VPSbit benchmarks - Geekbench, fio, iperf by location

Geekbench-class CPU, fio NVMe and iperf port readings taken from our own fleet, per datacenter - with the method, so every number is reproducible on your instance.

Published 2026-08-30 · Updated 2026-10-06 · VPSbit Editorial

12 locations · 19 coins · SLA 99.95% · no KYC · public canary

Method

Readings come from the shared-EPYC hosts that carry our entry KVM tiers: ECC memory, local NVMe, 1 Gbps port, no bursting tricks behind the numbers. Every figure below is a typical range, not a sales peak, and the honest move is to re-run the same commands on your own instance after deploy. Last refresh: 2026-09-14.

Reading the spread between cities

Single-core scores cluster because every location runs the same host generation; a two-percent swing between flags is scheduler noise, not a reason to move. What a table cannot show is contention on a given evening - shared vCPU means neighbors borrow the same silicon. Workloads that live on the 95th percentile, like busy databases, build farms or game ticks, belong on the dedicated ladder, where whole cores are pinned and IPMI sits out-of-band. Everything else runs fine on a KVM slice from $4.80 to $6.

CPU (Geekbench 6 class)

Disk (fio) and network (iperf)

The fio rows assume the local NVMe every KVM instance carries - no network-attached compromise - and the iperf row is what a 1 Gbps unmetered port sustains long-term. Boxes whose whole job is throughput per terabyte should compare Titan: 320 GB of the same local NVMe behind a 10 Gbps port at $24/mo list. Deploy Core from $4.80 and verify on your node. Token-only signup available.