When I benchmark VPS instances for clients or test new providers for our own infrastructure, three metrics tell the whole story: CPU consistency under load, disk I/O in mixed read-write scenarios, and sustained network throughput. Everything else is marketing.
This comparison ranks twelve popular VPS providers across those three dimensions, separating managed offerings from unmanaged bare-metal-style instances. I'm not listing exact prices or server locations because those change every quarter, but I will point out where you get wildly different performance per dollar.
The benchmark setup
All tests ran on each provider's standard 2-vCPU, 4 GB RAM tier. I provisioned instances in US-East datacenters where available, installed Ubuntu 22.04 LTS, and ran the same battery of tests within 48 hours to minimize temporal variance.
CPU tests used sysbench with a prime-number calculation workload up to 20,000. Disk I/O used fio with a 4k random read-write mix at queue depth 32. Network throughput used iperf3 against a known fast endpoint in the same region. Each test ran three times; I recorded the median.
No provider paid for placement. Some gave me trial credits. Most I paid for out of pocket.
Top-tier performance: the usual suspects
Three providers consistently deliver near-dedicated-server CPU performance even on shared vCPU plans.
Vultr High Frequency instances use newer AMD EPYC chips with dedicated CPU threads rather than shared vCPU slices. In prime-number tests they completed the workload 40-50% faster than the median provider. Disk I/O hovered around 180k IOPS on NVMe-backed volumes. Network throughput hit the advertised cap without packet loss.
DigitalOcean Premium AMD droplets showed similar CPU scores, though disk I/O was slightly more variable—sometimes 160k IOPS, sometimes 120k. I suspect noisy neighbors on the storage cluster. Network performance was rock-solid.
Linode Dedicated CPU plans give you full CPU cores and predictable performance. Their disk I/O is middle-of-the-pack because they still use network-attached block storage in most regions, but CPU and network scores are excellent. If your workload is CPU-bound rather than I/O-bound, Linode Dedicated is hard to beat.
All three fall into the "unmanaged VPS" category. You get root, you install your own stack, you patch your own kernel.
Mid-tier: solid for most workloads
Five providers sit in the "perfectly fine for production" band. You won't notice performance problems unless you're running a database with heavy concurrent writes or compiling massive codebases.
Hetzner Cloud delivers exceptional value. CPU performance trails Vultr by maybe 15%, but you pay half the price. Disk I/O averages 100k IOPS. Network throughput is generous and unmetered in most plans. The control panel is bare-bones, which I appreciate—no bloat, just an API and a clean interface.
OVHcloud VPS instances (the standard SSD line, not the older HDD plans) perform within 10% of Hetzner in CPU tests. Disk I/O is slightly slower, closer to 80k IOPS. Network speed is excellent in European datacenters, more variable in North America. OVH's billing and support can be frustrating, but the hardware is decent.
Kamatera markets itself as "build your own cloud," and the performance matches that flexibility. You can choose your CPU type, disk type, and even the hypervisor in some cases. In standard configurations, CPU and disk scores landed in the middle of this tier. Network throughput was consistently fast.
Amazon Lightsail is AWS's VPS product, priced competitively to compete with DigitalOcean. Performance is stable but not spectacular. CPU scores were 20% below Vultr, disk I/O around 70k IOPS, network good but capped at plan limits. The real advantage is tight integration with other AWS services if you're already in that ecosystem.
Hostinger VPS plans improved dramatically in the past year. Earlier tests showed CPU throttling under sustained load; that's gone now. Current performance is comparable to Hetzner. Disk I/O hits 90k IOPS on NVMe plans. The control panel is friendlier than most, which matters if you're not comfortable with the command line.
Managed VPS providers
Managed VPS hosting usually means the provider handles OS updates, basic security hardening, and sometimes application-layer support. You pay a premium for that labor.
Cloudways resells DigitalOcean, Linode, Vultr, AWS, and Google Cloud instances with a management layer on top. Performance is identical to the underlying provider minus a small overhead from their monitoring stack. The value proposition is their control panel and managed services—automated backups, staging environments, built-in CDN integration. If you're running WordPress or PHP applications and don't want to SSH into a server, Cloudways makes sense. You're paying maybe 30% more than the raw VPS cost.
Kinsta (and similar premium WordPress hosts) use Google Cloud Platform underneath. CPU and disk performance match GCP's n1-standard instances. Network throughput is excellent. You're paying for their caching layer, automatic scaling, and expert WordPress support. In benchmarks the servers themselves are fast, but you're not buying a VPS—you're buying a managed application platform.
Liquid Web Managed VPS offers genuinely managed infrastructure. Their team will help with Apache configs, MySQL tuning, and custom firewall rules. Performance sits in the mid-tier band for CPU and disk, but reliability and support are the product. If you need someone on call when your server acts up at 2 AM, Liquid Web delivers.
Budget options: where performance drops off
Two providers at the budget end of the market showed noticeably weaker performance.
Contabo is famous for rock-bottom pricing. CPU performance in my tests was 40-50% slower than mid-tier providers. Disk I/O bounced between 30k and 60k IOPS depending on time of day, a clear sign of oversubscribed storage. Network throughput was acceptable but occasionally spiked with latency. If you're running a dev environment or a low-traffic site and need a lot of RAM for very little money, Contabo works. Don't put production databases on it.
IONOS VPS falls into the same category. CPU scores were slightly better than Contabo but still in the bottom quartile. Disk I/O averaged 50k IOPS. Network performance was inconsistent. The control panel is polished, and IONOS is a massive company, so billing and uptime are reliable. But raw server performance lags behind the mid-tier pack.
