skip to content

How would you choose the fixed tick rate for a dedicated game server, and what does raising it cost the server, the network and the players' clients?

level: principalimportance: should knowfreq 38%

answer

  1. period shrinks as rate rises
  2. work per tick times ticks
  3. input waits for the next tick
  4. client prediction shares the step
  5. hard to change after tuning

basics

~20 s

Choose the lowest fixed tick rate that the game's pace and precision need, measured against per-tick cost at peak load. Raising it multiplies server CPU and often bandwidth, shrinks the time each tick may take, and changes tuning that clients share.

solid answer

~50 s

A dedicated server advances its world on a fixed tick, and the rate is a standing design choice. Higher rates cut how long an input waits before it is simulated, on average half a period, so 25 ms at 20 Hz but about 8 ms at 60 Hz, and they resolve fast interactions more finely. The costs scale with it: at 60 Hz per-tick work runs three times as often as at 20, so CPU per match roughly triples; the period shrinks from 50 ms to 16.7 ms, leaving less headroom for spikes; bandwidth grows if state is sent every tick; and clients predicting locally must step at the same rate, so physics and movement tuning depend on it. I would pick per genre, measure tick cost at peak load with headroom, decouple send rate and slow subsystems, and fix the rate early, since changing it re-tunes the game.

go deeper

for a junior

Recall that a dedicated server advances its world in fixed ticks, and that the tick period is one second divided by the tick rate.

for a middle

Explain the half-period input wait, why per-tick CPU cost scales with the rate, and why clients that predict must share the server's step.

for a senior

Measure tick cost at peak load, keep headroom below the period, and separate simulation rate from send rate and from slow subsystems.

for a principal

Own the rate as a standing decision: weigh responsiveness against hosting cost per match, set it before tuning, and plan per-mode rates deliberately.

## What the tick rate is A **dedicated server** runs the authoritative simulation of a match with no rendering. Its loop advances the world in **fixed ticks**: each tick reads queued player inputs, simulates one fixed step of movement, physics and game rules, and prepares state for clients. The **tick rate** is how many ticks run per second; its inverse is the **tick period**, the real time each tick may take. | Tick rate | Tick period | Average wait for an input to be simulated | |---|---|---| | 10 Hz | 100 ms | 50 ms | | 20 Hz | 50 ms | 25 ms | | 30 Hz | 33.3 ms | about 16.7 ms | | 60 Hz | 16.7 ms | about 8.3 ms | | 128 Hz | about 7.8 ms | about 3.9 ms | The last column assumes inputs arrive at random times, so on average each waits half a period for the next tick. There is no universal rate. Slower-paced and large-scale games often run somewhere in the 10 to 30 Hz range, while some fast competitive shooters run at 60 Hz or above. The rate is a **standing choice** a lead owns because so much else is tuned against it. ## What a higher rate buys - **Responsiveness** - inputs wait less time before the server acts on them, which players feel in fast, precise games. - **Temporal resolution** - fast-moving objects advance less per tick, so the server misses fewer brief interactions and outcomes match what players saw more closely. - **Fresher state** - when clients receive state more often, they see a more current world, though how often state is sent is a separate, replication-layer choice. ## What it costs 1. **Server CPU.** Most per-tick work is roughly constant per tick, so going from 20 to 60 Hz runs it three times as often. CPU per match roughly triples, and the number of matches a machine can host falls by about the same factor, which goes straight into hosting cost. 2. **Headroom.** The period shrinks from 50 ms to 16.7 ms. A spike that fits comfortably in a 50 ms tick can overrun a 16.7 ms one, and a tick that overruns its period puts the server behind real time. 3. **Bandwidth.** If full or delta state goes out every tick, outbound traffic grows with the rate. Many games therefore send state less often than they simulate. 4. **Client coupling.** Clients that predict their own movement must run the same fixed step as the server, or their predictions diverge and get corrected. Movement, physics and jump tuning are made at that step, so the rate is baked into the game's feel. 5. **Per-tick history.** Anything the server keeps for a fixed span of time, such as recent states, grows in proportion to the rate. ## How I would choose - **Start from the genre's needs**: how fast objects move relative to their size, how precise hits must feel, and how many entities a match holds. - **Measure the tick's cost** at peak entity counts on target server hardware and keep clear headroom below the period, rather than sizing on an average match. - **Model the economics**: matches per core at each candidate rate, multiplied out to hosting cost at expected concurrency. - **Decouple what does not need the core rate**: run AI decisions, economy or slow systems every few ticks, and send state at a rate chosen for bandwidth. - **Keep it one constant**, shared by server and client code, and decide it early, because changing it after tuning shifts every trajectory designers balanced. - **Allow per-mode rates** where modes differ sharply, such as a large casual mode and a small competitive one, accepting that each needs its own tuning check. ## The trade in one sentence The right tick rate is the lowest one at which the game still feels responsive and fair for its pace, chosen with measured per-tick costs and hosting economics in view and fixed before designers tune against it.

  • Why can you not raise a shipped game's server tick rate as a simple configuration change?
    Movement, physics and timing were tuned at the old step, and results depend on step length, so trajectories, jump arcs and timing windows shift. Clients predicting locally must change in lockstep with the server, which means a client update. The per-tick cost also multiplies, so hosting capacity must be re-planned before the change ships.
  • Why might a server simulate at a higher rate than it sends state to clients?
    Simulation precision and input responsiveness benefit from short steps, while bandwidth grows with each send. Running the core tick fast and sending state every second or third tick keeps simulation quality while capping traffic; clients interpolate between the states they receive. The exact scheme belongs to replication design, but the loop must allow the two rates to differ.

  • Choosing the highest tick rate the hardware allows without costing hosting
  • Assuming the tick rate must equal the clients' frame rate
  • Treating the tick rate as freely changeable after gameplay is tuned
  • Sizing tick cost on an average match instead of peak load
  • Believing a higher tick rate costs nothing if bandwidth is fine