Cloud Servers
Choosing Between ARM, Standard and Dedicated vCPU Servers
KapsuleHost Servers come in three processor families, and the right one for you is decided by three questions: does your software run on ARM, which region do you need, and does your workload need guaranteed cores.
Everything else is the same across all three. You get full root access, the same KPanel tools, the same firewall and patching, the same daily offsite backups and the same snapshots. The difference is the processor underneath, and that difference shows up in price, in performance consistency, in which regions you can pick, and in whether some specific piece of software will run at all.
The Three Families at a Glance
| ARM | Standard x86 | Dedicated vCPU | |
|---|---|---|---|
| Best for | Cost-efficient web apps, APIs, containers, background workers | Broadest compatibility, general workloads | Steady performance under sustained load |
| CPU | 64-bit ARM | Shared 64-bit x86 | Dedicated 64-bit x86 cores, not shared |
| Price position | Lowest for a given vCPU and memory | Middle | Highest, because the cores are reserved for you |
| Regions | Europe only | Europe and Singapore | All six regions |
| Stock | Frequently constrained, reserve queue available | Normally available | Normally available |
| Tiers | Photon, Quark, Lepton, Muon | Meteor, Comet, Atlas, Apollo, Vega, Pulsar | Quasar, Orion, Nebula, Nova, Sirius, Andromeda |
Every tier in every family comes in two forms: unmanaged Cloud VPS, where you run the operating system, and Managed Cloud Servers, where we do. That choice is independent of the family, and Managed Compared With Unmanaged Cloud Servers covers it.
Where to Compare Them
Sign in to KPanel, open Store in the left sidebar, and choose Cloud VPS or Managed Cloud Servers. The full ladder, including bare metal, storage and GPU, is at /plans/compute, grouped by family with an Unmanaged and Managed toggle at the top.

Each tier shows its specifications, its monthly price in your billing currency excluding GST, and a live stock badge. Each family also carries a short panel telling you when to choose it and what the trade-offs are, which is the same guidance set out below.
ARM: The Best Value
ARM gives you the most vCPU and memory for your money. Modern ARM processors are quick and power efficient, and the great majority of current software has native ARM builds: Linux itself, Node.js, Python, Go, PHP, Ruby, Java, Docker, nginx, PostgreSQL, MySQL, Redis, and almost all widely used open source tooling.
Choose ARM when:
- Your whole stack has 64-bit ARM builds. Most modern runtimes and databases do.
- You want the lowest price for a given vCPU and memory.
- Europe is a fine place for your server to live.
The trade-offs, stated honestly:
- ARM only, with no emulation. We do not emulate x86 on ARM. Software with no ARM build will not run, and the install fails cleanly rather than limping along at a fraction of the speed. That is deliberate: a refused build is easier to diagnose than a mysteriously slow one.
- Europe only. No United States, no Singapore.
- Frequently out of stock. ARM capacity refills from cancellations, so waits can run to several weeks at busy times. You can reserve rather than wait at the screen: see Regions, Availability and the Reserve Queue.
Standard x86: The Compatibility Choice
Standard servers use shared 64-bit x86 cores, the same architecture as most laptops and desktops. Everything that runs on ARM also runs here, plus everything that only ships as an x86 binary.
Choose Standard when:
- Your application depends on software with no ARM build.
- You run proprietary or precompiled x86-64 binaries with no ARM release.
- You are not sure whether your stack is ARM-ready and want the safe option.
- You need Singapore, which ARM does not offer.
Software That Will Not Run on ARM
This is the concrete reason Standard exists on the ladder despite ARM being cheaper. The order surface lists these as x86-only:
- wkhtmltopdf, widely used for PDF and screenshot generation. Its patched-Qt binaries are published for x86-64 only.
- The official Mautic Docker image, which is published for x86-64 only, with no ARM manifest.
Beyond those two, the same reasoning applies to any closed-source vendor binary, licensed media processor, legacy enterprise agent, or old container image that was only ever published for linux/amd64.
Not sure whether your stack is ARM-ready? Check your dependencies' release pages for an arm64 or aarch64 build, and check your Docker images for a linux/arm64 manifest. If everything you rely on has one, ARM will save you money. If even one critical piece does not, choose Standard and stop worrying about it.
The trade-offs:
- A little more expensive than ARM for the same vCPU and memory.
- Shared vCPU, which is fine for most workloads but not guaranteed under sustained full load.
Dedicated vCPU: Predictable Performance
On ARM and Standard tiers, CPU cores are shared with other tenants. That is efficient and perfectly good for most workloads, but it means your throughput under heavy load depends partly on what your neighbours are doing. Dedicated vCPU tiers give you cores that are reserved for you alone.
Choose Dedicated vCPU when:
- You need consistent performance under sustained load, with no noisy neighbours.
- You want the widest region choice: this is the only family in all six regions, including the United States.
- You are running production workloads that cannot tolerate shared-core variance: busy databases, high-traffic application servers, CI runners, video and media processing.
The trade-off: a higher price than shared ARM or x86 for the same vCPU count. You are paying for the guarantee.
Deciding in Four Questions
- Do you need a server in the United States? Then Dedicated vCPU, the only family there.
- Does your stack include software with no ARM build? Then Standard x86, in Europe or Singapore.
- Does your workload need guaranteed cores under sustained load? Then Dedicated vCPU.
- None of the above? Start with ARM for the best value, and move up a tier later if you outgrow it.
You can resize freely up and down within a family, but you cannot resize across families, because the underlying hardware is different. Choosing a family is the one decision worth spending a few minutes on. Choosing a tier within it is not: start smaller than you think and resize when the metrics say so.
Beyond the Cloud Families
The three families above are virtual servers. The ladder continues into single-tenant physical hardware, which is a different proposition again:
- Dedicated Servers (Bedrock, Slate, Granite, Basalt, Titan) for bare-metal performance.
- Storage Servers (Shelf, Trench, Abyss) for very large capacity with no egress fees. See Storage Servers.
- GPU Servers (Spark, Forge, Foundry, Reactor) for AI training and inference. See GPU and AI Compute Servers.
Bare-metal, storage and GPU servers carry a one-time setup fee at checkout in addition to the monthly price, because they are physical hardware provisioned for you. ARM, Standard x86 and Dedicated vCPU cloud servers have no setup fee at all. See Cloud Server Pricing and GST.
If you would like a second opinion on which family fits your workload, ask Kora inside KPanel or reply to any Kapsule email. Our New Zealand team would rather help you pick correctly than watch you fight an unsupported binary.