vCPU
A vCPU is a virtual processor core allocated to a workload. On most cloud hardware it maps to one hardware thread rather than a full physical core, so two vCPUs often share a single core's execution resources. It is the unit almost all game server hosting is priced and scaled in.
Also called
virtual CPU
,
Is a vCPU a Core or a Thread?
Usually a thread. Modern server processors run two hardware threads on each physical core, a technique called simultaneous multithreading (Intel calls it Hyper-Threading). The operating system sees each thread as a separate processor, and cloud providers sell each one as a vCPU.
The two threads are not two cores. They share the core's execution units and caches, and take turns when both are busy. Together they get more work done than one thread alone, because one can use the core while the other waits on memory. They do not get twice as much.
That is why "vCPU vs core" has no single answer. A 4 vCPU instance on most x86 hardware is two physical cores. Some instance families skip thread sharing and give one vCPU per full core, so the same 4 vCPUs there are four cores. When comparing offers, compare like with like: same vCPU count, same mapping.
In Practices - Insights from Edgegap's Blog
Article's key insights
Tick Rate Is a Cost Decision: In Competitive games, tick rate carries a significant CPU load and bandwidth cost, making tick rate one of the most consequential infrastructure choices in multiplayer development given its cost while having a significant impact on players experience of the game's "feel".
VALORANT's Commitment to 128 Hz: Riot rebuilt core engine systems from the ground up to get from a 50ms server frame time to under 2ms. A years-long engineering investment most studios cannot easily replicate.
CS2's Subtick System: Rather than increasing tick rate, Valve introduced microsecond timestamping of player actions, addressing the quantization problem without multiplying infrastructure costs.
Marathon's Middle Ground: Bungie's extraction shooter demonstrates that the right tick rate depends on genre requirements and operational capacity, not just on what the highest number is.
No Universal Answer: Apex runs at 20 Hz, VALORANT at 128 Hz, CS2 at 64 Hz with Subtick, and Marathon at 60 Hz. Each reflects a different balance of precision requirements, player expectations, and what the studio can sustain.
Why Isn't Every vCPU Equal, and When Does It Matter?
A vCPU is a billing unit, not a performance unit. Two offers at "1 vCPU" can differ in clock speed, in processor generation, and in architecture: x86 or Arm. The last one matters before anything runs, because a server built for amd64 will not start on an Arm host without an arm64 build.
For most game servers, the other differences rarely decide anything. A containerized server runs the same way on any modern host, so a build tested on one provider behaves the same on the next. Hardware has become a minor buying criterion, behind where the servers are and how quickly they start.
The exception is a large simulation on one server. When a single instance holds a big world and a hundred or more players, the speed of each core sets how many players that server can carry, and a faster processor buys real capacity. It is a real case, and a rare one.
How Many vCPUs Does a Game Server Need?
Fewer than most people expect. Much of a game server's work runs on one main thread: read inputs, advance the simulation, send state, once per tick. Engines move some work to other threads, such as physics or networking, but the main loop sets the pace. Beyond one or two vCPUs, adding more rarely makes a tick finish faster.
Small servers often need less than one. A 1v1 or a small co-op match can run on a fraction of a vCPU, and several such servers fit side by side on one host.
A multi-room server goes further: one server process hosts dozens of matches at once, each in its own room, sharing the process's vCPUs and memory. It suits games with light simulations, such as card, puzzle or turn-based games, and it changes what you size. The question becomes how many rooms a vCPU can carry, not how many vCPUs one match needs.
The reliable way to size is to measure. Run a profiled server build at your target player count and tick rate, and watch how much of each tick's time budget the main thread uses. The result worth keeping is players, or rooms, per vCPU: it is what turns a match into a cost.
What Does a vCPU Cost?
The bill is vCPUs × time, and what you pay depends on how that time is bought.
On demand. Billed only while a server runs. Edgegap's Edge Cloud charges $0.00115 per vCPU per minute, or $0.069 per hour.
Reserved. A host rented by the month, paid whether full or empty. On Edgegap's Private Fleet that works out to $0.0256 per vCPU per hour (platform data, last updated 30 September 2026).
Reserved is cheaper per hour when kept busy. On demand is cheaper for every hour it would sit idle. Across a month, the vCPU-minutes you pay for without a match on them often move the bill more than the rate does.
Reserved hosts are also where the large-simulation exception lives. Edgegap's Private Fleet offers a bare metal host type at 3.7 to 5.1 GHz (private fleet documentation). Path of Titans, Alderon Games' dinosaur MMO with up to 200 players per server and CPU requirements of 4 GHz and above, has run on Edgegap's Private Fleet with the custom hardware the MMO required.
A word from our sponsor (ourselves!)
A region is a guess about where your players will be. Edgegap's Edge Cloud is a distributed, multi-cloud network of 615+ locations across 17+ providers, available on demand. Each server launches at the best available location on the network when the match starts, not at the nearest of a handful of regions.
