How we score

Last updated October 3, 2026

Every score on Plugger is worked out the same way for every part in its category, from published specs and benchmark results. No store or brand can pay to change a score, a ranking or a pick.

The short version

Every score is a percentage of the best part in its category on the same site: 100 is the best, 50 is half as good by that measure. Scores only compare parts within one category.
Every part’s page lists its specs, so you can check them yourself.
Scores are approximate hints, not measurements of your computer. Your software, settings and the rest of your system matter.

What each score is based on

Plugger

Graphics Cards: From specs, as a percentage of the best graphics card here: hardware on paper, not measured game speed, which varies by maker and architecture.
Processors: From specs, as a percentage of the best processor here: hardware on paper, not measured game speed.
SSDs: From the drive's published specs, as a percentage of the best SSD here. Approximate: specs aren't a measurement.
Memory (RAM): From the kit's published specs, as a percentage of the best kit here. Approximate: specs aren't a measurement.

Plugger Creator

Workstation GPUs: This card's results in real apps (Blender, Premiere Pro and DaVinci Resolve), each compared with the best card measured in that app, and its VRAM. Approximate. An app the part hasn’t been measured in is filled with our estimate from its specs (never for DaVinci Resolve, which specs don’t predict well). A part with too few app results is scored from its specs instead.
Workstation Processors: This processor's results in real apps (Blender, Premiere Pro, DaVinci Resolve, After Effects and Photoshop), each compared with the best processor measured in that app. Approximate. An app the part hasn’t been measured in is filled with our estimate from its specs (never for DaVinci Resolve, which specs don’t predict well). A part with too few app results is scored from its specs instead.
Workstation SSDs: From the drive's published specs, with sustained speed and endurance for media work in mind, as a percentage of the best drive here. Approximate: specs aren't a measurement.
Workstation Memory: From the kit's published specs, with capacity for big projects in mind, as a percentage of the best kit here. Approximate: specs aren't a measurement.

Spec Score, in comparisons

The Spec Score compares two parts on their hardware specs alone (for example compute, memory and bandwidth for graphics cards, or cores and clocks for processors), weighted the same way for every part in the category. It tells you which part has more hardware on paper; it doesn’t measure speed in any particular game or app.

Real World Score, on Plugger Creator

Measured results from real apps, for graphics cards and processors. Each benchmark only appears on the parts it stresses:

Workstation GPUs:
- Blender: median scores from Blender Open Data (Blender 4.5, CC0), Sep 30, 2026.
- Puget Bench for Premiere Pro v1.0–1.1 (Premiere Pro 25.1–25.2): Overall Score (Standard), published by Puget Systems from at least 10 public results of whole systems, Sep 30, 2026.
- Puget Bench for DaVinci Resolve v1.2 (DaVinci Resolve 20.0–20.3): Overall Score (Basic), published by Puget Systems from at least 10 public results of whole systems, Sep 30, 2026.
Workstation Processors:
- Blender: median scores from Blender Open Data (Blender 4.5, CC0), Sep 30, 2026.
- Puget Bench for Premiere Pro v1.0–1.1 (Premiere Pro 25.1–25.2): Overall Score (Standard), published by Puget Systems from at least 10 public results of whole systems, Sep 30, 2026.
- Puget Bench for DaVinci Resolve v1.2 (DaVinci Resolve 20.0–20.3): Overall Score (Basic), published by Puget Systems from at least 10 public results of whole systems, Sep 30, 2026.
- Puget Bench for After Effects v1.0 (After Effects 23.2–25.2): Overall Score (Standard), published by Puget Systems from at least 10 public results of whole systems, Sep 30, 2026.
- Puget Bench for Photoshop v1.0 (Photoshop 24.0–27.5): Overall Score (Standard), published by Puget Systems from at least 10 public results of whole systems, Sep 30, 2026.

Blender Open Data is public domain (CC0). Puget Bench results are cited with Puget Systems’ permission, with the benchmark’s full name and version, and every comparison links to Puget’s own breakdown. Plugger isn’t paid by either.

Estimates. A benchmark row for a part still on sale never shows an estimate: without a trusted result it says “No trusted results yet”. A discontinued part without enough results may be estimated from its specs and measured parts of the same brand and era, and is always marked with an asterisk and a striped bar. We test each estimate method against parts with measured results, and don’t use one that isn’t close enough (that’s why DaVinci Resolve has no estimates).

Use-case rows

Memory bandwidth is worked out from speed and the number of memory channels the kit fills (two on desktops; one per module for registered workstation memory).
“Good for” on memory is a capacity guide for common workloads.
Endurance per dollar on SSDs is the rated lifetime writes (TBW) divided by the price shown.

