Choosing a VPS provider used to mean reading marketing claims and crossing your fingers. Not anymore. I ran standardized benchmarks across eight popular VPS hosts in September 2026—same tests, same instance sizes, same month—to measure what actually matters: disk speed, network throughput, CPU performance, and how fast support replies when something breaks.
This comparison focuses on entry-level and mid-tier plans because that's where most site owners and developers land. If you need bare-metal or dedicated hardware, you already know your requirements. For everyone else, here's what the tests showed.
Why these eight providers
I picked hosts that support tickets and forum threads mention most often: DigitalOcean, Linode (now Akamai), Vultr, Hetzner, OVHcloud, UpCloud, Contabo, and Kamatera. All offer hourly billing or monthly plans, Linux images, and root access. All support standard tools like fio, iperf3, and sysbench.
A few names are missing. AWS and Google Cloud live in a different budget universe. Smaller regional providers might be faster in your specific geography, but I focused on hosts with global data center footprints and English-language support.
Test methodology
Each provider was tested on a 2-vCPU, 4 GB RAM instance running Ubuntu 22.04 LTS. I used the same benchmark suite on each:
- Disk I/O:
fiowith random read/write patterns, 4K block size, 75% read / 25% write mix - Network throughput:
iperf3tests to three geographic endpoints (US East, EU West, Asia Pacific) - CPU performance:
sysbenchsingle-thread and multi-thread runs, prime number calculation to 20000 - Latency: ICMP ping tests from five monitoring locations
All tests ran between 2 AM and 6 AM UTC to minimize noisy-neighbor interference. Each test repeated three times; I kept the median result.
Disk I/O results
Disk speed matters more than most people realize. A slow disk turns a snappy WordPress site into a frustrating wait. It chokes database queries, delays log writes, and makes backups crawl.
The fastest disks came from UpCloud and Hetzner, both offering NVMe storage on their standard plans. DigitalOcean and Linode delivered solid mid-pack performance with SSD arrays. Contabo lagged noticeably—likely due to oversubscription or older hardware.
In support tickets I handled, slow disk I/O usually showed up as intermittent site hangs, MySQL slow query logs piling up, or backup scripts timing out. If your current VPS feels sluggish during traffic spikes, disk speed is often the culprit.
UpCloud's MaxIOPS plans push disk performance even higher, but you pay extra. For most workloads, the standard NVMe offerings from UpCloud or Hetzner provide plenty of headroom.
Network throughput and latency
Network speed determines how fast your users load pages and how quickly you can push updates or backups off-site.
Vultr and DigitalOcean topped the throughput tests, especially on transatlantic routes. Both provision generous bandwidth allocations and connect to high-quality transit providers. Linode matched them closely. Hetzner delivered excellent speeds within Europe but showed slightly higher latency to Asia-Pacific endpoints—no surprise given their EU focus.
Kamatera's network was the most variable. Some tests hit advertised speeds; others dropped packets or showed high jitter. That inconsistency is a red flag if you run real-time applications or video streaming.
Contabo's network was slower but stable. If you mostly serve a regional audience and price matters more than raw speed, that tradeoff might work.
CPU performance
CPU benchmarks measure how fast your VPS can crunch numbers, compile code, or handle PHP requests. Single-threaded performance matters for most web apps; multi-threaded helps with video encoding, build pipelines, and parallel batch jobs.
Linode and DigitalOcean used AMD EPYC processors across most regions, delivering strong single-thread scores. Vultr mixed Intel and AMD depending on location; performance was solid either way. Hetzner's EPYC instances punched above their price point—best value in this category.
UpCloud and OVHcloud landed in the middle. Both offered consistent performance but didn't stand out. Contabo again trailed the pack, likely due to higher oversubscription ratios. Kamatera's results varied by data center; their newer US locations were faster than older EU ones.
One note: CPU steal time was lowest on Hetzner and DigitalOcean, meaning fewer noisy neighbors stealing cycles. Contabo showed occasional steal spikes, which can cause unpredictable slowdowns.
Pricing breakdown
Price per month for a 2-vCPU, 4 GB RAM instance with 80 GB SSD storage varied widely. Hetzner and Contabo came in cheapest, often half the cost of DigitalOcean or Linode. UpCloud sat at the higher end but included features like automated backups and flexible resource scaling.
Hourly billing gives you flexibility to spin up test instances or handle short-term traffic spikes without paying for a full month. DigitalOcean, Linode, Vultr, and UpCloud all offer this. Hetzner and Contabo bill monthly, which locks you in but costs less.
Bandwidth caps matter. Most providers include 1-2 TB per month on entry plans. Hetzner includes 20 TB, which is absurd for the price. DigitalOcean and Linode charge overage fees; OVHcloud throttles speed instead of billing extra.
Snapshot and backup costs add up. Some hosts include automated backups (UpCloud, Kamatera); others charge 20% extra (DigitalOcean, Linode). Check the fine print.
Support quality
When your site goes down at 3 AM, support response time is the only metric that matters.
I opened identical tickets with each provider asking about PHP memory limits and custom kernel modules. DigitalOcean and Linode replied within 15 minutes with accurate answers. Vultr took about 30 minutes. Hetzner's first reply came in under an hour, but the answer was thorough and included config examples.
