Does a VPN slow down your internet? Honest numbers
Short answer: yes, slightly. WireGuard adds about 5-10% overhead. VLESS Reality adds 10-15%. Connecting to a server far from you costs more than the protocol overhead. Most of the time you will not notice unless you are running a speed test.
Long answer below, with numbers from our own fleet.
What causes the slowdown
Three things, in roughly this order of impact:
1. Server distance. Your traffic now takes a detour. If you are in Berlin and connect to a Fexyn server in Frankfurt (300km away), the detour adds maybe 5-10ms of latency. If you connect to Ashburn, Virginia (6,500km away), the detour adds 80-120ms. Latency adds up across the round-trips that make up a webpage load.
2. Encryption overhead. Encrypting and decrypting packets costs CPU. On modern hardware (any laptop or phone from the last 5 years), this is barely measurable for the throughput most home connections deliver. On older devices or when running at multi-Gbps speeds, the encryption overhead becomes visible.
3. Protocol-specific overhead. WireGuard adds about 60 bytes of overhead per packet (8 byte UDP header, 16 byte WireGuard header, 16 byte authentication tag, plus the encryption padding). VLESS Reality adds more because of the TLS handshake and the additional framing. On a network with 1500-byte MTU, this means roughly 4% of every packet is overhead rather than payload.
Our actual numbers
We measure protocol performance on our fleet quarterly. Last test was March 2026, on a 1Gbps fiber connection in Frankfurt connecting to our Frankfurt server (so server-distance latency is minimised — we want to isolate the protocol overhead).
| Test | Direct (no VPN) | Fexyn Bolt (WireGuard) | Fexyn Stealth (VLESS Reality+Vision) |
|---|---|---|---|
| Download throughput | 942 Mbps | 891 Mbps (-5.4%) | 824 Mbps (-12.5%) |
| Upload throughput | 935 Mbps | 878 Mbps (-6.1%) | 802 Mbps (-14.2%) |
| Latency to Cloudflare 1.1.1.1 | 7ms | 8ms (+1ms) | 11ms (+4ms) |
| Latency to a server in Tokyo | 248ms | 251ms (+3ms) | 257ms (+9ms) |
| Time to first byte (cnn.com) | 89ms | 95ms (+6ms) | 108ms (+19ms) |
These are best-case numbers — same-region server, low-load test. Real-world performance varies. The pattern is consistent: WireGuard is fast, Reality has noticeable but not crippling overhead, server distance matters more than protocol choice.
When the slowdown is noticeable
For most everyday use — browsing, email, video calls, streaming — the overhead is invisible. You will not notice WireGuard is on. Reality you might notice during the initial connection (the handshake takes about 100ms longer than WireGuard) but not during sustained use.
Where you will notice:
- Speed tests. Speedtest.net showing 940 Mbps without VPN and 824 Mbps with Fexyn Stealth is a real measurement. Whether it matters depends on whether you can use the difference.
- Real-time gaming. A few extra milliseconds of latency are perceptible in competitive shooters or fighting games. WireGuard is the right choice for gaming on networks where it works; Reality's extra ~5ms is not great.
- Large file transfers at gigabit speeds. A 50GB Steam download will take noticeably longer through Reality than direct, because the per-packet overhead compounds. Direct is faster if you do not need the VPN's other properties.
Why Reality is slower than WireGuard
Three reasons:
TCP vs UDP. WireGuard runs over UDP. Reality runs over TCP. TCP has head-of-line blocking — if one packet is lost, every packet behind it has to wait. On networks with even small amounts of packet loss (mobile networks, congested Wi-Fi, lossy international links), this turns into stalls that UDP avoids.
Bigger handshake. Reality does a real TLS 1.3 handshake to a real public site, then forwards the certificate, then establishes the inner VLESS session. WireGuard does a one-round-trip Noise handshake. Reality's handshake is more like 3-4 round trips and includes a remote dependency (the camouflage host has to respond fast).