Hardware facts

Some comparison rows aren’t scores but facts the makers publish, because they often decide which part is right for you:

AV1 encoding and 4:2:2 10-bit HEVC decoding (Plugger Creator): whether the graphics card, or a processor’s built-in graphics, has hardware for them, by chip generation, from the makers’ media engine specifications. Many Sony, Canon and Panasonic cameras record 4:2:2 10-bit HEVC, which edits far more smoothly with hardware decoding. Whether a particular app uses that hardware varies by app and version.
Local AI models (Plugger Creator): the largest common model size (7, 14, 32, 70, 120 billion parameters) whose 4-bit weights fit in the card’s memory, with room left for context and the display. It says what fits, not how fast it runs, and needs software that supports the card.
Upscaling (Plugger): the maker’s own upscaling and frame generation the card supports (DLSS, FSR or XeSS), by chip generation. AMD’s FSR and Intel’s XeSS also run on many other cards in games that include them.
Recommended PSU and card size (Plugger): the power supply the chip’s maker recommends, and the size of the maker’s own card where it sells one. Other brands’ cards vary in size and sometimes ask for more power, so check the listing. Where the maker doesn’t publish a figure we say “Not listed yet” rather than guess.
Cooler in the box, built-in graphics and socket support (Plugger processors): from AMD’s and Intel’s product pages. Socket support only states what the maker has published (AMD has committed to AM5 through 2027 and beyond).
PS5 storage upgrade (Plugger SSDs): checked against Sony’s published rules: an M.2 NVMe drive on PCIe 4.0 or 5.0 (5.0 drives run at 4.0 speed), 250 GB to 8 TB, 5,500 MB/s reads recommended, and a heatsink no taller than 11.25 mm.

Prices

Plugger doesn’t have store price data yet, so every price is the part’s launch MSRP, labelled as such. Comparisons say “launched at” for this reason. Check the store for today’s price. Commissions never change which parts we show, their scores or their order; see our Affiliate Disclosure.

Best value right now

The home page of Plugger and Plugger Creator picks one part per category that gives the most for the money today, for what that hardware is used for. Every part is judged by the same rule:

Plugger, Graphics Cards: most performance per dollar.
Plugger, Processors: most performance per dollar.
Plugger, SSDs: lowest cost per TB: room for your games is what you notice.
Plugger, Memory (RAM): most performance per dollar among 32 GB+ kits, today's gaming baseline.
Plugger Creator, Workstation GPUs: most performance per dollar.
Plugger Creator, Workstation Processors: most performance per dollar.
Plugger Creator, Workstation SSDs: lowest cost per TB among drives built for sustained writes and endurance.
Plugger Creator, Workstation Memory: lowest cost per GB: capacity is what big projects and scenes need.

Then, for every category:

Only parts on sale count, and only ones that clear a mainstream performance bar for that audience, so a weak part can’t win just by being cheap.
Prices are compared like for like: while any part has a store price, only parts with one count (a launch MSRP beside today’s store prices would look like a bargain it isn’t); with no store prices, launch prices are compared, and the pick says so. Used and discontinued parts never count here.
Picks are worked out again as prices update, and an open home page refreshes them, saying when they were last updated. A pick only changes when another part is clearly better value, so small price moves don’t swap it back and forth.
Each pick says what the next step up (the cheapest faster part) costs and adds, so you can judge whether spending more is worth it.
The score is the one comparisons use: on Plugger, hardware specs on paper (Spec Score), not measured game speed; on Plugger Creator, real app results where measured. No store or brand can pay to be a pick.

Best value used or refurbished

A separate row under it uses the same rules at used and refurbished eBay prices only (the same listings Where to buy shows), including discontinued parts. A part on sale only counts when its used price is below its new one.

New and used are never ranked together: a used part has no maker’s warranty, its returns are up to the seller, and one listing can sell within hours, so the rows answer different questions.
A discontinued part competes whenever it clears the same bar and has a used price; an old part can’t win just by being cheap.
A sold-out, out-of-stock or ended listing is removed, so the row can change or go empty between checks.

What a score can’t tell you

Whether a part fits your case, power supply, motherboard or budget. Check compatibility with the manufacturer before you buy.
How it runs your exact software. Benchmarks use fixed test projects; your projects may lean on different parts.
Noise, heat, build quality, drivers or support.

Spotted a mistake?

Tell us at support@plugger.site and we’ll check it and correct it.