This VPS sizing guide is here to help you answer one deceptively simple question: how much server do you actually need? Buy too little and your site swaps, stalls or crashes under load; buy too much and you pay every month for headroom you never touch. The goal of right-sizing is to match your VPS to your real workload, RAM, vCPU, NVMe storage and bandwidth, so performance is smooth and the bill stays sensible.
Below we break down each resource, explain how to estimate your needs, and give concrete recommended specs for the most common workloads, from a small WordPress blog to a busy WooCommerce store or a database backend. If you are still deciding between hosting types entirely, start with our hosting types decision guide, then come back here to size your VPS.
Why VPS Sizing Matters
Sizing matters because a VPS gives you dedicated, guaranteed resources, and you pay for what you reserve. Under-provisioning is the more painful mistake: when an application runs out of memory it starts swapping to disk (very slow) or the Linux OOM killer terminates processes to survive, which can take your site or database offline. Over-provisioning is less dramatic but wasteful, you burn budget on capacity that sits idle.
Getting it right is also about honesty in specifications. Two VPS plans with the same “4GB RAM, 2 vCPU” label can perform very differently depending on the CPU generation, whether cores are dedicated or oversold, and whether storage is enterprise NVMe or slow shared disk. That is why we favour transparent specifications before you buy, so a given number actually means something.
How to Use This VPS Sizing Guide
Work through this VPS sizing guide in order. First, identify your workload (a single site, several sites, a store, a database, an app). Second, estimate RAM, since it is usually the first ceiling you hit. Third, check CPU and vCPU needs, then storage and bandwidth. Finally, match your numbers to the recommended-specs table near the end. If you are migrating an existing server, measure your current usage first (we show the commands below) rather than guessing.
How Much RAM Does Your VPS Need?
For most web and database workloads, RAM is the first bottleneck, so start here. Work out how much RAM you need by adding up the memory used by your application, its database, the operating system, and a comfortable buffer for caching and traffic spikes. When RAM runs out, the system swaps or kills processes, both of which feel like an outage to your visitors.
Here are sensible rules of thumb, not guarantees, since plugins, themes and traffic all shift the numbers:
- Small static or personal WordPress site: around 1–2GB.
- A real, plugin-rich WordPress site: 2–4GB.
- WooCommerce or another e-commerce store: 4–8GB or more.
- A dedicated database or a busy application: 8GB and up.
- Multiple sites on one VPS: add the individual estimates together, plus overhead.
Before migrating, always measure what you actually use today. On Linux, free -h shows memory in human-readable units; for a full explanation of the output columns (used, free, available, buffers/cache) see the Linux free(1) manual. The “available” figure is the honest one to watch, not “free”.
free -h # RAM: watch the "available" column
top # or htop, for live CPU and memory per process
df -h # disk space used vs free per mountHow Much CPU and vCPU Do You Need?
When it comes to how much CPU you need, quality beats quantity for most websites. PHP and WordPress requests are largely single-thread-bound, so the speed of a single core and whether that core time is genuinely dedicated matter far more than a big vCPU count on paper. In practice, 2 fast, dedicated vCPUs will often out-perform 4 slow, oversold ones that you are quietly sharing with noisy neighbours.
A higher vCPU count earns its keep for genuinely parallel work: handling many concurrent requests, running background jobs, video encoding, compiling, or CI builds. For a typical brochure or blog site, prioritise per-core speed; for a busy store or app with lots of simultaneous users, add cores. We explain the trade-off in detail in dedicated vs shared vCPU, why fast cores beat many. UK Speed uses AMD EPYC processors with dedicated vCPU, so each core does more real work than an oversold budget core of the same “count”.
How Much Storage and IOPS Do You Need?
Size your storage for your data plus realistic growth, plus backups, logs and temporary files, then leave headroom. A disk that runs near-full slows down and risks failed writes, so treat 70–80% used as your practical ceiling and plan to scale up before you hit it.
