Shared vCPU vs Dedicated vCPU
Shared vs dedicated vCPU: what steal time costs you
A shared vCPU is scheduled against other tenants, so under contention your process waits — visible as steal time in top. A dedicated vCPU is pinned to a physical core reserved for you. For bursty workloads the difference is invisible; for anything latency-sensitive, tick-based or continuously busy, it is the single most important specification on the page.
| Dimension | Shared vCPU | Dedicated vCPU |
|---|---|---|
| Steal time | Possible under contention | Effectively zero |
| Sustained load | May be throttled | Unrestricted |
| p99 latency | Variable | Consistent |
| Price | Lower | ~40–60% higher |
| Good for | VPN, DNS, small bots, dev | Databases, games, trading, APIs |
Which should you choose?
Shared vCPU
Shared cores are genuinely fine for idle-most-of-the-time workloads — a VPN endpoint does not care.
Dedicated vCPU
Pay for dedicated cores whenever tail latency or sustained throughput matters to someone other than you.
Frequently asked questions
How do I check steal time?
The st column in top or vmstat. Anything consistently above 2–3% means you are competing for the core and should move to dedicated vCPU.
Is a dedicated vCPU a whole physical core?
On our platform it is a reserved physical thread with no other tenant scheduled against it. Terminology varies by provider — always ask what "dedicated" means before comparing prices.