A Garry’s Mod dedicated server does not have one universal, official RAM or CPU minimum. Facepunch specifies the supported server environment, SteamCMD installation, app ID 4020, launch settings and the need for enough disk space for the content you install. As practical starting points, plan on 1–2 GB RAM and 1–2 fast vCores for a small lightly modded server, about 2 GB and 2 vCores for a public TTT or Prop Hunt server, and 3–4 GB with 2–4 vCores for a larger DarkRP server. These are workload-based allocations, not guarantees: player slots, Lua scripts, Workshop collections and additional server processes determine the capacity you actually need.
What are the actual Garry’s Mod dedicated-server requirements?
For a dedicated server, the most important requirements are a supported operating system, SteamCMD, Garry’s Mod app ID 4020, a Steam Game Server Login Token, fast CPU performance, sufficient memory and SSD space for every file your server downloads. Facepunch does not publish a single numeric server specification for RAM or CPU. That means a server sized for four friends in Sandbox is a very different machine from a 40-slot DarkRP community with databases and hundreds of addons.
The commonly quoted 4 GB or 8 GB figures come from Steam’s Garry’s Mod client listing, not from a dedicated-server specification. Valve’s current listing (2026) gives the client a 2 GHz processor minimum, 2.5 GHz recommended processor, 4 GB RAM minimum, 8 GB recommended RAM, 10 GB minimum storage and 25 GB recommended storage. Those figures include running the graphical game and should not be copied into a server plan.
How much RAM does a Garry’s Mod server need?
Use the following as starting allocations for one server process. They are independent planning guidance rather than Facepunch requirements. Leave room for the operating system and monitoring tools, then measure actual use after your addons and player count are in place.
| Workload | Starting RAM | Starting CPU | Typical reasons to add capacity |
|---|---|---|---|
| Private Sandbox or TTT with a few addons | 1–2 GB | 1–2 fast vCores | More players, larger maps or additional Lua scripts |
| Public TTT or Prop Hunt with Workshop content | About 2 GB | About 2 vCores | A large collection, more slots or simultaneous downloads |
| Larger or heavily scripted DarkRP | 3–4 GB | 2–4 vCores | 30 or more players, databases, many addons or multiple processes |
Memory use rises as maps, materials, Workshop addons and Lua systems load. A server can start successfully with the lower number and still fail later when a new map or collection is introduced. Watch resident memory and swap activity during a busy session; sustained swapping is a sign that adding RAM will help more than adding another virtual core.
Do I need 4 GB or 8 GB RAM for a GMod server?
Neither number is an official dedicated-server minimum. Four and eight gigabytes are Steam client figures. A small, lightly modded server may run within 1–2 GB, while a heavily scripted public server may need 4 GB or more; the right allocation depends on content and concurrency. If the host also runs a web panel, database, voice service or another game server, count those services separately.
What CPU do I need for a GMod server?
Prioritize fast per-core performance and consistent clock speed over a large count of slow cores. Source engine simulation and many Lua addons place substantial work on a main thread, so two fast vCores can outperform several slower vCores for one busy instance. Additional cores still matter when the machine runs multiple server processes, databases, monitoring agents or other workloads.
- For a private Sandbox or small TTT instance, begin with 1–2 fast vCores.
- For a public TTT or Prop Hunt server, begin with about 2 fast vCores.
- For a larger DarkRP server, begin with 2–4 vCores and verify that the host delivers reliable single-core performance.
Do not judge capacity by advertised core count alone. On a VPS, ask whether vCPUs are dedicated or oversubscribed and whether several tenants share the same physical CPU. On a self-hosted machine, check per-core utilization, server frame time and tick stability while players are active. CPU saturation often appears as delayed entity updates or a falling tick rate before the process consumes all system memory.
How much storage does a Garry’s Mod dedicated server use?
Facepunch’s requirement is content-based: “Enough disk space for the content you wish to install.” The installed footprint is therefore not a fixed server number. Budget space for the base dedicated-server files, every map and Workshop addon, logs, crash dumps, update staging and backups or snapshots.
- Base installation: SteamCMD downloads app 4020 and its current server files.
- Content: maps, materials, sounds and Workshop collections can add substantially more data than the base files.
