NodeRings Docs

Dedicated CPU plans

Pinned physical cores with no sharing, plan categories, vCPU counting, and billing

Dedicated CPU plans (plan codes starting with DC-) give your VM host CPU cores that are pinned to your VM and never shared with, or oversubscribed by, any other VM. They sit alongside the existing Shared Compute (SC-*) plans and use the same purchase flow, VM management, and billing.

Choose dedicated when you need consistent CPU performance: databases, game servers, CI runners, latency-sensitive APIs, or any workload where a noisy neighbor is unacceptable.


Dedicated vs shared compute

Shared Compute (SC-*)Dedicated CPU (DC-*)
CPU allocationvCPUs scheduled by the hypervisor across shared host coresPhysical host cores pinned to your VM
OversubscriptionPossible; the provider can sell more vCPUs than physical threadsNever; each physical core belongs to exactly one VM
NeighborsOther VMs can compete for the same coresNo other VM ever runs on your cores
Host typeRuns on the provider's shared nodesRuns only on the provider's dedicated nodes
BillingFixed or usage-basedIdentical (see Billing)

On a dedicated plan, both SMT threads (hyper-threads) of every physical core you get are pinned to your VM. Nothing else — not even a sibling thread — is sold to anyone else.


Plan categories

Dedicated plans come in four categories, so you can match the vCPU : RAM ratio to your workload:

CategoryBest for
General PurposeBalanced vCPU and RAM for most workloads
CPU OptimizedCompute-heavy work: builds, encoding, game servers
Memory OptimizedRAM-heavy work: databases, caches, in-memory analytics
Storage OptimizedDisk-heavy work: large datasets, object stores

Plan codes follow the same scheme as shared plans: TIER-CATEGORY-VCPUsC-RAMGB-DISKGB (for example DC-CPU-8C-8G-320 = Dedicated Compute, CPU Optimized, 8 vCPU, 8 GB RAM, 320 GB disk).


How vCPUs are counted

A plan's vCPU count is the number of logical threads your VM gets — the same unit shared plans use. Under the hood, allocation happens in whole physical cores:

  • On hosts with SMT (hyper-threading), one physical core has 2 threads. An 8 vCPU dedicated plan gets 4 physical cores — all 8 threads pinned to your VM.
  • Both threads of a core always go to the same VM. A physical core is never split between two customers.
  • On hosts with SMT disabled, 1 vCPU equals 1 physical core.

For example, a host built on an AMD EPYC 9654 has 96 physical cores and, with SMT enabled, 192 hardware threads — NodeRings exposes it as 192 vCPUs of dedicated capacity. A DC-CPU-8C plan on that host pins 4 of its physical cores (8 threads) exclusively to your VM.

This is the standard industry approach for dedicated instances: threads are the unit you buy, cores are the unit of isolation.


Create a dedicated VM

The flow is the same as for shared plans, at /platform/virtual-machines/create.

Pick the Dedicated CPU tier

On the plan selector, choose Dedicated CPU instead of Shared Compute. The tier tab shows the dedicated categories and plans available in your chosen region or from your chosen provider.

Choose category and plan

Pick a category (General Purpose, CPU Optimized, Memory Optimized, or Storage Optimized) and a plan size. Plan rows show the same details as shared plans: code, vCPU, RAM, storage, and price.

Choose OS and buy

Select the operating system, authentication method, and hostname, then confirm the launch summary — exactly as described in Virtual Machines → Create a VM.

NodeRings automatically places the VM on one of the provider's dedicated nodes with enough free cores. Everything after purchase — power actions, console, backups, snapshots — works the same as any other VM.


Billing

Dedicated plans bill exactly like shared plans:

ModelHow you pay
FixedUpfront monthly flat rate (subscription)
Usage-basedHourly while the VM exists, with a monthly cap on base compute

Bandwidth egress is metered separately with a free tier, and addons (backups, snapshots, additional IPs) attach the same way. See Billing for details.

A stopped dedicated VM still owns its pinned cores — they are a reservation, not runtime usage. Delete the VM to release the cores and stop billing.


Next