Skip to content
Cloud Provider Benchmarks

Benchmarks

Cloud price-performance: normalised VM and GPU pricing across providers

Price lists use different shapes, currencies and billing increments. This page normalises them: VM catalogs are decomposed into implied per-vCPU and per-GiB prices, GPU offers into per-GPU-hour prices, and billing granularity is shown alongside, because per-minute billing changes the cost of short jobs.

Updated:

Direct answer

Normalise cloud prices to per-vCPU-hour, per-GiB-hour and per-GPU-hour in a single currency, account for billing granularity, and only then combine with measured performance. Figures here carry their source and access date; currencies are not converted silently.

What is measured

  • Implied per-vCPU-hour and per-GiB-hour prices (least-squares fit)
  • Per-GPU-hour list prices
  • Billing granularity and minimums

Billing granularity

Billing granularity
Provider Granularity Sub-hour billing Details Currencies
PortvMind Cloud Per minute Yes Per-minute pay-as-you-go. Stopping a resource stops the compute charge; allocated disks and stored data keep accruing. Monthly figures in the calculator are 720-hour estimates, excluding tax. New accounts receive a free usage allowance (published as ₺2,000 on the Turkish site and $45 on the English pricing page). Source TRY, USD
Amazon Web Services Per second Yes On-demand Linux EC2 instances are billed per second with a 60-second minimum; other operating systems and services have their own meters. Source USD
Microsoft Azure Mixed Unverified Usage-based pay-as-you-go with service-specific meters; reservations and savings plans available. Billing granularity for virtual machines is not recorded here as verified. Source USD, EUR
Google Cloud Per second Yes Compute Engine VMs are billed per second with a one-minute minimum; sustained- and committed-use discounts apply to eligible machine types. Source USD
DigitalOcean Per second Yes GPU Droplets are billed per second with a minimum charge of 60 seconds or $0.01, whichever is higher; powered-off GPU Droplets continue to bill until destroyed. Source USD
Vultr Per hour No Hourly billing. GPU plans bill on a 730-hour month (actual calendar hours); standard compute instances cap monthly billing at 672 hours. Source USD
Hetzner Per hour No Cloud servers are billed per started hour up to a monthly price cap. Object Storage has an hourly base price with included storage and egress quotas. Source EUR, USD

Implied per-vCPU and per-GiB prices

Least-squares fit (hourly = a·vCPU + b·GiB) for providers whose full price catalog has been captured.

Implied unit prices
Provider / family Region Per vCPU-hour Per GiB-hour R² n
PortvMind · general purpose tr-ist-01 TRY 0.0570 TRY 0.2392 0.999 10
PortvMind · memory optimised tr-ist-01 TRY 0.5273 TRY 0.2278 1.000 6

Full catalogs for the other providers have not been captured in this release; they will be fitted with the same method once added. For GPU prices see the GPU price-performance page. GPU price-performance

Verified results