- Operations: logs and crash dumps grow over time, especially on public servers.
- Updates: SteamCMD and Workshop updates need temporary working space while files are replaced.
- Recovery: persistent communities need space for configuration backups, database dumps and snapshots.
An SSD improves file access, map changes and update behavior. SATA and NVMe drives are both suitable storage types; Facepunch does not specify a required model or capacity. A 1TB SATA SSD is a sensible headroom-oriented purchase for a server expected to accumulate maps, collections, logs and backups, but 1TB is not an official minimum. For a small test instance, a much smaller SSD can be sufficient if you actively manage collections and backups.
Can I run a GMod server on Linux?
Yes. Facepunch documents Linux hosting through SteamCMD and the srcds_run launcher. On 64-bit Linux, SteamCMD itself is 32-bit, so the official instructions note that 32-bit compatibility libraries may be required. Install those libraries before treating a SteamCMD error as a Garry’s Mod problem.
Linux installation outline
- Install SteamCMD and the required 32-bit compatibility libraries for your distribution.
- Use SteamCMD to authenticate as appropriate and install or update app ID
4020. - Place your server configuration, maps and Workshop settings in the downloaded server directory.
- Launch with
srcds_run, supplying the game mode, map, player limit and token options. - Confirm that the process remains running, downloads the expected Workshop content and writes logs without filling the disk.
Linux is often chosen for unattended servers because it works well with service managers and remote administration. The practical requirements remain the same: fast CPU time, enough memory for the loaded content, SSD headroom and a stable network location.
How do I install and launch a Garry’s Mod server on Windows?
Windows servers use SteamCMD for installation and updates, then launch the dedicated executable srcds.exe. The documented process requires a Steam Game Server Login Token; Facepunch explicitly marks the token as mandatory.
- Download and unpack SteamCMD, then use it to install or update Garry’s Mod app
4020. - Open the server’s command prompt or create a shortcut that runs
srcds.exe. - Add the launch options your workload needs, such as
+maxplayers,+gamemode,+mapand+sv_setsteamaccount. - Insert your Steam Game Server Login Token with
+sv_setsteamaccount; without it, the documented server setup is incomplete. - Start the process and check the console for map loading, Workshop downloads and network registration before inviting players.
Keep update and backup space on the same drive or on a separately managed volume. A full disk can prevent both SteamCMD updates and normal log writing, making an otherwise correctly sized server appear unstable.
Which operating systems and branches are supported?
Facepunch documents Windows and Linux dedicated servers. macOS dedicated-server files are no longer shipped, so a Mac cannot use the current official dedicated-server distribution directly. Use a supported Windows or Linux host instead.
The official documentation also describes development, prerelease and x86-64 beta branches. Treat those branches separately from stable: beta builds can be unstable, and an addon or configuration that works on stable may fail after a branch change. Keep a backup before switching branches and avoid testing a beta build on the only public instance.
How should I compare a self-hosted machine, VPS and managed hosting?
Compare the resources and policies that affect simulation, downloads and recovery rather than looking only at a headline RAM number.
| Comparison point | Why it affects a Garry’s Mod server |
|---|---|
| Allocated RAM | Determines how much room remains for maps, Lua addons, the operating system and other services before swapping begins. |
| Single-core CPU performance | Influences Source simulation and Lua execution; nominal vCPU count does not reveal per-core speed. |
| SSD capacity and I/O | Controls Workshop and map storage, update staging, log growth and the speed of file-heavy operations. |
| Player-slot policy | Some plans limit slots or charge for higher limits; verify that the advertised allocation matches your target population. |
| Workshop support | Confirm that collections, downloads and launch parameters are supported and that the provider does not block required traffic. |
| Backups and snapshots | Determine whether you can restore configurations, maps and databases after corruption or a failed update. |
| Network location and policy | Choose a region close to players and check bandwidth, DDoS handling and any traffic restrictions. |
| Shared processes | On a self-hosted box or VPS, other servers and databases consume the same CPU, memory and disk budget. |
A managed plan can simplify updates and backups, while self-hosting gives you direct control over hardware and the operating system. A VPS sits between those options: it is easier to relocate than a physical machine, but performance depends on the provider’s CPU scheduling and storage layer. Ask for measured or guaranteed allocations rather than assuming “four vCPUs” means four dedicated physical cores.