Capacity is only half the story; speed matters just as much. IOPS (input/output operations per second) and latency determine how quickly your database can read and write, and this is where NVMe storage transforms performance versus older SATA SSDs or spinning disks. Databases, WooCommerce carts and any write-heavy, busy site feel this most. Not all flash is equal, either, our comparison of enterprise NVMe versus consumer NVMe explains why sustained, consistent IOPS matter more than a headline peak. Use df -h to see current disk usage before you choose a size.
How Much Bandwidth Do You Need?
Most websites use surprisingly modest bandwidth: text, images and normal browsing rarely trouble a generous monthly allowance. Bandwidth becomes a real planning factor when you serve large media, offer downloads, stream video, or run a VPN. Two numbers matter here, and they are different things: port speed (how fast data can move at any instant, for example 1Gbps) and monthly data transfer (the total volume included per month).
For a typical site, a fast port with a comfortable transfer allowance is plenty; keep this part qualitative and only reach for high-transfer plans if you know you push a lot of data. If you plan to consolidate several projects, our guide to hosting multiple websites on one VPS covers how their combined traffic adds up.
Choose VPS Specs by Workload
Use the table below as illustrative starting points, not fixed rules. Your real numbers will vary with traffic, plugins, caching and how the application is built. When in doubt, pick the row that matches your workload and start one step above your current usage.
| Workload | RAM (start) | vCPU | NVMe storage |
|---|---|---|---|
| Small / personal WordPress | ~2GB | 1–2 | 25–40GB |
| Business WordPress / small WooCommerce | ~4GB | 2 | 50–80GB |
| Busy WooCommerce / multiple sites | ~8GB | 4 | 100GB+ |
| Dedicated database / SaaS backend | 8–16GB+ | 4+ | Fast NVMe, sized to data |
| Game server / Discord bot | 4–8GB | Strong single-core | Modest, varies |
| AI / LLM inference | GPU VPS | VRAM-driven | Large, model-dependent |
Two quick notes. Game servers and Discord bots vary enormously by player count and framework, but they reward fast single-core performance more than a big core count. AI and LLM work is a different category: it is driven by GPU VRAM rather than ordinary RAM and vCPU, so choose a GPU VPS sized to the model you intend to run.
When to Scale Up Your VPS
The right time to scale up is when your metrics say so, not when you feel nervous. Watch for consistent signals over days, not one-off spikes:
- Memory “available” regularly near zero, or visible swap usage.
- CPU pinned near 100% during normal traffic, not just at backup time.
- Rising response times or database queries queuing under load.
- Disk consistently above 70–80% full.
A key advantage of a VPS is that scaling up is quick, you resize the plan rather than migrating to a new machine. That is exactly why you should start right-sized rather than massively overbuying: you can grow the moment the numbers justify it.
Best Practices for Right-Sizing
- Measure before you migrate, use
free -h,htopanddf -hon your current server rather than guessing. - Start slightly above your current need so you have headroom for spikes, then monitor.
- Prioritise per-core speed and dedicated vCPU for PHP/WordPress; add cores for parallel workloads.
- Choose NVMe storage for anything database-heavy, and never run the disk near-full.
- Enable caching (page and object caching) to make a smaller VPS go further.
- Remember the hardware behind the number, EPYC cores and enterprise NVMe mean a given spec delivers more real performance.
Conclusion
Right-sizing is not guesswork, and this VPS sizing guide gives you a repeatable method. Size RAM to your app, database, OS and cache headroom; favour fast dedicated vCPU over a big oversold core count; put busy databases on enterprise NVMe; and treat bandwidth qualitatively unless you serve heavy media. Start right-sized, monitor, and scale up when the metrics ask for it.
Do this next:
- Check your current usage with
free -h,htopanddf -h. - Match your workload to the recommended-specs table above.
- Pick a UK Speed EPYC + NVMe VPS one step above your current need.
- Monitor for a week or two, then scale up only if the numbers say so.
Found your size? Get a UK Speed VPS
Every UK Speed plan runs on AMD EPYC and enterprise NVMe with dedicated resources and transparent specs – so the RAM and vCPU you pick perform at their best. Right-size now and scale up anytime.