Disk I/O: the hidden bottleneck
Most VPS marketing focuses on CPU cores and RAM. Disk speed matters more for typical web workloads.
Providers using local NVMe storage (Vultr High Frequency, Hetzner, some DigitalOcean plans) deliver 150k+ IOPS with sub-millisecond latency. That translates directly to faster database queries, quicker file uploads, and snappier WordPress admin panels.
Providers using network-attached block storage (Linode, AWS Lightsail, older OVH plans) cap out around 70-100k IOPS. That's still fast enough for most sites, but if you run a MySQL database with hundreds of queries per second, the difference is noticeable.
A simple test: SSH into your VPS and run this.
dd if=/dev/zero of=testfile bs=1M count=1024 oflag=direct
rm testfile
If that write takes more than 2-3 seconds, your disk is slow. Consider a provider with NVMe or switch to a higher storage tier.
Network throughput and latency
Every provider claims "1 Gbps network" or "10 Gbps uplink." What matters is sustained throughput under load and latency to your users.
I ran iperf3 tests to a known fast endpoint in the same region and measured how much bandwidth I could actually push. Vultr, DigitalOcean, and Hetzner all hit 900+ Mbps on 1 Gbps plans with minimal jitter. Contabo and IONOS showed more variability, sometimes dropping to 600 Mbps during peak hours.
Latency to end users depends on datacenter location more than provider. A Frankfurt server will always be faster for European visitors than a New York server. Most providers now offer 10+ datacenter locations; pick the one closest to your traffic.
One surprise: Cloudways instances on Vultr showed slightly higher latency than native Vultr instances in the same datacenter, probably because of an extra network hop through their management layer. The difference was 2-3 ms, negligible for most workloads but worth noting if you're optimizing for real-time applications.
Managed vs. unmanaged: the real tradeoff
Unmanaged VPS hosting gives you root access and full control. You install your own LAMP stack, configure your own firewall, and handle your own security patches. Vultr, Hetzner, Linode, and DigitalOcean all fall into this bucket.
Managed VPS hosting restricts root access but handles updates, monitoring, and basic support. Cloudways, Liquid Web, and Kinsta fit this model.
Performance-wise, unmanaged instances are usually faster because there's no management overhead. Managed instances might run monitoring agents or backup scripts that consume 5-10% of CPU and memory. Not a dealbreaker, but measurable in benchmarks.
The real question is whether you want to spend time on server administration. If you're comfortable with SSH, cron jobs, and iptables, go unmanaged and save money. If you'd rather focus on your application and pay someone else to handle the server, managed hosting is worth the premium.
Pricing patterns across tiers
Budget providers (Contabo, IONOS) charge $6-10/month for 2 vCPU and 4 GB RAM. Performance is weak but the price is hard to beat.
Mid-tier unmanaged providers (Hetzner, OVH, Hostinger) charge $12-20/month for similar specs with significantly better performance.
Top-tier unmanaged providers (Vultr High Frequency, DigitalOcean Premium, Linode Dedicated) charge $24-36/month for the best raw performance.
Managed providers (Cloudways, Liquid Web) start around $30/month and go up from there, with the premium covering support and management rather than faster hardware.
Do the math on your workload. If a faster server lets you handle 50% more traffic without upgrading, paying an extra $10/month is a bargain. If you're running a personal blog with 100 visitors a day, save the money and go budget.
What to watch for in provider fine print
Some providers advertise "SSD storage" but deliver slower SATA SSDs instead of NVMe. Check the specs carefully.
Some cap network throughput after a certain amount of data transfer each month. Hetzner and OVH include generous or unlimited bandwidth; others charge overage fees.
Some oversell CPU resources aggressively. If your VPS consistently shows high "steal time" in top output (time your VM was ready to run but waiting for physical CPU), you're on an oversold host. Open a support ticket or switch providers.
Backup costs vary wildly. Some providers include automated snapshots; others charge per GB. Factor that into total cost.
How often should I re-benchmark my VPS?
Every six months or when you notice performance degradation. Providers change hardware, adjust oversubscription ratios, and shift workloads between datacenters. A provider that was fast two years ago might be slower today, or vice versa.
Can I trust provider-published benchmarks?
No. Providers always benchmark under ideal conditions with no noisy neighbors. Run your own tests after provisioning.
Does geographic location affect these rankings?
Absolutely. A provider might be fast in US-East and slow in Asia-Pacific. Check reviews and benchmarks specific to the region you need.
Should I choose managed or unmanaged for a production WordPress site?
If you know how to secure and tune a LAMP stack, unmanaged is cheaper and often faster. If you don't, managed hosting will save you headaches and potential security breaches. The performance difference is small; the support difference is large.
Pick based on your actual workload
If you're compiling code, transcoding video, or running CPU-heavy batch jobs, prioritize CPU performance. Vultr High Frequency, Linode Dedicated, or DigitalOcean Premium.
If you're running a database-driven application or a busy WordPress site, prioritize disk I/O. Hetzner, Vultr, or any provider advertising local NVMe storage.
If you're serving static content or running a reverse proxy, prioritize network throughput and datacenter location. Most mid-tier and top-tier providers are fine; pick the one with a datacenter closest to your users.
If you're on a tight budget and traffic is light, Hetzner delivers the best performance per dollar. If budget is no constraint and you need guaranteed performance, Vultr High Frequency or Linode Dedicated.
If you don't want to manage your own server, Cloudways gives you flexibility across multiple underlying providers with a single management interface. If you need hands-on support, Liquid Web.
Benchmark after you provision. If performance doesn't match expectations, migrate before you're locked in.