Heavier crypto. Both protocols use modern AEAD ciphers (ChaCha20-Poly1305 typically), which are fast. But Reality also wraps everything in TLS 1.3, which adds an outer layer of encryption that WireGuard does not have. The CPU overhead is small on modern hardware but it adds up at gigabit throughput.
Why we still ship Reality
On a clean network, WireGuard is faster and simpler. There is no reason to use Reality if WireGuard works.
The point of Reality is networks where WireGuard does not work. Russia, China, Iran, Pakistan, UAE, Saudi Arabia, parts of Turkey. In those networks, WireGuard's handshake gets blocked within seconds. Reality's handshake — which is a real TLS handshake to a real public site — survives. The protocol guide has the full technical detail.
The Fexyn client tries Bolt (WireGuard) first by default. If Bolt is blocked, it switches to Stealth (Reality+Vision) automatically. You get WireGuard's speed when it works and Reality's censorship resistance when you need it. The protocol decision is made per-network, not per-account.
Things that hurt VPN speed more than the VPN
Worth checking before blaming the VPN:
The remote server. If a website is slow without a VPN, it will be slower with a VPN. Test against a known-fast destination (Cloudflare, Google) before assuming the VPN is the problem.
Wi-Fi congestion. Crowded apartment buildings, hotel Wi-Fi, public Wi-Fi: the access-point's contention is often the bottleneck, not the internet connection. The VPN cannot help there.
ISP throttling. Some ISPs throttle specific destinations (Netflix, YouTube) or specific protocols (BitTorrent). A VPN routes around this. If you see better speeds with the VPN on than off for a specific service, your ISP was throttling that service. Cox and Comcast have done this historically; some Indian and Brazilian ISPs do it now.
Server load. All VPN providers' servers can get crowded. Switching servers usually fixes this. Fexyn auto-balances connections across our fleet, but you can also manually pick a different server if the default is slow on your route.
MTU mismatch. WireGuard with the wrong MTU setting can fragment every packet, killing throughput. The default settings should be right for most users; if you see weirdly slow throughput on otherwise-fast connections, MTU is worth checking. We document this in our support docs.
How to minimise the slowdown
If you are paying the VPN overhead, get the most for it:
- Use the closest server that works. Frankfurt for European users; Cyprus for Gulf and Middle East; Helsinki for Scandinavia and northern Russia; Ashburn for North America.
- Use Bolt (WireGuard) when you can. Only use Stealth where Bolt is blocked. The auto-switch logic handles this; the manual override is for users who know their network.
- Test, don't assume. Run a speed test with the VPN connected to the server you actually use. If it's wildly different from our numbers above, something is wrong — different server, network issue, or your hardware is the bottleneck.
- Mobile networks need different math. Cellular networks have variable packet loss. Bolt over UDP can struggle here; Stealth over TCP is sometimes better despite the protocol overhead, because TCP retransmission handles the loss more gracefully than UDP loss.
Frequently asked
Will a VPN actually speed up my internet?
In specific cases, yes. If your ISP throttles Netflix or YouTube, a VPN routes around that and you see real speed improvement. If your ISP routes traffic poorly (some emerging-market ISPs use bad peering), a VPN's path can be faster than direct. Otherwise, a VPN can only be as fast as direct, never faster.
Why is my VPN suddenly slower than yesterday?
Usually one of: server load (try a different server), network congestion at your local ISP (try direct, see if direct is also slow), or destination-specific throttling (test against a different destination). Restart the VPN client first; that fixes a surprising number of mystery slowdowns.
Does WireGuard or OpenVPN have less overhead?
WireGuard. By a meaningful margin. OpenVPN's per-packet overhead is roughly 2x WireGuard's, plus OpenVPN's TLS handshake is heavier and the protocol is more verbose. WireGuard is the modern choice when both work.
Will a faster server location help if I'm streaming?
Yes, mostly. Streaming services serve content from CDNs that have nodes close to most VPN exits. The bottleneck is usually the path between you and your VPN server, not between your VPN server and the streaming CDN. Closer VPN server = better streaming.
Try Fexyn free for 7 days — Bolt (WireGuard) for speed when it works, Stealth (VLESS Reality+Vision) for the markets where it does not.