Dedicated vCPU VPS: VDS hosting with dedicated CPU cores and a 10 Gb/s port

What is a dedicated vCPU VPS?

A dedicated vCPU VPS, also sold as a VDS (virtual dedicated server), is a KVM virtual machine whose vCPUs are not shared with other customers. A normal VPS gets a slice of time on a physical core that other virtual machines also use. On a VDS the cores you pay for are kept for your server, so the CPU time you measure at 03:00 is the CPU time you get at 20:00.

RS Computers LLC sells three such plans from its own network, AS213900, in Amsterdam (Netherlands), Dublin (Ireland) and Prishtina (Kosovo), at the same price in each city:

• VDS Small: 4 dedicated vCPU, 8 GB RAM, 240 GB NVMe, 10 Gb/s port, 50 TB of traffic a month, €40 a month
• VDS Medium: 8 dedicated vCPU, 16 GB RAM, 480 GB NVMe, 10 Gb/s port, 75 TB of traffic a month, €80 a month
• VDS Large: 16 dedicated vCPU, 32 GB RAM, 960 GB NVMe, 10 Gb/s port, 100 TB of traffic a month, €160 a month

Every plan comes with full root access, a dedicated IPv4 address plus IPv6, free weekly backups that you restore yourself with one click, and automatic deployment in under 60 seconds after payment. If you are still deciding between a VPS, a VDS and a physical machine, read our VPS vs VDS vs dedicated server comparison first; this page is about the VDS plans only.

Shared vCPU and dedicated vCPU: the real difference

A physical server has a fixed number of CPU cores. The hypervisor, KVM in our case, hands those cores out to virtual machines as vCPUs. With shared vCPUs, the host sells more vCPUs than it has cores, on the sound assumption that most servers sit idle most of the time. With dedicated vCPUs, it does not.

Sharing works well until several neighbours get busy at once. Then your server wants the CPU, the host gives that moment to someone else, and Linux records the wait as steal time. The kernel documentation defines it as time spent in other operating systems when running in a virtualised environment (proc_stat(5)). People call the cause a noisy neighbour. From inside your server it looks like an application that became slow for no reason: nothing in your logs, no change on your side.

When a shared vCPU is the right choice

Our VPS Nano, VPS Micro and VPS Mini plans use shared vCPUs, and that is correct for what they run: small websites, a WireGuard VPN, bots, development and staging servers, a mail server for a few mailboxes. These workloads use the CPU in short bursts and wait for the network or the disk the rest of the time. Paying for reserved cores there buys nothing you would notice.

When you need dedicated cores

• Databases. PostgreSQL and MySQL answer in milliseconds, so a few milliseconds of waiting for a core show up directly in the slowest requests.
• Build servers and CI runners, which hold every core at 100% for minutes at a time.
• Game servers, where one late tick is a visible stutter for every player.
• Trading bots and other latency-sensitive jobs that must react at a fixed moment, not when a core is next free.
• LLM inference on CPU, which uses all cores for the whole length of every answer.

We have seen the effect in our own lab. In the test behind our VDS sizing guide for a database and API (29 October 2026, PostgreSQL 17 with pgbench and wrk in a container on our Amsterdam lab host), one run happened to land while the host was busy: the 99th percentile response time at 50 clients went from 60 ms to 217 ms with nothing changed on our side. That lab was a container, not a VDS, so it is not a measurement of these plans. It shows what contention for CPU does to the slowest requests.

How to measure CPU steal yourself

You do not have to take any host's word for it, ours included. These three commands work on Debian 13 and Ubuntu 24.04. Run them while your own CPU is busy, because an idle server asks for nothing and so loses nothing.

# as a normal user: live view, press 1 for one line per vCPU, q to quit
top

# as a normal user: one line per second, ten lines
vmstat 1 10

# as root, once: mpstat is part of the sysstat package
apt install sysstat

# as a normal user: every vCPU, averaged over 5 seconds
mpstat -P ALL 5 1

In top, look at the %Cpu(s) line near the top: the last value, marked st, is steal. In vmstat, read the st column at the far right and ignore the first line, which is an average since boot. In mpstat, read %steal. This is the output we published on 9 October 2026 from a KVM server of ours in Amsterdam with 8 vCPUs and 16 GB of RAM, the same size as VDS Medium, while sysbench held all 8 vCPUs at full load:

# output, vmstat under load (shortened)
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
 9  0      0 3917904      0 249804    0    0     0     0 2166  242 100  0  0  0  0
 8  0      0 3917908      0 249804    0    0     0     0 2135  112 100  0  0  0  0
# output, mpstat
Average:     CPU    %usr   %nice    %sys %iowait    %irq   %soft  %steal  %guest  %gnice   %idle
Average:     all   99.95    0.00    0.02    0.00    0.00    0.02    0.00    0.00    0.00    0.00

That is 0% steal with every vCPU at 100%. The host was quiet during that test, so treat it as one sample and not as a promise. The full method, with disk and network checks, is in our VPS benchmark guide. As a rule of thumb, 0 to 1% most of the time is normal on a shared vCPU and steal above 5% for long stretches points to an oversubscribed host (SSD Nodes benchmarking guide). If your current server shows that pattern, our guide to a slow VPS helps you rule out your own configuration before you move.

