VPS Benchmarks (2026): What Providers Publish, What This Site Measures, and How to Test a Server Yourself

This site measures nothing. It reads what each provider publishes about its servers (network caps, transfer allowances, disk types, IOPS limits where they exist), links the page or API it came from, and shows you how to measure the rest on a rented server in an hour. The three specification pages below are built that way; this page is the map.

✓ Published figures only
✓ Every number linked to its source
✓ Updated September 2026

The rule

A number appears on this site only if a reader can open the linked page or API and see the same number. A test one editor ran on one server on one afternoon does not meet that rule, so no such test is published. What providers do publish is thinner than a benchmark table but has the advantage of being checkable: caps, allowances, disk labels, the IOPS figures UpCloud publishes for its own storage tiers, and, for the hyperscalers, per-disk-tier IOPS and throughput limits.

Specification pages

Storage specs by provider

What each provider states about its disks (NVMe, SSD, RAID-10), the IOPS figures UpCloud publishes for its tiers and the per-disk-tier ones Azure and Google Cloud publish, and the fio command to measure your own.

Hetzner Cloud network, traffic and specs

Hetzner's own statements: shared 10 Gbit hosts, about 300 to 500 Mbit/s per server with no guarantee, 20 TB in Europe and 1 TB in the US, prices from its price API.

Azure VM specs and prices

B-series and D-series sizes with Microsoft's published vCPU, memory, disk IOPS and network figures, and consumption prices from the Azure Retail Prices API.

DigitalOcean Droplet specs

The five Droplet families and their ladders, three CPU options, the v5 processors and DigitalOcean's own "internal testing" qualifier, per-second billing, snapshots at $0.06/GB and backups at 20% or 30%.

Hostinger VPS specs

The four KVM plans with both the advertised and the renewal price, AMD EPYC, NVMe and 1 Gbps stated for the line, weekly backups with manual snapshots, and the paid-upfront term.

CloudCone VPS specs

Two lines out of one Los Angeles datacentre: SSD VPS from $2.33 a month billed $28 a year, and SC2 from $5.20 a month at $0.00699 an hour, with the non-refundable prepay clause that goes with the yearly one.

Contabo VPS specs

The Cloud VPS ladder, port speeds from 200 Mbit/s to 1 Gbit/s with the fair-usage note, snapshot counts per plan, and the euro-denominated pricing with the exchange-rate caveat.

IONOS VPS specs

Five VPS+ plans with the introductory and regular prices side by side, NVMe sizes, unlimited traffic, and the one-year term the promotional rate depends on.

Hostwinds specs

Unmanaged Linux from $4.99 with the support line Hostwinds prints, managed rows that show two prices, the 3 Mbps overage cap, and the 72-hour refund.

Cloudways specs

Five clouds, the Flexible Micro card that prints $11/mo next to $0.0357/hr, no root, and Cloudflare Enterprise as an add-on on Flexible.

A2 Hosting / hosting.com specs

Managed Linux 4 GB at $23.20/mo for a 1-year term then $58 renewal, no root, and the money-back page that excludes VPS.

CPU specs by provider

What providers actually state about their processors: Vultr publishes a CPU vendor and vCPU type per plan, Linode a plan class, OVHcloud a vCore count, and most others only a family name.

How to load test a site

Why a fixed-VU test measures itself rather than the server, what k6 calls coordinated omission, why cached and uncached pages are two different benchmarks, and how to tell whether the ceiling is PHP, the database or the disk.

Network and traffic specs

Per-plan port speeds and transfer allowances as published, the overage rates where a provider states one, and the iperf3 method for measuring the server you actually rent.

What providers publish

The table is limited to figures with a public source read in September 2026; a blank means the provider states nothing beyond a storage type on the pages this site read. Prices are monthly.

ProviderNetwork figureDisk figureWhere it comes from
Linodenetwork_out cap per plan: Nanode 1 GB 1,000 Mbps, Linode 2 GB 2,000, 4 GB 4,000, 8 GB 5,000; transfer 1 TB, 2 TB, 4 TB, 5 TBDisk size onlytypes API
VultrBandwidth allowance per plan: vc2-1c-1gb 1,024 GB, vc2-2c-4gb 3,072 GB, vhp-1c-1gb-amd 2,048 GBDisk size and a type label: SSD (vc2), HIGHFREQUENCY (vhf), AMDHIGHPERF (vhp)plans API
HetznerShared 10 Gbit host connection; "you can expect about 300-500 Mbits" per server; no guaranteeNVMe label onlyHetzner spec page (plan FAQ)
AzurePer-size network bandwidth in Microsoft's VM size tablesPer-disk-tier IOPS and MB/s (a 32 GiB Standard SSD E4: up to 500 IOPS, 100 MB/s; Premium SSD P4: 120 IOPS, burst to 3,500); per-size uncached limits (B2ats v2: 3,750 IOPS, 85 MBps)Azure spec page and storage page
Google CloudIOPS per GiB by Persistent Disk typestorage page
InterServer10 Gbps shared port per slice; 2 TB transfer per sliceSSD (cloud) or SATA (storage line) labelInterServer guide (VPS page)
DigitalOcean, OVHcloud, Contabo, Hostwinds, RackNerd, Kamatera, BuyVMTransfer allowance per plan; DigitalOcean notes Premium Droplets have "higher network throughput"Storage type label onlyProvider pricing pages, as listed on the storage page

Two things to read in that table. The hyperscalers publish limits because they sell disks and network as metered products with defined ceilings; the plan-priced hosts publish labels because their disks and network are shared and they do not promise a share. And a cap is not a typical value: Linode's 1,000 Mbps on a Nanode is the most it will pass, not what a transfer will get at 3 pm.

