How we speed-test VPNs (and why most reviews get it wrong)

· 7 min read

Most VPN speed claims come from one test, on one server, at one time of day. That produces numbers that cannot be reproduced. Our method is slower but survives a second look.

The four variables that actually matter

Change any one of these and your result moves more than the choice of provider does.

  • Protocol — WireGuard-based protocols beat OpenVPN by a wide margin
  • Distance — latency rises roughly with physical distance to the server
  • Server load — a busy server at 9pm behaves nothing like the same server at 4am
  • Your own line — measure an unprotected baseline before every single run

Our procedure

We take an unprotected baseline, then run three measurements per server, on three servers per region, at both an off-peak and a peak hour. We report the median as a percentage of baseline rather than raw megabits, because raw numbers only describe our office line.

What good looks like

On a nearby server, a modern VPN should retain 85 to 95 percent of your unprotected download speed. Long-haul connections between continents typically retain 50 to 75 percent. Anything below half on a nearby server points to a protocol or configuration problem, not the provider.

Frequently asked

Will a VPN slow down my internet?

Slightly. Expect a 2 to 10 percent loss on a nearby server with a WireGuard-based protocol, which you will not notice while browsing or streaming.

Which protocol is fastest?

WireGuard and its variants, NordLynx and Lightway. Use OpenVPN only when a network blocks the faster protocols.

Find your match in 30 seconds

Filter 8 tested VPNs by use case, budget and must-have features.

Open the VPN finder →

Keep reading