During GDC 2025 (March 17-21), we will be conducting a >10,000-player test. Before we begin, let’s talk about the tech behind it!
(Questions are collected from various friends, and we welcome more questions, tests, and feedback from those interested.)

Q:
Is it really possible for 10,000 players to exist in the same world at once? You claim to achieve this without sharding, but isn’t server meshing essentially a form of sharding?
Additionally, I’d like to know about latency and throughput, and whether your technology relies on determinism.

A:
Our technology shifts the bottleneck from spatial density to logical density.
Traditional methods like sharding or server meshing, which divide the game world by geographical regions, distribute players across different servers—preventing true interaction.

Instead, we parallelize by “tasks” rather than geographic space, meaning the number of players in an area does not directly impact latency. The key factor is how many objects a task affects and the type of computation involved.

Latency: Under ideal network conditions, player interactions stay below 100ms.
Throughput: In previous tests, we managed 2,000–3,000 logins per second.
Determinism: We do not rely entirely on deterministic synchronization, as it demands strict network conditions and limits developer flexibility.
Beyond Traditional Game Servers
Typical game servers often only relay packets, forwarding client requests to other players.
Our approach lets the server process all game logic directly, ensuring high performance and minimizing additional packet delays.

Let’s push the boundaries of massive real-time interactions.