How to test a server yourself

The measurement that matters is the one on the plan you will buy, in the region you will use, at the hours you care about. It costs an hour of a rented server. Vultr, Linode and DigitalOcean bill by the hour; on a monthly-billed host such as InterServer you pay the month.

  1. Create the server in the target region with the distribution you will run, and install the tools: apt install sysbench fio iperf3 on Debian or Ubuntu, dnf install sysbench fio iperf3 on AlmaLinux or Rocky (sysbench may need the EPEL repository).
  2. CPU, single thread: sysbench cpu --threads=1 --time=60 run. The figure to keep is events per second. Run it again with --threads equal to the vCPU count to see how the shares scale.
  3. Disk, random 4K: fio --name=randrw --filename=/root/fio.test --size=2G --rw=randrw --rwmixread=70 --bs=4k --iodepth=32 --numjobs=1 --direct=1 --runtime=60 --time_based --group_reporting, then delete the test file. Keep read and write IOPS separately. Use --direct=1, or you measure the page cache rather than the disk.
  4. Network: iperf3 -c <host> -t 30 -P 4 against a public iperf3 server in the region that matters to you, or better, against a second server of your own in the location your users are, since a public server's own load is part of the result. Add -R for the download direction.
  5. Repeat each test three times at different hours and keep the median. The spread between runs is as informative as the median: it is the shared-host effect.
  6. Destroy the server. On an hourly plan the bill is a few cents.

Providers that publish a test endpoint make the network part easier: Linode's per-region speedtest hosts, InterServer's speed test page with download files from its three datacenters, and the latency test buttons on Kamatera's and Hostwinds's datacenter pages are listed on the low-latency page.

Reading vendor numbers

  • "Up to" and caps. A ceiling, enforced by the provider's rate limiter. It tells you what you cannot exceed and nothing about the typical case.
  • Burst. Azure's "burst to 3,500 IOPS" is a credit-based allowance for short peaks; the base figure is what a sustained load gets.
  • Shared port. InterServer's 10 Gbps and Hetzner's 10 Gbit are the host's uplink, shared by every server on it; Hetzner is the only one of the two to publish a per-server expectation.
  • Storage labels. NVMe, SSD and RAID-10 describe hardware, not a rate. Two NVMe plans from two providers can differ by an order of magnitude in what the provider lets a single VM pull.
  • Third-party benchmark sites. Useful as a second opinion when they show the command, the date, the region and the plan; useless when they show a score without those four.

Frequently Asked Questions

Does BestUSAVPS run its own VPS benchmarks?

No. Every figure on this site is a specification or price a provider publishes, with a link to the page or API it came from. Earlier versions of this section presented test results; they were removed because a number this site cannot show the reader how to verify is not worth publishing. The three pages under Specification pages below are the replacement.

Which VPS providers publish performance figures?

In September 2026: Linode's types API gives a network_out cap in Mbps and a monthly transfer allowance per plan (Nanode 1 GB: 1,000 Mbps and 1 TB; Linode 4 GB: 4,000 Mbps and 4 TB). Vultr's plans API gives disk size, a disk type label (SSD, HIGHFREQUENCY, AMDHIGHPERF) and a bandwidth allowance per plan. Hetzner's plan FAQ says hosts have a shared 10 Gbit connection and "you can expect about 300-500 Mbits" per server, with no guarantee. Azure and Google Cloud publish disk IOPS and throughput per disk tier and VM size. The plan-priced hosts (Vultr, Linode, DigitalOcean, Hetzner, OVHcloud, Contabo, InterServer, Hostwinds, RackNerd, Kamatera, BuyVM) state a storage type and no IOPS figure; UpCloud is the exception, publishing over 100k read IOPS at 4k for MaxIOPS against up to 10,000 for standard storage.

How do I benchmark a VPS myself?

Rent the plan for an hour on a provider that bills hourly, install sysbench, fio and iperf3 from the distribution's packages, and run the three commands in the How to test section below: a single-thread CPU run, a 4K random read and write run against a test file on the server's disk, and a throughput test to a public iperf3 server or to your own second machine. Run each three times at different hours, keep the median, and delete the server. On a $5 plan the whole exercise costs a few cents.

Why do two VPS with the same specs perform differently?

Because a vCPU is a share of a physical core and a virtual disk is a share of a physical array; what the other tenants on the host are doing at that moment decides your share. That is why a single measurement, by anyone, is a snapshot rather than a property of the plan, and why the figures providers do publish are caps and allowances (network_out, bandwidth, IOPS limits) rather than typical values. Dedicated-CPU plans remove the CPU share; nothing removes the shared network.

Is NVMe faster than SSD on a VPS?

NVMe drives have much higher random I/O ceilings than SATA SSDs, and providers that publish disk figures (Azure, Google Cloud) show that. Whether a given NVMe-labelled plan is faster than a given SSD-labelled plan depends on how the provider shares and limits the array, which plan-priced hosts do not publish. Treat the label as a floor for the hardware, not a promise of a number; the storage specs page lists the labels per provider.

Where did the old per-provider benchmark pages go?

They are being replaced one by one with specification pages built from provider documentation and APIs, in the format of the three linked below. Until a provider's page is rebuilt it is not linked from this hub.

Sources

Figures in the table are from the linked APIs and pages as read on 2026-09-22 (Linode, Vultr) or as verified on the linked specification pages in September 2026 (Hetzner, Azure, Google Cloud, InterServer). The commands in How to test are the tools' standard invocations; the parameters are this site's suggestions.