VPN Speed Testing Methodology

Test setup

Tests run on a Windows PC with an AMD Ryzen 7 9800X3D processor and 32 GB of DDR5-6000 memory. The wired AT&T fiber connection is rated at 1 Gbps in both directions and delivers approximately 950 Mbps download and upload without a VPN.

I normally select a VPN server in the San Francisco Bay Area and use the site’s test server in Seattle. If a free plan does not offer that VPN location, I use an available server and note the exception on the VPN’s page.

I prefer WireGuard or a comparable provider protocol. If a different protocol is used, it is identified on the VPN’s page. Each test uses four parallel connections and measures download and upload separately for 10 seconds each.

Schedule and daily results

Automated tests run at different times of day, with up to five runs per VPN plan. All completed test results are recorded.

Download and upload medians are calculated separately for each UTC calendar day. Rankings use the latest completed day’s median download speed, with paid and free plans ranked separately. Charts show 30 days of daily medians; days without data remain empty.

Speeds are rounded for display only. Calculations use the stored values before display rounding.

Independent testing

I personally purchase paid subscriptions through normal public sales channels. I do not use special testing accounts supplied by VPN providers. Providers cannot purchase ranking positions.

Interpreting the results

Results reflect the tested conditions. Your location, connection, device, VPN server, protocol and network conditions can change the speeds you receive. The connection’s practical limit of approximately 950 Mbps also limits what I can measure; results near that level do not establish a VPN’s maximum throughput.

The ranking compares speed, not overall VPN quality. Privacy, security, reliability, server coverage, features and price also matter when choosing a VPN.