How to Choose a VPS Size: vCPU, RAM and Disk Explained
October 7, 2026 · 5 min read
Picking a server size is mostly about memory. CPU and disk matter, but running out of RAM is what makes a small server crawl. Here's how to size it without guesswork.
The three numbers
- vCPU — how much work the server can do at once. Web apps and bots rarely need more than one or two.
- RAM — the working memory. Databases, Docker containers and game servers eat it first.
- Disk (NVMe) — your OS, apps and data. Easy to estimate: add up what you'll store, then leave headroom.
Kvantis plans at a glance
| Plan | vCPU | RAM | Disk | Price |
|---|---|---|---|---|
| Atom | 1 | 1 GB | 25 GB | $10/mo + VAT |
| Spark | 1 | 2 GB | 55 GB | $18/mo + VAT |
| Core | 2 | 4 GB | 80 GB | $34/mo + VAT |
| Turbo | 4 | 8 GB | 160 GB | $62/mo + VAT |
| Titan | 6 | 16 GB | 320 GB | $120/mo + VAT |
High Frequency and High Performance plans are available under All plans when you need faster CPUs.
Matching workloads to plans
- Atom — a Telegram or Discord bot, a WireGuard VPN, a small static site.
- Spark — WordPress, a Node.js or Python web app, a Vaultwarden password manager.
- Core — several Docker containers, Nextcloud for a small team, a modest database.
- Turbo — game servers, busier databases, CI builds.
- Titan — heavy workloads, large databases, CPU-only AI models.
How to tell if you picked right
Log in and run:
free -h # memory in use and available
htop # live CPU and memory per process
df -h # disk usage
If available memory is regularly near zero, or the system is swapping, move up a size. If it sits mostly idle, you can go smaller.
Start small — it costs almost nothing to test
Billing is hourly, so trying a size for an afternoon costs cents. Deploy the smaller plan, load it with your real workload, and if it struggles, deploy the next size and move over.
Pick a plan and deploy — billed by the hour, paid in crypto.