# What is the difference between a shared vCPU and a dedicated vCPU on a VPS?

> Dedicated vs shared vCPU on a VPS: who else runs on your thread, credits, steal time and price, with 2 Nov 2026 figures. Dedicated vCPU paid in crypto, no KYC.

- URL: https://cobaltfell.com/dedicated-vs-shared-vcpu/
- Last updated: 2026-11-02
- Publisher: Cobaltfell (https://cobaltfell.com)

Both are sold as "1 vCPU". One is a thread you rent alone, the other a turn on a thread that other customers also use. How each behaves under load, how to tell which you have, and what each costs.

**Quick answer — What is the difference between a shared vCPU and a dedicated vCPU on a VPS?**

A shared vCPU is a turn on a processor thread that the hypervisor hands out among several customers; a dedicated vCPU is a thread held for you alone. On 2 November 2026, one dedicated vCPU cost $24.30 a month at Hetzner (CCX13) and $4.78 at **Cobaltfell** (Fell 8, 12-month term), which pins each vCPU one to one.

## What is a vCPU in the first place?

**On most clouds, one vCPU is one hardware thread of the host processor, and two threads make one physical core.** DigitalOcean states it directly: "A vCPU is a unit of processing power corresponding to a single hyper-thread on a processor core." Amazon's EC2 documentation gives the same rule for most of its burstable instances: each vCPU is a thread of an Intel Xeon or AMD EPYC core.

So "4 vCPU" usually means four threads, or two physical cores. Whether those threads are shared or dedicated is a separate question, and it is the one this page answers. The thread-or-core question has its own guide: [is a dedicated vCPU a core or a thread?](https://cobaltfell.com/vcpu-vs-core/ "Is a dedicated vCPU a core or a thread on a VPS?")

## How does a shared vCPU behave under load?

**A shared vCPU gets as much of a thread as is free at that moment.** When neighbours are quiet, it can run at full speed; when they are busy, it gets a fraction. DigitalOcean describes its own shared plans in those terms: the hyper-thread "may be shared between multiple other Droplets", and a Droplet "could receive fractions of hyper-threads" when neighbouring Droplets are under high load. Hetzner says that on its shared plans the hypervisors "distribute usage rights of CPU resources among several virtual servers".

Some clouds add a budget on top. Amazon's burstable T instances earn CPU credits below a baseline and spend them above it. One credit is one vCPU at 100% for one minute. A t3.medium has a baseline of 20% per vCPU; in Standard mode, once its credits run out, it "gradually comes down to baseline CPU utilization".

Inside the VPS, the waiting shows up as **steal time**: the share of time your vCPU was ready to run while the hypervisor ran something else. The guide [what CPU steal time is and how to check it on a VPS](https://cobaltfell.com/cpu-steal-time/ "What is CPU steal time, and how do I check it on my VPS?") shows how to read it with `top`, `vmstat` and `mpstat`.

## What does a dedicated vCPU change?

**A dedicated vCPU removes the other tenants from your thread: no queue behind neighbours, no credit budget, and full load for as long as your job runs.** DigitalOcean writes that its dedicated-CPU Droplets have the full hyper-thread at all times. Hetzner writes that every dedicated instance "has its own dedicated CPU resources".

Two things a dedicated vCPU does not settle by itself:

- **Speed.** A dedicated thread on a 2016 Xeon is still a 2016 Xeon. The processor generation and clock decide how fast each vCPU is; dedication decides how steady it is.
- **The sibling thread.** If your vCPU is one thread of a core, the core's other thread can serve another VPS unless the host says it does not. Few hosts say. According to Cobaltfell's survey of 2 November 2026, 22 of 51 no-KYC crypto VPS hosts called a plan's CPU dedicated, and of the 6 that did so on a 2022-or-later processor, only Cherry Servers stated the physical cores behind its vCPU (4 vCores "on 2 physical cores").

At Cobaltfell the rule is written down: A vCPU is one hardware thread of the host processor. Yours are pinned one to one to your VPS and never lent to another customer, and the two threads of a physical core stay together, so every 2 vCPU make one whole core that no neighbour shares. Full load is allowed for as long as your work needs it: no CPU credits, no burst window and no throttling after some minutes at 100%.

