Buying a dedicated server comes down to five decisions: the processor, the amount of memory, the type and size of storage, the network port, and the city it sits in. Get those right and almost everything else, from the operating system to the control panel, can be changed later. This guide walks through each one with the questions to ask yourself.
Start with the workload, not the spec sheet
Write down what the server will run and roughly how busy it will be. A WordPress site with 20,000 visits a month, a 100-player Rust server, a Proxmox host for eight virtual machines and a 30 TB backup target need very different machines. Each of those has a resource that runs out first:
- Websites and APIs usually hit memory and disk IO before CPU.
- Game servers are limited by single-core speed.
- Virtualization hosts run out of memory, then cores.
- Databases want memory first, then fast storage.
- Backups and media libraries need raw disk capacity.
Size the server for that first bottleneck and give the others reasonable headroom.
Processor: cores vs clock speed
Server CPUs trade clock speed for core count. A four-core Xeon E3 or E-2000 series chip runs each core fast, which suits game servers, single-threaded applications and small sites. A dual-socket Xeon E5, Xeon Scalable or AMD EPYC system offers 16 to 64 or more cores at lower clocks, which suits virtualization, containers, build servers and anything that handles many requests in parallel.
Two practical rules help. First, older generations are not a problem for most jobs: a Xeon E5 v3 or v4 from a few years ago still handles typical web and application loads easily and costs much less. Second, count physical cores, not threads. Hyper-threading adds perhaps 15 to 30 percent of throughput, not double.
Browse servers with 16 or more cores if you are building a virtualization or CI host, or game hosting servers if you care most about per-core speed.
Memory: how much RAM you need
Running out of RAM is the most common cause of a slow server, because the operating system starts swapping to disk. Some rough starting points:
| Workload | Starting RAM |
|---|---|
| Single website, small mail server, VPN | 8 to 16 GB |
| Several sites with a control panel such as cPanel or Plesk | 16 to 32 GB |
| One or two game servers | 16 to 32 GB |
| Database server for a growing application | 32 to 128 GB |
| Virtualization host | 64 GB and up, sum of all VMs plus 10 percent |
Memory is cheap relative to the rest of a server, so leaning one step higher than you need is usually money well spent. For large hosts, see servers with 128 GB or more.
Storage: NVMe, SSD or hard drives
Storage type affects how a server feels more than almost any other choice.
- NVMe SSDs connect over PCIe and are the fastest option. Choose them for busy databases, search indexes and build pipelines.
- SATA SSDs are a large step up from hard drives and fast enough for nearly every website and application server.
- Hard drives (HDD) are slow at random access but give you the most capacity per dollar. They are ideal for backups, archives and media.
Look at the drive count as well as the type. Two identical drives let you run RAID 1, so the server keeps working if one fails. A popular layout for storage servers is a small SSD for the operating system and large hard drives for data. Whatever you choose, keep backups somewhere else; RAID protects against a dead disk, not against deleted files. Compare SSD and NVMe servers with HDD storage servers.
Network: port speed and bandwidth
The port speed is the maximum rate the server can send and receive, for example 1 Gbit/s. Bandwidth or traffic is how much data you can move in a month. "Unmetered" means you are not billed per gigabyte; you can use the port continuously at its stated speed.
For most sites a few hundred megabits per second is far more than enough. Streaming, large downloads, backup replication and busy game communities benefit from a 1 Gbit/s port or faster. The port speed and traffic terms appear on every listing and in the specification sheet.
Location: put the server near your users
Network latency rises by roughly one millisecond of round-trip time for every 100 km of fiber. Pick the city closest to most of your users, or a central hub if they are spread out. In North America, Dallas and Chicago reach both coasts well. In Europe, Frankfurt and Amsterdam are the safe defaults. Data residency rules may decide for you: Canadian or EU personal data is simplest to manage on a server in Canada or Europe.
Before you order, test real network paths with the looking glass. It runs ping, traceroute and MTR from each data center to any address you enter.
Management, operating system and support
A dedicated server comes with full root or administrator access. You choose the operating system at checkout; Ubuntu, Debian, AlmaLinux and Rocky Linux are common, and Windows Server, Proxmox and VMware ESXi are usually available. Hardware faults are handled by the data center. Software, security updates and backups are your responsibility unless you add a management service, so budget a little time each month for patching.
A quick checklist
- Name the workload and its first bottleneck.
- Pick cores or clock speed to match.
- Size RAM one step above your estimate.
- Choose NVMe, SSD or HDD, and decide whether you need RAID.
- Check port speed and traffic terms.
- Pick the city nearest your users and test it in the looking glass.
- Plan off-server backups.
When you are ready, filter the full server list by these choices, or ask us to suggest a configuration.