Skip to content

VPS comparison / Benchmarks

VPS benchmarks

Equal vCPU counts do not mean equal speed. These are measured results for servers we price, each from a test Spare Cores ran on the server type itself, with the plan's monthly bill beside it. Where To Deploy did not run the tests. One group so far: Hetzner's four 2 vCPU servers.

Lowest bill per 1,000 multi-core points
$4.38
CAX11, Ampere Altra
Fastest single core
1,970
CCX13, dedicated cores
Most multi-core points
3,378
CPX22, 2.9 times the lowest

Hetzner Cloud's 2 vCPU servers, measured

The four 2 vCPU server types Hetzner sells: the same provider, vCPU count and tester, on three kinds of processor and both shared and dedicated cores. Bills are for one server running all month with 100 GB of transfer out a month and a public IPv4, the VPS comparison's default workload.

Hetzner Cloud's 2 vCPU servers: measured results by Spare Cores and monthly bills, lowest bill first
Server Processor as tested Geekbench 6, single-corepointsGeekbench 6, multi-corepointsPostgreSQL read-only, 2 clientstransactions a minute Your bill Bill per 1,000 multi-core points Measured
Hetzner CX232 vCPU, shared · 4 GB RAM Intel Xeon (Skylake)Hetzner lists the CX line as Intel or AMD, so your server may get the other one (Hetzner's page) 6541,177123 $7.09/moPlan $6.49/mo $6.02 28 Aug 2026
Hetzner CAX112 vCPU, shared · 4 GB RAM Ampere Altra (Arm, Neoverse-N1) 9561,734183 $7.59/moPlan $6.99/mo $4.38 29 Sep 2026
Hetzner CPX222 vCPU, shared · 4 GB RAM AMD EPYC (Genoa) 1,8643,378309 $23.59/moPlan $22.99/mo $6.98 27 Aug 2026
Hetzner CCX132 vCPU, dedicated · 8 GB RAM AMD EPYC (Milan) 1,9702,523220 $51.09/moPlan $50.49/mo $20.25 27 Aug 2026

What the numbers say

The Arm server costs least per point. The CAX11 on Ampere Altra (Arm, Neoverse-N1) processors measures 1,734 multi-core points for a $7.59 bill, $4.38 per 1,000 points. Software and container images built only for x86 need an Arm build to run on it; the Arm plans in the VPS comparison say which providers sell them.

A faster shared processor is not a better deal per point. The CPX22, on AMD EPYC (Genoa), scores 2.9 times the CX23's single-core result and 2.9 times its multi-core result, for 3.3 times the bill. Choose it when one request has to finish faster, not to get more work per dollar.

Dedicated cores buy steadiness, not more points. The CCX13's cores are not shared with other customers, and it has the fastest single core here, yet its multi-core result is below the CPX22's at 2.2 times the bill. On a shared server, a busy neighbour can slow your app at any time; that is what the dedicated plan's price pays for, and no single test run shows it.

How these were measured

Results: Spare Cores, raw data, licensed CC BY-SA 4.0. This table adds our prices and bills to them and is shared under the same licence.

More groups follow when we find results measured the same way on servers we price. To compare every VPS plan by its bill, use the VPS comparison.

Find a tool or provider

Compare hosting costsToolVPS price comparisonToolFree VPS hosting offersToolDocker hosting: PaaS, VPS or serverlessToolPaaS pricing calculatorToolKubernetes cost calculatorToolChoose a cloud regionToolBare metal vs cloud cost crossoverToolAI inference routing by workloadToolMCP hosting by trust boundaryToolSupabase vs Neon for managed PostgresToolFree Docker hosting: what each tier allowsToolFree web app hosting: what breaks firstToolStorage-heavy VPS plansGuideWindows VPS hostingGuideHosting for high-egress workloadsGuideCheapest production-ready Docker hostingGuideAPI and backend hostingGuideStatic and front-end hostingGuideDevelopment vs production high-availability costGuideUS, UK and EU hosting regionsGuideHosting alternativesGuideVercel alternativesGuideRailway alternativesGuideRender alternativesGuideAll hosting providersProviderHosting price glossaryTermAkamai Connected CloudProviderAkamai Cloud (Linode)ProviderAlibaba CloudProviderAmazon Web ServicesProviderBack4appProviderCivoProviderClever CloudProviderCloud 66ProviderCloudflare PagesProviderCloudflare WorkersProviderContaboProviderDeno DeployProviderDigitalOceanProviderFirebase HostingProviderFly.ioProviderGitHub PagesProviderGoogle CloudProviderGreenCloudVPSProviderHerokuProviderHetznerProviderHostBrrProviderHostingerProviderIBM CloudProviderIONOSProviderKamateraProviderKoyebProviderLeapcellProviderLeasewebProviderMicrosoft AzureProviderNeonProviderNetcupProviderNetlifyProviderNorthflankProviderOneProviderProviderOracle CloudProviderOVHcloudProviderPlatform.shProviderPorterProviderPythonAnywhereProviderQoveryProviderRailwayProviderRareCloudProviderRenderProviderReplitProviderScalewayProviderSevallaProviderSupabaseProviderTencent CloudProviderUpCloudProviderUpsunProviderVercelProviderVultrProviderZeaburProviderData egressTermData ingressTermInter-AZ transferTermInter-region transferTermIncluded data transferTermPooled data transferTermEgress overageTermUnmetered transferTermShared vCPUTermBurstable vCPUTermDedicated vCPUTermCPU creditsTermControl plane feeTermWorker nodeTermLoad balancer costTermBlock storageTermObject storageTermEphemeral storageTermStorage overageTermPublic IPv4TermScale to zeroTermCold startTermMinimum billable durationTermAlways-on minimumTermPromotional vs renewal pricingTermCommitted use and reserved instancesTermPlatform feeTermPer-seat pricingTermBYOCTermList, calculated, usage-based and quote pricesTermHow we calculate your billAbout