | 305.2 | 268.4 | 5 |
| 3 | Chile | 291.8 | 252.0 | 7 |
| 4 | United Arab Emirates | 287.3 | 143.6 | 6 |
| 5 | Thailand | 247.9 | 198.2 | 8 |
| 6 | Vietnam | 159.4 | 138.7 | 9 |
| # | Country | Download (Mbps) | Upload (Mbps) | Ping (ms) |
|---|---|---|---|---|
| 1 | United Arab Emirates | 412.7 | 32.1 | 17 |
| 2 | Qatar | 358.9 | 28.6 | 19 |
| 3 | Kuwait | 263.4 | 25.9 | 21 |
| 4 | South Korea | 148.2 | 18.4 | 24 |
| 5 | Denmark | 142.6 | 21.0 | 22 |
| 6 | Vietnam | 64.8 | 17.3 | 27 |
Latest reading
Why your measured speed is lower than the plan you pay for
How Wi-Fi, your end device and the number of parallel streams all shape the final number.
Jitter matters more than you think on video calls
Low average latency can still feel terrible when the variation between packets is large.
Submarine cables and those slow-internet days
What happens to international capacity when a cable breaks, and how long recovery takes.
How standalone 5G changes latency
Latency compared across 4G, non-standalone 5G and standalone 5G in the same test area.
Measuring gigabit links properly
Why a single stream cannot fill a fast link, and how many streams is the right number.
Home networks: tune yours before upgrading your plan
Router placement, frequency band and ageing devices are usually the real bottleneck.
Market reports
Providers compared country by country, with detail down to individual provinces.
Vietnam
Fixed and mobile speeds by quarter, broken down by major provider.
Read the report →Southeast Asia
The regional picture: the gap between cities and rural areas is narrowing.
Read the report →5G trends
How average speeds have shifted two years into commercial rollout.
Read the report →Video quality
The share of connections fast enough for 4K streaming, by region.
Read the report →How the numbers are produced
Every test is a real session between a user's device and the nearest server. Anomalous sessions are discarded before anything is aggregated.
The server with the lowest latency among the nearby candidates wins, not the one that is geographically closest — because the path a packet takes is what actually determines delay.
A single TCP connection rarely fills a fast link, because of window size and latency. Several streams running at once give a much truer picture of what the connection can do.
No. The opening moments of each phase are discarded so the result is not dragged down by TCP still accelerating, which means the number reflects the link at steady state.
Published figures are grouped by region and by provider. This demonstration build keeps your history in your own browser and sends it nowhere.