low cost VPS server: ARM or x86?

The exact decision

This page answers one purchasing question: whether switching the processor architecture makes a matched low cost VPS server cheaper when the provider, advertised cores, memory, storage, contract view and storefront tax setting stay the same. The netcup catalogs show matching x86 and ARM64 G12.5 tiers from 4 GB through 64 GB RAM, with the same published pre-VAT monthly amount at every matched tier. [A] [B]

The software check is narrower than an installation test. It records whether five pinned Docker Official Image tags publish both linux/amd64 and linux/arm64 variants in their official manifests. It does not prove that plugins, native extensions, private images, backup agents or license servers in your own stack work on ARM. [C] [D] [E] [F] [G] [H]

Published inputs

These are catalog inputs, not benchmark results. The catalogs label the x86 storage as SSD and the ARM64 storage as NVMe; matching the advertised quantity does not establish matching storage performance. The provider uses 1 TB = 1024 GB in its own storage calculator. [A] [B]

Calculation: architecture alone changes no base-price line

Keep currency and tax treatment separate from architecture. The comparison below uses the pre-VAT monthly amounts printed beside the German 19% VAT display, then multiplies each monthly amount by the selected 12-month view. It does not add backup, software licenses, operator time or migration work because no matching amount for those items was collected here. [A] [B]

Example at the smallest matched tier: monthly cost per advertised GB of RAM = €8.26 ÷ 4 = €2.065 per GB-month for both architectures. Storage unit conversion = 64 ÷ 1024 = 0.0625 TB for both advertised capacities. ARM64 saving for the selected term = €99.12 − €99.12 = €0.00. [A] [B]

Inference: architecture is not a price lever in this specific dated catalog match. That does not show equal speed, energy use or application capacity. A buyer who selects ARM here is exchanging software-compatibility risk for an architecture choice, not receiving a lower published base charge. Only a workload test on the purchased systems could support a performance or capacity comparison.

Compatibility check: inspect the pinned image manifest

Do not treat a generic product-page claim about Linux support as proof for your exact deployment. Resolve the actual image tag used by the compose file or release manifest, then check that tag rather than an unpinned product name. Docker documents multi-platform images as a manifest list that points to platform-specific variants; the registry chooses a matching variant when one exists. [C]

A listed ARM64 variant clears only the image-availability gate. Before purchase, repeat the check for every image, inspect separately downloaded binaries and agents, and identify any package built during deployment. If one required component has no ARM64 artifact, keep x86 or replace that component before ordering. This is a preflight rule, not a claim that the listed stack was installed successfully.

Quick answer

Is an ARM low cost VPS server cheaper than x86?

Not in this matched netcup snapshot. The five published G12.5 tier pairs have the same pre-VAT monthly amounts, so their calculated 12-month base totals are also the same. Choose ARM only if the exact software manifest passes and you have a reason to test that architecture; otherwise x86 avoids introducing an architecture change without a published base-price saving. [A] [B]

Downloadable comparison table

The CSV keeps the verification date, official URLs, plan inputs, shown calculations and pinned image tags together. Blank fields would mean unknown, not free; this table has no blank price input because both matched catalog pages exposed the required base amounts.

The file is static and opens in a spreadsheet. Recheck every linked source before using it for a later purchase.

Before ordering an ARM server

1. Copy every exact container tag and separately downloaded binary from the deployment configuration. Do not substitute a product family name for the deployed tag.
2. Open each official tag manifest and confirm that linux/arm64 is listed. Keep the source URL and check date beside the tag.
3. Check native extensions, backup agents, monitoring agents and licensed software separately. They can have a different support matrix from the main container.
4. Compare the final cart using the same location, network option, contract and tax setting. The catalog match on this page is not a customer-specific checkout quote.
5. If the manifest gate passes, run an authorized ARM installation and workload test before claiming performance or operational savings. No such test result is included here.

Evidence and number sources

All pages and API responses were checked September 23, 2026. That is the verification date, not a vendor offer-expiry date. Published prices and image manifests can change.

