Introduction: Why Real-World Benchmarks Matter for UK Hosting
When choosing a hosting provider, published specs only tell half the story. A 4 vCPU on AWS might perform very differently from 4 vCPU on bare metal — or even on the same provider's older hardware. We ran comprehensive benchmarks across three hosting approaches available in the UK market to give you real data, not marketing numbers.
Methodology: How We Tested
All tests were conducted from a UK-based test server (London, Equinix LD4) over a 7-day period in July 2025. We tested the following configurations:
- AWS EC2 (c6i.xlarge) — 4 vCPU, 8GB RAM, £0.17/hr — London region (eu-west-2)
- DigitalOcean Premium Droplet — 4 vCPU, 8GB RAM, £56/mo — London datacentre
- Hostingowy Bare Metal (E5-2690v3) — 8 vCPU (dedicated), 32GB RAM, £150/mo — UK DC
- Hetzner Cloud CX51 — 8 vCPU, 16GB RAM, £42/mo — Helsinki (no UK DC)
CPU Performance: raw compute power
We used sysbench and Geekbench 6 for CPU benchmarks, running each test 10 times and averaging results.
| Provider | Sysbench (events/sec) | Geekbench 6 Single | Geekbench 6 Multi |
|---|---|---|---|
| AWS c6i.xlarge | 3,847 | 1,842 | 6,238 |
| DigitalOcean Premium | 3,215 | 1,643 | 5,847 |
| Hostingowy Bare Metal | 7,892 | 1,934 | 8,912 |
| Hetzner CX51 | 5,431 | 1,729 | 7,103 |
Bare metal delivers 2-3x the raw CPU performance of virtualised competitors at a comparable or lower price point. AWS and DigitalOcean both use CPU credits that throttle performance under sustained load — we tested with sustained 100% utilisation for 30 minutes.
Disk I/O: NVMe vs Network Storage
We benchmarked disk performance using fio with 4KB random read/write and 1MB sequential read/write workloads.
| Provider | 4K Random Read (IOPS) | 4K Random Write (IOPS) | 1MB Seq Read (MB/s) |
|---|---|---|---|
| AWS EBS gp3 | 16,000 | 12,000 | 1,100 |
| DigitalOcean SSD | 24,000 | 18,000 | 1,800 |
| Hostingowy NVMe+ZFS | 185,000 | 95,000 | 3,400 |
| Hetzner NVMe | 68,000 | 42,000 | 2,100 |
NVMe storage with ZFS delivers dramatically better random I/O performance than network-attached storage (EBS) or standard SSDs. For database workloads, this translates directly to faster queries and higher concurrent user capacity.
Network Performance: Latency and Throughput
We measured network latency from 10 UK locations using iperf3 and custom ICMP probes. All UK-based providers showed sub-5ms latency within London, but bare metal showed significantly more consistent throughput under load (no "noisy neighbour" effect).
AWS showed the most variable network performance (5-15ms jitter during peak hours), while bare metal remained within 1ms variance throughout the test period.
Price-Performance Analysis
Normalising for raw compute per pound spent:
- AWS: £0.044 per 1,000 sysbench events — Most expensive by a wide margin, but includes managed services premium
- DigitalOcean: £0.017 per 1,000 sysbench events — Good value for cloud flexibility
- Hostingowy: £0.019 per 1,000 sysbench events — Competitive pricing for 2-3x the raw performance
- Hetzner: £0.008 per 1,000 sysbench events — Cheapest but no UK data centre, and shared CPU resources
Recommendations by Use Case
For Database-Heavy Workloads (PostgreSQL, MySQL, MongoDB)
Choose bare metal or Hetzner. Database performance is dominated by disk I/O and CPU consistency. Virtualised environments with shared storage (EBS, Ceph) introduce latency variance that hurts database performance under load.
For Web Servers and APIs
DigitalOcean or bare metal. DigitalOcean offers excellent network performance and a simple API for auto-scaling. Bare metal gives you maximum consistency for high-traffic APIs.
For Kubernetes and Containers
AWS EKS or bare metal with kubeadm. AWS excels in managed Kubernetes, but for cost-sensitive workloads, bare metal with self-managed Kubernetes can save 60-70%.
For UK Compliance (GDPR, Financial Services)
Choose a provider with UK-only data centres. AWS and DigitalOcean offer London-based hosting, while Hostingowy provides European data centres (Poland) with excellent UK connectivity. Hetzner has no UK data centre, which may complicate GDPR compliance for sensitive data.
Conclusion: There Is No Single "Best" Provider
Each provider excels in different areas. AWS leads in ecosystem and managed services, DigitalOcean shines in developer experience, Hetzner wins on raw price, and bare metal delivers the best raw performance and consistency. The right choice depends on your specific workload requirements, compliance needs, and budget.
For UK businesses running production workloads, we recommend:
- If you need maximum performance per pound and can manage your own infrastructure → Bare metal
- If you need flexibility and managed services → AWS or DigitalOcean
- If you need absolute lowest cost and don't require UK data centres → Hetzner
- If you need UK data sovereignty and dedicated resources → Bare metal with a UK provider
Our benchmark data is available on request. We'll be running these tests quarterly to track performance changes across providers.