VDS: dedicated CPU cores, explained

Every shared VPS on the market - ours included - plays the same game: the host has more virtual CPUs allocated than physical cores, because most servers idle most of the time. It works, it keeps prices low, and for the majority of workloads nobody ever notices. Our new VDS plans are for the workloads that do notice. A VDS gets physical EPYC cores pinned to the VM: the hypervisor schedules your virtual cores onto the same dedicated host cores every time, with top scheduling priority.

What pinning changes in practice

Consistency. A shared vCPU delivers excellent performance on average but can wobble when a neighbor spikes. A pinned core delivers the same performance at 3 a.m. and at peak hour. That matters for game servers where a tick that takes 60 ms instead of 40 ms is a visible stutter, for databases where p99 latency is the number that pages you, and for anything real-time: voice, trading bots, CI runners with strict timeouts.

The plans

Eight fixed presets, all NVMe-backed, all with a dedicated IPv4 - from Freya (2 dedicated cores, 4 GB RAM) at €6.99/mo through Tyr (4 cores), Thor (6), Heimdall (8), Allfather (12), Baldur (16) and Fenrir (24) up to Surtr with 32 dedicated cores, 96 GB RAM and nearly a terabyte of NVMe at €94.99/mo. The pool of dedicated cores is finite by nature - when it is sold out, it is sold out, and we would rather tell you that at checkout than oversell a product whose entire point is not being oversold.

Do you actually need one?

Honest answer: if you are hosting a website, a Discord bot, a small modpack or a dev box, no - a standard KVM VPS at a third of the price will serve you identically. Buy a VDS when you have measured CPU steal, when your workload is latency-sensitive under sustained load, or when you resell capacity and want your customers isolated from noisy neighbors. Both product lines run the same NVMe storage, the same network, the same panel, the same API - the only difference is who owns the cores.