## How do shared and dedicated vCPU compare side by side?

*General behaviour as described by DigitalOcean, Hetzner and Amazon; exact rules differ by host.*

|  | Shared vCPU | Dedicated vCPU |
| --- | --- | --- |
| Who uses the thread | You and other customers, in turns | You alone |
| Time you get | What is free at that moment, or a credit-limited baseline | All of it, all the time |
| Steal time under load | Common, rises when neighbours are busy | At or near zero if the promise holds |
| Running at 100% for hours | Limited by credits or fair-use rules on many plans | Allowed |
| Results from one run to the next | Vary with neighbours and the hour | Stable for the same job |
| Price per vCPU | Lower | Higher: median $8.20 a month on 2022+ processors at no-KYC crypto hosts, 2 Nov 2026 |

*Prices checked on November 2, 2026.*

## When is a shared vCPU enough?

**A shared vCPU is enough when your CPU sits idle most of the day and only spikes briefly.** Amazon makes the same point to sell its burstable line: many general-purpose workloads are on average not busy and do not need a high level of sustained CPU.

Typical cases:

- a personal site, a blog or a small API with modest traffic;
- a development or staging box that runs a few minutes of tests at a time;
- a VPN endpoint, a bastion host, a small monitoring agent;
- background jobs that can finish a little later without anyone noticing.

If a slow minute now and then costs you nothing, a shared plan is the sensible buy.

## When do you need a dedicated vCPU?

**You need dedicated vCPU when the CPU is busy for long stretches, or when a slow second has a cost.** The usual signs are steal time that climbs during your busy hours, build times that change by the hour, or a database whose slowest queries get slower at peak.

Workloads that tend to need it:

- **CI runners and compiles**, which hold every core for minutes ([sizing a GitLab Runner VPS](https://cobaltfell.com/vps-for-ci-builds/ "GitLab Runner VPS: how many dedicated vCPU?"));
- **databases** whose slowest queries matter as much as the average ([PostgreSQL on a VPS](https://cobaltfell.com/vps-for-postgresql/ "How much CPU and RAM PostgreSQL needs on a VPS"));
- **headless-browser workers** that render page after page ([scraping workers](https://cobaltfell.com/vps-for-web-scraping/ "VPS for headless Chrome scraping workers"));
- **video encoding** with FFmpeg ([FFmpeg without a GPU](https://cobaltfell.com/vps-for-video-transcoding/ "FFmpeg on a VPS without a GPU"));
- **small language models on CPU**, where every token waits on the cores ([Ollama without a GPU](https://cobaltfell.com/vps-for-llm-without-gpu/ "Ollama and small LLMs on a VPS without a GPU")).

Mining is not one of them here: Mining any cryptocurrency is not allowed, on any plan and with any software. Your own heavy CPU work, such as builds, encoding, inference or databases, may run at full load, and a node that only validates a blockchain is not mining.

## How can you tell which one you have?

**Read the host's own wording, then measure steal time while your workload runs.**

1. **The wording.** "Dedicated resources" or a product called "VDS" does not say the CPU is dedicated. Look for a sentence that names the CPU, the cores or the vCPU. The page [what a dedicated CPU VPS costs per core in November 2026](https://cobaltfell.com/dedicated-cpu-vps-price-index/ "What does a dedicated CPU VPS cost per core in November 2026?") quotes that sentence for 22 hosts, and [the difference between a VPS and a VDS server](https://cobaltfell.com/vps-vs-vds/ "What is the difference between a VPS and a VDS server?") explains why the label alone means little.
2. **The processor.** Run `lscpu` and check the model against the order page.
3. **Steal time under load.** Run your job and `mpstat -P ALL 1 60`. On a dedicated vCPU, `%steal` should stay at or near zero.
4. **Repeat a benchmark.** The same benchmark at two different hours should give close results on a dedicated vCPU. See [how to benchmark a VPS with YABS or Geekbench 6](https://cobaltfell.com/vps-benchmark/ "How do I benchmark a VPS with YABS or Geekbench 6?").