OVHcloud and Contabo took 4-6 hours for a first reply. Kamatera's support was hit-or-miss: sometimes fast, sometimes a full day wait. UpCloud responded quickly but their answers occasionally missed the technical depth I needed.
All eight providers offer community forums or knowledge bases. DigitalOcean's documentation is the gold standard—clear, current, and usually the first Google result. Linode's docs are nearly as good. Hetzner's community wiki is excellent but less polished. Contabo's docs are sparse; expect to search elsewhere.
Control panel and API
Every provider here offers a web dashboard for spinning up instances, managing DNS, and viewing graphs. DigitalOcean and Vultr have the cleanest UIs. Hetzner's dashboard feels a bit dated but gets the job done. Contabo's interface is clunky; simple tasks take more clicks than they should.
API quality matters if you automate deployments or integrate with CI/CD pipelines. DigitalOcean, Linode, and Vultr have well-documented REST APIs with client libraries in every major language. Hetzner's API is solid but less widely supported by third-party tools. Contabo's API exists but expect rough edges.
All eight support Terraform providers, which is what I use for infrastructure-as-code. DigitalOcean and Vultr work smoothly; Kamatera required workarounds.
Data center locations
If your audience is regional, choose a provider with a data center nearby. Latency matters more than raw bandwidth for interactive sites.
DigitalOcean, Linode, and Vultr cover the most geographic regions—North America, Europe, Asia, and increasingly Latin America and Africa. Hetzner focuses on Europe (Germany, Finland) and recently added US locations. OVHcloud has strong European and Canadian presence. Contabo operates mostly from Germany with a few other spots. UpCloud and Kamatera have smaller footprints but hit key metro areas.
Check latency from your target users before choosing a location. A VPS in Frankfurt won't serve Tokyo visitors as well as one in Singapore, no matter how fast the disk.
So which provider should you pick?
It depends on your priorities. Here's how I'd break it down:
Best overall performance: UpCloud or DigitalOcean. Both deliver fast disks, strong network throughput, and responsive support. You pay extra, but the consistency is worth it for production sites.
Best price-to-performance ratio: Hetzner. Fast NVMe storage, modern CPUs, and absurd bandwidth allocations at half the cost of competitors. Support is slower but competent. If you can handle occasional troubleshooting yourself, this is the pick.
Best for beginners: DigitalOcean or Linode. Excellent documentation, active communities, and support that actually helps. The extra few dollars per month buy peace of mind when you're learning.
Best for budget-conscious projects: Contabo. Performance lags behind the pack, but if you're running low-traffic dev sites or personal projects, the price is hard to beat. Just don't expect blazing speed or instant support.
Best for API automation: Vultr. Solid performance, clean API, good geographic coverage. Sits in the sweet spot between price and features.
Avoid if: You need guaranteed uptime or real-time support. None of these providers hit enterprise-grade SLAs. For that, you need managed hosting or cloud platforms with paid support tiers.
What to watch out for
A few gotchas across the board:
Noisy neighbors: Shared virtualization means other tenants can impact your performance. CPU steal time and inconsistent disk I/O are the usual symptoms. Hetzner and DigitalOcean showed the least noise; Contabo and Kamatera the most.
Bandwidth throttling: Some providers throttle outbound speed after you hit your monthly cap instead of billing overages. Read the terms.
IPv4 costs: IPv4 addresses increasingly cost extra. Most providers include one with your instance, but additional IPs now run $3-5 per month. Plan accordingly if you run multiple SSL sites without SNI.
Backup restore time: Automated backups are great until you need to restore one. Test the restore process before you rely on it. I've seen backup restores take 30 minutes on DigitalOcean and three hours on Contabo for the same data size.
DDoS protection: Basic mitigation is included everywhere, but large attacks will get your IP null-routed. If you're a frequent target, add Cloudflare or a dedicated DDoS mitigation service in front.
Testing your own VPS
Don't take my word for it. Run your own benchmarks:
# Install tools
sudo apt update && sudo apt install -y fio iperf3 sysbench
# Disk I/O test (random read/write mix)
sudo fio --name=randrw --ioengine=libaio --iodepth=16 --rw=randrw \
--bs=4k --direct=1 --size=1G --numjobs=4 --runtime=60 --group_reporting
# CPU single-thread
sysbench cpu --threads=1 --time=60 run
# CPU multi-thread
sysbench cpu --threads=$(nproc) --time=60 run
# Network test (requires iperf3 server on another host)
iperf3 -c <remote-host> -t 30
Run tests multiple times and at different hours. A VPS that flies at 3 AM might crawl at 3 PM if the host oversells capacity.
What matters most
Speed benchmarks tell part of the story, but not all of it. A VPS that scores highest on synthetic tests but goes offline twice a month is worse than a slightly slower host that stays up.
Pick based on your actual workload. A static site needs fast disks and good bandwidth but minimal CPU. A build server needs strong multi-thread performance. A database needs low latency and consistent I/O.
Test the provider's support before you commit. Open a ticket, ask a technical question, see how long you wait and whether the answer makes sense. That tells you more than any benchmark.
Price matters, but the cheapest option often costs you in other ways—slow support, oversold hardware, surprise fees. A few extra dollars per month for better performance and reliable support is usually worth it.
Your hosting foundation supports everything else you build. Choose carefully.