Dedicated vCPU plans compared: RS Computers, Hetzner CCX, Contabo VDS, netcup and DigitalOcean

We checked each provider's own pricing page on 10 November 2026 and picked the plans closest to 4 vCPU with 8 GB and 8 vCPU with 16 GB. Prices are in the provider's currency and with the VAT treatment its page shows. Terms differ, so read the price column carefully: some figures need a 12 or 24 month commitment.

Read fairly, the table says this:

• netcup is cheaper than us at both sizes, at roughly half our price on a 12-month term, with about half the disk space of the matching RS Computers plan.
• Contabo gives more RAM for the money. Cloud VDS M has 8 cores and 32 GB for less than our VDS Medium if you commit for 24 months. It has half the disk of VDS Medium and a 500 Mbit/s port.
• Hetzner CCX has twice our RAM at each core count and costs more: CCX23 is about twice the price of VDS Small, and CCX33 is well above VDS Medium, both before VAT. Its page prices backups separately.
• DigitalOcean is the most expensive of the group for the same cores and RAM, with the smallest disks, and charges 20% or 30% extra for backups.

What RS Computers adds at its price: a 10 Gb/s port on every VDS, 50 to 100 TB of traffic with no overage fees, free weekly backups with one-click restore, monthly billing with no long commitment and no setup fee, and servers in Amsterdam, Dublin and Prishtina. If you need the lowest price per core and can commit for a year, netcup is hard to beat. If you need a fast port, more disk per plan, a month-to-month bill or a server in Ireland or Kosovo, compare our plan cards at the top of this page.

We left out two providers. Vultr sells dedicated-vCPU plans, but its pricing pages refused our automated check on 10 November 2026, so we print no figures for it. OVHcloud's VPS page does not say whether its vCores are dedicated, so it does not belong in this table.

What the 10 Gb/s port and the traffic allowance mean in practice

The port is the top speed of your server's network connection. The allowance is how much data you can move in a month at that speed. Our three VDS plans include 50, 75 and 100 TB a month.

These are plain arithmetic for a 30-day month (1 TB = 1,000 GB). The point of a 10 Gb/s port is not to run flat out all month. It is that a database dump, a container image pull, a backup to another city or a traffic spike finishes in seconds or minutes, and then the line is quiet again. In our published test of 9 October 2026, iperf3 from a server of ours in Amsterdam reached 4.9 to 6.5 Gbit/s to public test servers in Amsterdam and about 3.2 Gbit/s to London; what you see depends on the server at the other end as much as on ours.

There are no overage fees. If a server uses its whole allowance before the month ends, it stays online at a reduced speed until the next month starts, and the invoice does not change.

Who a dedicated vCPU server is for

• Production databases and APIs. Our VDS sizing guide works through cores, RAM and connection pools with pgbench, and the database setup guide covers PostgreSQL, MySQL and Redis.
• Busy WooCommerce shops. Cart and checkout pages are never page-cached, so PHP workers need cores that are there when a buyer clicks. See the e-commerce server guide.
• CI runners and build servers. A compile job that owns its cores takes the same time on every run. Start with a staging server with your own CI runner.
• Game servers. FiveM on a VDS and the wider game server guide explain why tick rate depends on steady single-core speed.
• n8n in queue mode. Several workers pulling jobs from Redis keep cores busy all day; the layout is in our n8n queue mode guide.
• Self-hosted AI apps. Open WebUI, Dify and small models on CPU want every core for as long as an answer takes. Read running an LLM without a GPU and Open WebUI on a server.

Start on a VPS and upgrade to a VDS later

You do not have to guess the size on day one. You can start on VPS Mini (4 vCPU, 4 GB RAM, 80 GB NVMe, shared vCPUs) and move up when the numbers say so: steal time during busy hours, a database that no longer fits in memory, or a disk that is filling up. VDS Small keeps the same 4 vCPU count, makes the cores dedicated, doubles the memory to 8 GB, triples the disk to 240 GB NVMe and moves the port from 1 Gb/s to 10 Gb/s.

The upgrade is done from the client area: choose the bigger plan, and the change, including the reboot, takes about a minute. Two limits to know: an upgrade goes to a bigger plan, and it stays in the same location, so a server in Dublin is upgraded in Dublin. Plan the minute of downtime for a quiet hour.

Payment, company and guarantees

RS Computers LLC was founded in Kosovo in 2018 and is registered there under business number 810142307, with its office at Rruga Ali Vitia 345, 10000 Prishtina. We operate our own network, AS213900, and are a RIPE NCC member. Every VDS is covered by a 99% monthly uptime SLA with service credits, and new customers can ask for a refund within the first 7 days under our Refund Policy.

You pay with crypto at checkout, in Bitcoin, Ethereum, USDT, Monero, BNB, TRON or another widely used coin, or by bank transfer using the bank details printed on your invoice. Prices are fixed in euros either way. The steps and the common mistakes are on our crypto VPS page. Questions before ordering go to Telegram or email, and a person answers.