The Hardware
Behind the Speed

Game servers are fundamentally different from web servers, databases, and most cloud workloads. A web server handles thousands of small, parallel requests, spread that across as many cores as you like and throughput scales with it. A game server runs a single game world: one physics simulation, one AI tick, one netcode loop. All of that runs on one or two CPU cores at any given moment. Adding more cores doesn't help. Clock speed does.

This is why we don't benchmark our hardware on multi-threaded throughput or virtualisation density. We benchmark it on what matters to you: how quickly one core can execute a game tick, how consistently it maintains that speed under sustained load, and how low the variance is (what server administrators call "server tick jitter", the variation that causes lag spikes even when average performance looks fine).

Our Normal tier's EPYC processors at 4.5GHz+ handle the majority of games without breaking a sweat. But some games, Satisfactory is the most extreme example, scale their CPU load with factory complexity in ways that can genuinely saturate a single core on a lower-clocked machine. Heavily modded ARK or Rust servers with high NPC counts, complex AI, or many active vehicles behave similarly. For those workloads, the Ryzen 9 at 5.5GHz+ on our Extreme tier is meaningfully faster, not 10% faster, but 20–30% faster at sustained peak single-thread load, which is often the difference between smooth 30-tick gameplay and noticeable rubber-banding.

NVMe storage matters too, though differently. World saves, map loads, and chunk generation are all disk-bound operations. A world save on a spinning disk can cause a momentary server freeze while the write completes; on NVMe, the same operation is nearly invisible. For games that save frequently (Valheim, ARK, Satisfactory, 7 Days to Die) or generate chunks dynamically (Minecraft, Terraria), NVMe is the difference between saves that are painless and saves that players notice.