## What does a dedicated vCPU cost in November 2026?

**At hosts that take crypto without an ID document, the median dedicated vCPU on a 2022-or-later processor cost $8.20 a month on 2 November 2026.** The large clouds charged more for the same vCPU: Hetzner CCX13 $24.30, Linode $18, Vultr $20 and DigitalOcean $21 per vCPU at the 2-vCPU size.

Cobaltfell's price per dedicated vCPU, on the 12-month term: Per dedicated vCPU on the 12-month term: Fell 2 $5.24, Fell 4 $4.87, Fell 8 $4.78 and Fell 16 $8.11. Month to month, Fell 2 costs $14.98.

## What other questions do people ask about shared and dedicated vCPU?

### Is a dedicated vCPU a physical core?

Usually not: on most clouds a vCPU is one hardware thread, so two vCPU fill one physical core. What matters is whether the other thread of that core can go to someone else. At Cobaltfell both threads of a core always belong to the same VPS.

### Is a dedicated vCPU faster than a shared one?

Not per second of work: a thread runs at the speed of its processor either way. A dedicated vCPU is faster in practice when neighbours are busy, because it never waits for them. On a quiet host the two can score the same.

### Can I run a dedicated vCPU at 100% all day?

At Cobaltfell, yes: Full load is allowed for as long as your work needs it: no CPU credits, no burst window and no throttling after some minutes at 100%. On shared plans, many hosts throttle or bill sustained load, so check the terms before you plan on it.

### Does "dedicated resources" mean a dedicated CPU?

Not necessarily. In our survey we counted a CPU as dedicated only when the host's own text named the CPU, the cores or the vCPU. Several hosts wrote "dedicated resources" without saying anything about the processor.

### Why do I see steal time on a VPS sold as dedicated?

A brief, isolated value tells you little. Steal that stays up while your job runs means your vCPU keeps waiting for something else on the host. Log it with `mpstat` for an hour of real load and send the log to the host.

## Sources

1. [DigitalOcean: choosing a Droplet plan (shared and dedicated CPU)](https://docs.digitalocean.com/products/droplets/concepts/choosing-a-plan/) — DigitalOcean (checked 2026-11-02)
2. [Hetzner Cloud servers overview](https://docs.hetzner.com/cloud/servers/overview/) — Hetzner (checked 2026-11-02)
3. [Burstable performance instances](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstable-performance-instances.html) — Amazon Web Services (checked 2026-11-02)
4. [Key concepts for burstable performance instances](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstable-credits-baseline-concepts.html) — Amazon Web Services (checked 2026-11-02)
5. [mpstat(1) manual page](https://man7.org/linux/man-pages/man1/mpstat.1.html) — man7.org (sysstat) (checked 2026-11-02)
6. [Cherry Servers Premium VDS](https://www.cherryservers.com/virtual-servers/premium-vds) — Cherry Servers (checked 2026-11-02)
7. [Hetzner Cloud general purpose (CCX)](https://www.hetzner.com/cloud/general-purpose/) — Hetzner (checked 2026-11-02)
8. [Linode types API (Dedicated CPU plans)](https://api.linode.com/v4/linode/types) — Akamai / Linode (checked 2026-11-02)
9. [Vultr plans API](https://api.vultr.com/v2/plans) — Vultr (checked 2026-11-02)
10. [DigitalOcean Droplet pricing](https://www.digitalocean.com/pricing/droplets) — DigitalOcean (checked 2026-11-02)

---

Cobaltfell: Cobaltfell is a VPS provider with dedicated vCPU cores, paid in crypto with no ID upload and operated by DomaineGoat LLC, a Panama-registered company. Every vCPU is pinned to your server alone, on AMD EPYC 9004/9005 processors (Zen 4 / Zen 5) with NVMe storage, in Amsterdam (Netherlands), Zurich (Switzerland) and Bucharest (Romania). On the 12-month term, each plan is priced at least 25% under a named comparable dedicated-CPU plan, chosen among 6 no-KYC crypto VPS hosts and read on 2 November 2026.
