A properly configured VPN on a good provider costs you 5 to 10 percent of your connection speed. If you’re losing half, something specific is wrong, and it’s almost always one of a dozen specific things.
This list is ranked by hit rate. Most people are fixed by step three; nobody should need all twelve.
First, get a real measurement
Thirty seconds of diagnosis beats an hour of guessing. Run a speed test three times: bare connection, VPN on your nearest server, VPN on the server you actually use. Now you know whether the VPN is even the problem (if the bare number is bad, call your ISP, not us), how much the tunnel itself costs, and how much the distance adds. Keep the numbers; every fix below is measured against them.
The big three (most cases end here)
1. Switch servers. The single highest-yield fix. Individual servers get crowded, and the “fastest” auto-pick isn’t always right. Try two or three alternatives in the same city or country and remeasure. If a neighboring server doubles your speed, you were just on a busy box; favorite the good one.
2. Switch protocols to WireGuard-class. If your app is on OpenVPN (or worse, a legacy protocol from an old tutorial), moving to WireGuard, NordLynx or Lightway routinely multiplies throughput, especially on routers, TVs and older machines where encryption overhead bites hardest. This is the fix with the biggest single jumps; our WireGuard vs OpenVPN guide explains why.
3. Connect closer. Physics charges by distance. A server in your country, or the country next door, will nearly always outrun one across an ocean. If you need a faraway country for content, that’s fine, but compare against a near server to see how much is geography versus a fixable problem.
The configuration tier
4. Check what else the tunnel is carrying. Cloud backups, game updates and seeding quietly saturate connections, and the VPN takes the blame. Pause the heavy apps and remeasure; if speed returns, the tunnel was never the bottleneck.
5. Wi-Fi before blaming the VPN. Test once on Ethernet or standing next to the router. Marginal Wi-Fi plus VPN overhead crosses thresholds that either alone wouldn’t; the cure is the radio, not the subscription.
6. Try UDP vs TCP. If you’re on OpenVPN TCP (common after using stealth modes), you’re paying double-acknowledgment overhead. Switch back to UDP or WireGuard when you’re off the hostile network that needed TCP 443.
7. Turn off what you don’t need. Double VPN, Tor-over-VPN and obfuscation modes all trade speed for properties you may not currently need, as our double VPN explainer quantifies. Streaming through Secure Core is a choice; make it on purpose.
8. Split-tunnel the speed-critical app. If one application needs every megabit (a console download, a video call) and doesn’t need privacy, route it outside the tunnel deliberately, per our split tunneling guide. The right answer to “the VPN slows this one thing” is sometimes “then exempt that one thing.”
The environment tier
9. The router is the bottleneck. If the VPN runs on your router, its CPU sets the ceiling: old hardware doing OpenVPN caps whole homes at a fraction of line speed. Switch the router config to WireGuard, or move the VPN to the devices, and watch the ceiling lift. This is the classic whole-house slowdown from our router guide.
10. Your ISP throttles VPN traffic, or everything, at peak. Compare speeds at 8am versus 9pm. Evening-only slowdowns are congestion or traffic management; ironically, on ISPs that throttle classified video traffic, the VPN often makes things faster, and on ISPs that deprioritize detected VPN protocols, obfuscation can recover the loss.
11. Antivirus or firewall is inspecting the tunnel. Security suites that filter network traffic sometimes double-process VPN packets. Temporarily disable the network-filtering module, remeasure, and if it was the culprit, add the VPN app to its exclusions rather than living without either.
12. The app is stale or the install is haunted. Last and least frequent: update the app (providers ship performance fixes constantly), or reinstall cleanly if settings have accumulated years of experiments. The fresh-install fix is rare but real, which is why it’s ranked twelfth and not first.
If you’ve done all twelve
At this point the math is the message: your provider’s infrastructure is the ceiling. Speed is bought with server capacity, and the gap between a 5/5 speed provider and the budget tier shows up exactly here, in the evenings, on popular locations. Our comparison’s speed column exists for this moment, and the top of it (NordVPN among the leaders) is the upgrade path with a 30-day window to verify on your own line. Test NordVPN here.
One last calibration so you fix the right problem next time: expected VPN cost is 5-10% on nearby servers with WireGuard, 15-30% across oceans, more under obfuscation. Anything inside those bands isn’t broken; it’s physics doing its modest billing. Anything far outside them is on this list.
Special cases worth their own line
Three situations bend the standard list. Obfuscated and stealth modes: expect a real penalty by design (the wrapping costs cycles and padding), so judge those connections against the bands for obfuscation, not against naked WireGuard; if you’re on stealth without needing it, that’s fix 7 in disguise. Double VPN and Secure Core: the multi-hop guide quantifies the haircut; again, a mode for moments. And mobile data: phone VPN speeds ride carrier weather (CGNAT, congested towers), so test the same server on Wi-Fi before concluding anything about the provider.
One measurement habit completes the toolkit: when you find a configuration that hits your line’s potential, screenshot the speed test with the server and protocol visible. Future slowdowns get compared against your own baseline instead of against hope, which converts every future diagnosis into a two-minute comparison.
(And if this page fixed it, adopt the two-minute prevention: favorite the fast servers you found, leave the protocol on the WireGuard-class option, and re-run the three-test baseline after major app or OS updates. Slowness caught at the baseline stage never becomes an evening.)
Speed, more than any other VPN property, rewards measurement over mood; the numbers end debates the adjectives start.
One closing reassurance for the band of readers who arrived genuinely worried: slow is the most common, least serious and most fixable VPN complaint there is. The tunnel that crawls today almost always flies by tonight, three steps from now.
Routers deserve the last word since they hide the worst cases: a whole-house slowdown that survives every app-level fix is almost always the router’s CPU pleading for WireGuard, and the firmware switch is an evening that feels like an upgrade purchase.
Want to compare all VPNs side by side? Check our full VPN comparison table with scores across 18 criteria.
Slow VPNs are almost never mysteries: a busy server, a legacy protocol or simple distance explains the bulk of cases, and the measured three-test diagnosis up top turns vague frustration into a five-minute fix. Walk the list in order, respect the physics bands, and reserve the provider upgrade for when the data says the infrastructure is the ceiling. Speed problems are the most fixable complaint in VPN; this page is the order of operations.