Why a VPN Slows You Down and How to Get the Speed Back

Encryption is barely involved. On any processor with hardware AES support the overhead is a few percent. The real losses come from four other places — distance to the server, how loaded that server is, a wrong MTU, and shaping by your provider. Here's how to identify which one is yours in about ten minutes, and what to do about each.

What each cause actually costs you

Why a single download won't speed up at high latency

This is the most underrated cause. The throughput of one TCP connection is capped by the window size divided by the round-trip time. With a standard 64 KB window and 100 ms of latency, one stream tops out around 5 Mbit/s no matter how fast your plan is. The arithmetic is simple: 64 KB ÷ 0.1 s ≈ 5.2 Mbit/s.

Modern systems scale the window, so the real ceiling is higher, but the relationship holds: more latency, slower ramp-up for a single connection. Two things everyone notices follow from it:

• Speed tests look fine while a file downloads slowly. A speed test opens several streams at once; a downloader usually opens one.
• A far server is slower than a near one on the same plan. Light travels about 200,000 km/s in fibre, so every 100 km adds roughly 1 ms of round-trip time — and the real figure is 1.5–2× that, because routes aren't straight lines.

Practical rule: pick the nearest server, not the one that sounds impressive.

Six-step diagnosis

1. Measure your baseline without the VPN. Wired, or next to the router. That's your ceiling; the VPN can't beat it.
2. Measure with the VPN on a near server. Losing 10–25% of the baseline is normal.
3. Change server within the same country. If speed jumps, the problem was node load, not geography.
4. Try a different network. Wi-Fi versus mobile data. Fast on mobile and slow at home means your home provider is shaping.
5. Check MTU. Symptom: small requests are instant, large files stall and break. Lower it to 1380 in the client.
6. Measure at different times of day. A dip strictly in the evening is congestion on your provider's side.

Why it's slow only in the evening

Between 7 and 11 pm the provider's uplink is busy, and operators apply traffic shaping: heavy flows get lower priority. If your provider can recognise VPN traffic by its protocol signature, that's the flow that ends up at the back of the queue.

Protocol matters here. Older protocols — OpenVPN, IKEv2, plain WireGuard — have a recognisable packet pattern, and network equipment tells them apart from ordinary web traffic. VLESS with Reality masking looks like a routine HTTPS session to a genuine site, leaving nothing to attach a rule to. How that inspection works: what DPI is. Protocol comparison: VLESS, OpenVPN or WireGuard.

Speed on a phone

Mobile adds its own causes:

• Weak signal. At one or two bars the radio spends more time retransmitting, and any dip is amplified. Test where reception is solid.
• Battery saver. It limits background activity and can slow reconnection.
• Switching between Wi-Fi and mobile. The tunnel rebuilds at the moment of transition. A client with auto-reconnect does it in seconds; without one the connection hangs.

What doesn't affect speed, despite what you'll read

• The provider's total server count. Your speed depends on one node, not on how many exist.
• The cipher. The difference between modern ciphers isn't perceptible on a phone.
• How many of your devices are connected, unless the service caps bandwidth per account. Your own line is the limit.

What BessyConnect does about speed

Three things that move the needle: dedicated servers on 10 Gbit/s links rather than shared hosting, a spread of locations so there's always a near one, and VLESS with Reality, which operators can't sort into a separate queue. Plus automatic reconnection when your network changes, so the tunnel doesn't hang while you move between Wi-Fi and mobile data.

• iPhone, iPad, Mac, Apple TV and Apple Watch — App Store
• Android — Google Play
• Android TV and set-top boxes — Google Play

FAQ

How much should a VPN slow me down? On a nearby server, losing 10–25% of your unprotected speed is normal. A drop of several times means an overloaded node, a wrong MTU, or shaping by your provider.

Why does a speed test look fine while video buffers? Speed tests open several connections; a player usually opens one, and a single stream is capped by window size divided by latency. Distance hits streaming hardest.

Does changing protocol help? Yes, when your provider recognises the traffic. A masking protocol looks like ordinary HTTPS, so no separate rule applies to it.

What is MTU and why change it? It's the largest packet size. If it exceeds what the path allows, packets fragment and throughput collapses. 1380–1420 suits most mobile networks.

Can a VPN ever be faster than no VPN? Yes, in one case: when your provider routes traffic badly or throttles a specific service. The tunnel then straightens the path or removes the limit.

Related articles

How to Set Up a VPN on Android with BessyConnect

How to set up a VPN on Android with BessyConnect: install from Google Play, sign in, allow the VPN connection, pick a server country and enable always-on VPN. More

VPN on a Router: How to Set It Up and Is It Worth It

How to set up a VPN on router firmware: why VLESS+Reality won't run on stock routers and why BessyConnect apps are simpler and safer for the whole home. More

Why Your VPN Disconnects by Itself: It's the Phone

VPN keeps disconnecting by itself? Nine times out of ten the phone unloads the tunnel to save battery. Where autostart hides in MIUI, EMUI, ColorOS and iOS. More