Skip to main content

Methodology explainer

Why UpgradeOrSave never blends workloads into one overall hardware score

Gaming, rendering, local AI and productivity answer different questions. Here is why each keeps its own scale and reference.

By UpgradeOrSave EditorialEditorial update Published

Editorial dates describe the text. They do not date the measurements. Measurement acquisition dates are unavailable in this view.

Many hardware sites summarise a product with one number. It is convenient, and it is also the fastest way to make a bad upgrade decision. A single score has to decide, on your behalf, how much rendering matters compared with gaming or local AI, and it has to put results measured on different scales onto one axis. UpgradeOrSave does neither.

One workload, one scale

Each workload index answers one question about one kind of work. The rendering index tells you how a product performs in Blender rendering; it tells you nothing about frame rates in games or speed in local language models. Combining them would require a weighting that matches nobody's real use, so we show each workload separately and let you choose the one you care about.

Even one workload can have several domains

Within a workload, results are grouped into reference domains. A domain is defined by its frozen reference and test context, and it is anchored so that its reference hardware scores one hundred. Two domains can share a workload name and still not share a scale: Blender rendering on one GPU backend is a different domain from Blender rendering on another. A score is only comparable with scores in the same domain.

When we compare two products, we use the highest-priority domain in which both have a ready result. If they share no domain, we say so and do not compute a percentage between them.

What this means on our pages

  • Rankings order one domain at a time, and switching domain switches the scale.
  • Hardware pages list each workload result separately, with its reference version.
  • Upgrade checks state the performance difference only for the workload you selected.
  • Missing results stay missing; another workload's result never stands in for them.

Why it matters for an upgrade

Two products that look close in one workload can be far apart in another. If you render in Blender and occasionally play games, a blended score could point you to a card that is excellent at the work you do least. Keeping workloads apart costs a little convenience and protects the decision that actually matters to you.

What this means for an upgrade

  • Pick the workload that drives your upgrade and compare within it.
  • Do not compare scores across different reference domains.

Limitations and uncertainty

  • Some workloads do not yet have a current reference for every category; those workloads show no index for that category.

Evidence and methodology

Source-rights checks and active research holds apply before evidence reaches this page. An unavailable result is omitted rather than estimated.

No commission-tracked links appear in this article. Any future affiliate links require separate activation and a clear disclosure near the link. Read our disclosure.