minecraft server hosting
Game server hosting
Tick rate is a single-thread problem. Everything else is noise.
What we would deploy
€76.23/mo
€108.90−30% annual
A game server loop runs on one core, and the number that decides whether your players complain is how fast that core is and how consistently it can run without being interrupted. We pick the highest-clocked silicon on the market, pin it, move kernel housekeeping off it with isolcpus and tickless mode, then put game-layer filtering in front of it so an attack on your community does not become a support ticket for you.
Why this configuration
NITRO runs Zen 5 at 5.9 GHz with cores pinned and isolated, which keeps frame-time jitter under a millisecond even with a full server. The L7 filtering profile understands Source-engine, FiveM, Minecraft and TeamSpeak traffic patterns, so it drops the junk without dropping your players.
What we would deploy
Sizing
Size on single-thread headroom, not core count. One authoritative game loop runs on one core. A 128-tick shooter with 12 slots wants one NITRO core at full clock and 4 GB. Rust or large Source servers past 100 players need two cores' worth of headroom, because the loop spikes on entity churn. Minecraft with 40 players and a plugin stack: 4 cores, 12 GB, most of it heap. Keep steady-state load under 60%.
Tick budget arithmetic
A tick is a hard deadline. At 64 tick you have 15.6 ms per frame; at 128 tick, 7.8 ms. That budget covers physics, hit registration, AI and network serialisation for every connected player. Overrun once and the server drops a tick. Players read that as rubber-banding and lost shots. Profile p99 frame time, never the mean. The mean stays flat while round start, map change and mass entity spawns blow straight through the budget. Size so steady-state frame time sits under half the budget. The spikes then still land inside it.
What L7 game filtering actually does
Volumetric scrubbing stops the pipe filling. It does nothing about a query flood made of valid packets. Source engines answer A2S_INFO from any address, so an attacker sends small queries and your tick loop pays for the replies, or reflects them at somebody else. Filtering at layer 7 means the scrubber speaks the protocol: challenge-response on queries, per-source rate limits on handshakes, malformed packets dropped before they reach the game process. The 15 Tbps of inline capacity is the floor. Protocol awareness is the part that keeps the loop on schedule.
What goes wrong
- Buying burst or shared vCPU for a tick loop. Sustained load drops the boost clock, p99 frame time roughly doubles, and players feel it before your graphs show it. Pin the cores on NITRO.
- Running the game server and its MySQL or Redis backend on the same cores. An fsync stall or a GC pause lands inside the tick window and the server drops frames under load.
- Sizing on average frame time. The mean stays flat while round start, map rotation and mass entity spawns overrun the budget. Provision against p99, measured during those events, not during idle.
Tune the machine
- CPU scheduling — Pinned + isolcpus + tickless · €18
- DDoS protection — Game-layer L7 filtering · €24
- Everything else — Second-resolution metrics + alerting · €5
More
Built for specific jobs
Trading and low-latency VPS
Distance to the matching engine, and nothing between you and the wire.
Private VPN and proxy endpoints
Your own exit, in a jurisdiction you picked on purpose.
Seedboxes and bulk storage
Terabytes that stay cheap and a port that stays open.
Kubernetes worker nodes
Cheap per core, dense, and identical every time.
AI inference and fine-tuning
A whole GPU, passed through, on a commitment that makes the maths work.
Hosting resellers
Your brand on the panel, our hardware underneath.