• A — x86 plans: netcup vServer (VPS) catalog. G12.5 plan names, x86 architecture, cores, RAM, storage, pre-VAT monthly amounts, German 19% VAT display and selected 12-month catalog view.
• B — ARM64 plans: netcup ARM server catalog. Matching G12.5 ARM64 plan names, resources, prices, German 19% VAT display and selected 12-month catalog view.
• C — manifest method: Docker multi-platform documentation. Platform-specific image variants and the linux/amd64/linux/arm64 platform notation.
• D — NGINX tag: Docker Hub official tag API. The pinned tag and listed image architectures.
• E — PostgreSQL tag: Docker Hub official tag API. The pinned tag and listed image architectures.
• F — WordPress tag: Docker Hub official tag API. The pinned tag and listed image architectures.
• G — Redis tag: Docker Hub official tag API. The pinned tag and listed image architectures.
• H — MySQL tag: Docker Hub official tag API. The pinned tag and listed image architectures.

What people on these servers say in their own words

Reproduced verbatim and unedited, each with its discussion title, link and date. These are experience reports from people who run such servers, not official provider wording and not this page's conclusion. Where a quote and the evidence above disagree, the linked official page states the published rule and the quote stays as one person's experience. Most of these quotes are about ARM servers in general, not about the plans compared above.

How these are counted: one sample is one discussion thread, not one person. Several people replying inside the same thread count once. The six quotes come from six threads.

Reported in more than one thread: one thing — what decides an ARM choice is software and image compatibility, not the price line. Quote 1 (thread one) refuses ARM until a standard Debian Aarch64 image boots, quote 2 (thread two) says getting into ARM is hard outside cloud offerings that absorb the problem, quote 4 (thread four) recalls a period when open source projects did not guarantee ARM support, and quote 5 (thread five) points at the cost of cross builds aimed at servers.

Reported in one thread only, so not a pattern: quote 3, where every container image he wanted already had an ARM build, and quote 6, where ARM instances outperformed the x86 ones he was using. Both are one thread each, and neither cancels out the four above.

1 · Refuses ARM until a standard image boots · compatibility

Not with this compatibility mess. I would trust ARM for a server/pc system, the day I can boot standard Debian Aarch64 image on them. Until then x86 it is.

Arm wants a bigger slice of the chip business · 16 February 2026

Where this gets stuck: the blocker is the install image, not the monthly price, so a cheaper ARM plan can still be unusable. Way around it: confirm the exact image your provider offers boots before you buy, which is the manifest check described above.

2 · Hard to get into ARM outside cloud offerings · compatibility

However, it's hard to get into ARM other than using cloud offerings. Because those cloud offerings have handled for you the problematic case of ARM generally operating as "closed platform" even when everything is open source.

Bought myself an Ampere Altra system · 30 June 2025

Where this gets stuck: a managed platform hides the firmware and support gaps that a bare VPS leaves to you. Way around it: check who absorbs that work on your plan — the provider or you — before comparing ARM prices across plan types.

3 · Every image he wanted already had an ARM build · compatibility

I build out my server in Docker and I’ve been surprised that every image I’ve ever wanted to download has an ARM image.

CLI agents make self-hosting on a home server easier and fun · 12 January 2026

Where this gets stuck: a container-first stack hides the problem until one image in the stack lacks a build. Way around it: check the architectures listed for every image in your compose file, not just the base image.

4 · ARM support used to be less certain · compatibility

Also, at least until some time ago, ARM compatibility was not always guaranteed by open source projects (I think nowadays it became ubiquitous).

Raspberry Pi 500 Review: The keyboard is the computer, again · 9 December 2024

Where this gets stuck: "nowadays it became ubiquitous" is his assessment, not a guarantee for your dependency list. Way around it: verify each package you actually install, rather than assuming the ecosystem moved as a whole.

5 · The cross builds are where it hurts · compatibility

Remembering the Mac ARM transition pain wrt Docker and Node/Python/Lambda cross builds targeting servers, there's a lot to be said for binary compatibility.

AMD entered the CPU market with reverse-engineered Intel 8080 clone 50 years ago · 24 December 2025

Where this gets stuck: the cost lands in build time and tooling, not in the invoice. Way around it: price one cross build of your own stack before committing to an ARM plan; that is the number a monthly comparison will not show.

6 · ARM instances beat the x86 ones he was using · performance

One of the most eye opening things to me is how terrible ECS x86 performs vs graviton instances with Fargate. I've found much lower IO and honestly seems like some instances are 3rd generation Intel instance class.

AWS EC2 Instance Timeline · 20 July 2025

Where this gets stuck: the comparison is against a specific instance generation on one platform, so it does not transfer to a low cost VPS row. Way around it: compare on the same workload and the same generation, and treat performance notes as separate from the price comparison.