What should I monitor after launch?
Initial sizing is only a hypothesis. Collect data during an empty server, a normal peak and the largest event you expect to host.
- Memory: record steady resident usage, peak usage after map changes and whether swap is touched.
- CPU: watch per-core utilization and server frame time, not only total host percentage.
- Tick stability: note tick loss, hitches and delayed entity updates when players and scripted systems are active.
- Disk: track free space, Workshop growth, log directories and backup retention.
- Network: check upload capacity and latency during downloads and busy play sessions.
Increase RAM when the process or host is paging, increase CPU capacity when a core is consistently saturated, and increase storage when updates or logs approach the free-space limit. If several independent servers share one host, measure them together rather than sizing each process in isolation.
When should I upgrade RAM, CPU or storage?
Upgrade RAM when content causes paging
Add memory if peak usage leaves little headroom, the operating system starts swapping or map and addon loading produces out-of-memory errors. A 16GB DDR4 kit can be a practical server RAM upgrade for a host that runs several instances or companion services, but it is not a universal requirement; verify the workload and the machine’s memory type first.
Upgrade CPU when simulation time is the bottleneck
Move to faster cores or a host with more dependable CPU allocation when one core remains busy and players report stutter despite free RAM. Simply adding vCores may not help a single instance if the underlying per-core performance is unchanged.
Upgrade storage before it becomes an outage
Expand the SSD when Workshop collections, logs, backups or update staging reduce free space to a narrow margin. Keep old logs under a retention policy and test that backups can be restored; a large disk without recovery procedures does not protect a persistent community.
What are the most common sizing mistakes?
- Treating client requirements as server requirements: the 4 GB/8 GB RAM and 10 GB/25 GB storage figures describe the Steam client.
- Buying by slot count alone: 20 players with heavy Lua systems can stress a server more than a larger lightly modded session.
- Ignoring Workshop growth: every map, material and sound in a collection consumes storage and may require update staging space.
- Counting vCores without checking speed: shared or oversubscribed CPU time can produce poor tick stability.
- Leaving no recovery margin: a full disk or untested backup can turn a routine update into extended downtime.
- Running several processes on the same estimate: add the memory, CPU and storage needs of each process and its supporting services.
What is the practical starting point for a new server?
For a private, lightly modded Sandbox or TTT server, begin at 1–2 GB RAM and 1–2 fast vCores on an SSD. For a public TTT or Prop Hunt server with a Workshop collection, begin around 2 GB RAM and 2 fast vCores. For a larger DarkRP installation, begin at 3–4 GB RAM and 2–4 vCores, then adjust after observing peak memory, per-core CPU load, tick stability and disk growth. Those allocations are starting points, not promises of a particular player count.
Install app 4020 with SteamCMD, use a supported Windows or Linux environment, provide the mandatory Steam Game Server Login Token, and reserve storage for content, logs, updates and backups. Re-measure whenever you add slots, switch game modes, expand a Workshop collection or run another process on the host.
Frequently Asked Questions
Can a Garry’s Mod server run on a home internet connection?
It can, provided the connection offers adequate upload bandwidth, stable latency and a reachable inbound port. You must also account for power outages, changing public IP addresses and the lack of provider-grade DDoS protection; a hosted location is usually easier for a persistent public community.
How many players can 2 GB of RAM support?
There is no reliable player number tied to 2 GB. Addon memory, map size, Lua behavior and other services can change the result, so treat 2 GB as a starting allocation for a public TTT or Prop Hunt workload and validate it under your real peak load.
Does changing from Windows to Linux reduce the server’s RAM requirement?
The operating system changes the overhead and administration workflow, but it does not remove the memory used by maps, Workshop files or Lua addons. Linux can be efficient for unattended hosting, while the server workload still determines the required allocation.
Do Workshop collections need extra disk space after every update?
Collections consume space for their installed files, and updates need temporary working room while content is downloaded and replaced. Keep free capacity beyond the current collection size and monitor growth rather than allocating only the initial download footprint.