Performance comparison
Official
Community
Striped = Linux
Native DLSS FSR XeSS = upscaler set by the preset
Charts show grouped summaries for clean comparisons. Spec columns come from the reviewed GPU catalog; expand a row for silicon details, price history, and the exact runs behind each number.
Open data
Get the data as JSON
Every number on this page comes from static JSON files built from the reviewed YAML. Use them in scripts, spreadsheets, or your own charts. No key or sign-up needed.
dashboard.json
The grouped official and community summaries behind this chart, plus the game, resolution, and preset lists.
gpus.json
The GPU catalog with full hardware specs, each GPU's official and community summaries, and links to its per-GPU files.
official.json
Every official run, one record per resolution and preset profile, with its OS details.
community.json
Grouped community summaries, with each run's machine details, credit, and submitter when they chose to give one.
Compare distros yourself
The chart shows one bar per OS, never one per distro. To compare distros, kernels, or Proton versions, download community.json, official.json, or a per-GPU summary.json: every run carries os, distro, kernel, mesa, and proton fields.
<source>/<gpu_id>/summary.json
One GPU's runs and summaries, for example official/rtx_5080_16gb_fe. GPU IDs are listed in gpus.json.
Command line
JavaScript
Contribute benchmark evidence
How community submissions work
Exact machine runs in; clean, comparable GPU summaries out.
Use a pull request to submit a run. The validator checks the profile and a maintainer reviews the evidence before it is included.
Your results become public data
- Anyone can use them under CC BY 4.0: copy them, make their own charts, put them on their site, use them commercially, or train AI on them.
- Sites have to say where the data came from. A site or app that uses it must credit Open GPU Benchmarks and link back, so readers can find the source, report problems, and contribute.
- You get credit here. Your name, or the source you found, stays with the result on this site, in the data, and in the Git history.
- You can't take it back. You can ask for a result to be removed from this site, but copies other people already made stay public.
- It covers everything you submit, including machine details and raw captures, so leave out anything you don't want public, and only submit what you're allowed to share.
Contribute benchmark evidence
How community submissions work
Exact machine runs in; clean, comparable GPU summaries out.
Use a pull request to submit a run. The validator checks the profile and a maintainer reviews the evidence before it is included.
Your results become public data
- Anyone can use them under CC BY 4.0: copy them, make their own charts, put them on their site, use them commercially, or train AI on them.
- Sites have to say where the data came from. A site or app that uses it must credit Open GPU Benchmarks and link back, so readers can find the source, report problems, and contribute.
- You get credit here. Your name, or the source you found, stays with the result on this site, in the data, and in the Git history.
- You can't take it back. You can ask for a result to be removed from this site, but copies other people already made stay public.
- It covers everything you submit, including machine details and raw captures, so leave out anything you don't want public, and only submit what you're allowed to share.
Fork and prepare
Fork the repository and use the submission template as a starting point.
Add profiles
Create result_YYYYMMDD_<game_id>.yaml in data/community/<gpu_id>/ and add each resolution/preset as a dashed result entry. New game? Add it to data/games.yaml in the same pull request. Your username is optional, and a result you found online only needs a link to where you found it.
Validate locally
Run the build and validator before opening your pull request.
Review evidence
Maintainers check the capture, profile, machine details, and significant outliers.
Approve and group
Merged runs are grouped into generated summaries for that GPU.
Ready to contribute? Copy the submission template, choose your GPU result folder, then follow the contribution guide.
Community YAML example
# data/community/rtx_5080_16gb_fe/result_20261003_cyberpunk_2077.yaml
# Required fields only. Everything else is optional.
gpu_id: rtx_5080_16gb_fe # catalog file name in data/gpus/
game_id: cyberpunk_2077 # id in data/games.yaml; add new games there
date: '2026-10-03' # the day you ran it, same as the file name
result:
- resolution: 1440p
graphics_preset: High
avg_fps: 79.1
p1_low: 65.0 # when your tool reports it
system:
os: windows # windows or linux
proof:
method: presentmon # the capture tool
# Found it in a video or review? Say where:
# credit:
# source_url: https://example.com/gpu-reviewReview checklist
Required
- GPU (the catalog ID)
- Game ID from data/games.yaml (add a new game there), resolution, and settings
- Avg FPS (1% low when your tool reports it)
- Operating system (system.os)
- Capture tool (proof.method)
- Frame generation off (the default)
- For a result you found: where you found it (credit.source_url)
Optional, if you want to share it
- Raw PresentMon capture or a summary link
- Laptop TGP or GPU power
- Your GitHub username
- Exact card or device name
- CPU, memory, driver, and power mode
Validate every profile: the leading dash makes result a YAML list. Every profile is checked independently and only matching GPU, power, game, resolution, and preset entries are averaged. A submission more than 25% away from a matching official baseline must be corrected, explained by a different profile, or removed during review.
Featured GPU picks
One favorite from each GPU maker. Reviews and guides are on the blog.
Affiliate links. Store buttons below are affiliate links: if you buy through one, the site may earn a commission at no extra cost to you. It never changes the benchmark numbers.