Choosing a dedicated server line for your workload

Worldstream's server list on worldstream.com carries use-case filter shortcuts, Deals, Storage, Virtualization, DevOps, AI/ML, Database, and Gaming, alongside the more familiar filters for data centre location, delivery time, uplink speed, and processor family (Intel Xeon, Intel Xeon E, AMD EPYC, and others). The use-case filters are the faster starting point if you're not already set on a specific model: each one narrows the list toward configurations suited to that workload's real bottleneck, rather than leaving you to compare core counts and clock speeds cold. This article walks through what each use case typically needs and why, so the filtered results make sense rather than feeling arbitrary.

Storage: capacity and drive bays over raw CPU

Storage-oriented workloads, file servers, backup targets, media libraries, bulk object storage nodes, are bottlenecked by how much data fits and how it's laid out, not by CPU horsepower. What matters is the number of drive bays available, the mix of capacity and speed across those bays, and the RAID configuration protecting the data. A modest processor paired with generous, well-configured storage will outperform a high-core-count server with a couple of small drives for this kind of workload. See Choosing storage media: NVMe vs. SSD vs. HDD by workload for how to match drive type to the specific access pattern once you've picked a chassis with enough bays.

Virtualisation: high core count and RAM

A server running many virtual machines needs to divide its resources across all of them simultaneously, so the two figures that matter most are core count and total RAM, both of which set a hard ceiling on how many VMs (and how generously specced each one) the host can support. Storage speed matters too, but usually second: enough NVMe throughput to avoid I/O contention across VMs, rather than the outright capacity that a storage-first workload needs.

DevOps: balanced CPU plus fast local storage

CI/CD pipelines, build servers, and container registries don't fit neatly into a single bottleneck. Build and test jobs benefit from solid multi-core throughput to run steps in parallel, while build caches, container layers, and artifact storage benefit heavily from fast NVMe, repeated reads and writes to the same cache directories are exactly where storage latency shows up as wasted pipeline time. See Choosing a CPU for CI/CD, virtualisation and Kubernetes nodes for a closer look at matching CPU choice to this kind of mixed workload.

AI/ML: cores, memory and data throughput

Training and inference are hungry for memory and for data throughput, so the questions that matter are core count, enough system RAM to stage datasets, and storage fast enough to keep the compute fed rather than idle. The processor family filter alongside the use-case shortcuts covers Intel Xeon and AMD EPYC options. The specific models on offer change over time, so treat the AI/ML filter on the live configurator as the current source of truth rather than any fixed model list. If your workload needs GPU acceleration specifically, ask support what is available before you order.

Database: single-core performance plus fast storage

Most database engines are more sensitive to single-core clock speed and per-query latency than to raw core count, a single query thread typically runs on one core, so a faster core often beats a server with more, slower cores. Storage latency matters just as much: NVMe drives cut the time spent waiting on disk for reads and writes, which shows up directly as query latency. See Choosing storage media: NVMe vs. SSD vs. HDD by workload for more on matching storage to this pattern.

Gaming: low latency and steady clock speed

Game servers are latency-sensitive in a way that's closer to database workloads than to virtualisation or storage: what matters is a high, steady single-core clock speed and consistent frame-to-frame performance, rather than the highest core count available. Network latency to your player base matters here too, which is where the data centre location filter becomes as important as the hardware filter itself.

From use case to order

Once the use-case filter has narrowed the list to a handful of relevant configurations, the rest of the decision comes down to the specific processor, memory, storage and network options shown on each listing's configure page. See Configuring a dedicated server when you order it for what's actually adjustable at that stage and how the ordering flow works.