For business Forum ✈ TG
Log in
Dataset · CC BY 4.0 · version 2026.10.13

FPSHQ GPU ranking: open data in CSV and JSON

Ranking of 212 GPUs by FPSHQ: the top desktop card is the GeForce RTX 5090 (10,000 points), while the GeForce RTX 4060 ranks #64 at 37% of the leader. The dataset covers models from 2012–2025 and downloads as CSV and JSON under CC BY 4.0.

Download the dataset

Direct links are permanent: the file always holds the latest version. Limit: 30 downloads per minute per address; cache the file on your side (ETag, Last-Modified).

Download CSV 25 KB Download JSON 103 KB
Version
2026.10.13 ·
License
CC BY 4.0
SHA-256 (CSV)
26d94782ae3fa230b5d3bd8d2a0792838001a660de9309564bc0ccde37c9abf4
FPSHQ open API
“game + GPU” lookups in JSON

Key figures

Dataset description as of Oct 13, 2026; the numbers below are recalculated with every new version.

212
Rows
13
Columns
2026.10.13
Version
from Oct 13, 2026
2012–2025
Data period
GeForce RTX 5090
Leader (desktop)
Bench Score 10,000
37.0%
RTX 4060 vs leader
#64
CC BY 4.0
License
free, attribution required
In short

Which graphics card is the most powerful in the FPSHQ ranking, and where can I download the full ranking?

According to the FPSHQ ranking as of Oct 13, 2026, the fastest desktop graphics card is the GeForce RTX 5090 with a Bench Score of 10,000; the desktop median is 3,265 points. The ranking includes 212 GPUs: 124 desktop, 63 laptop and 25 integrated, without board-partner versions (ASUS, MSI and others are merged into the reference model). The relative_pct column shows the score as a percentage of the leader — for example, the GeForce RTX 4060 is at 37%. Bench Score is an FPSHQ index built from review benchmarks and specifications; it shows relative gaming performance, not FPS in a specific game. The file updates automatically, the version changes only when the data changes, and CC BY 4.0 requires a link to FPSHQ.

updated · source: How to cite this dataset, FPSHQ methodology

Top 15 GPUs by Bench Score: % of the leader

FPSHQ Bench Score as a percentage of the most powerful model; the toggle switches the segment.

GeForce RTX 5090GeForce RTX 5090100.0% · 10,000GeForce RTX 4090GeForce RTX 409089.0% · 8,900GeForce RTX 5080GeForce RTX 508082.0% · 8,200NVIDIA RTX 6000 AdaNVIDIA RTX 6000 Ada82.0% · 8,200GeForce RTX 4080 SuperGeForce RTX 4080 Super76.6% · 7,660GeForce RTX 4080GeForce RTX 408076.0% · 7,600Radeon RX 7900 XTXRadeon RX 7900 XTX74.0% · 7,400GeForce RTX 5070 TiGeForce RTX 5070 Ti71.0% · 7,100Radeon RX 9070 XTRadeon RX 9070 XT71.0% · 7,100Radeon RX 7900 XTRadeon RX 7900 XT68.2% · 6,820NVIDIA RTX 5000 AdaNVIDIA RTX 5000 Ada68.0% · 6,800GeForce RTX 4070 Ti SuperGeForce RTX 4070 Ti Super67.0% · 6,700Radeon Pro W7900Radeon Pro W790065.0% · 6,500GeForce RTX 3090 TiGeForce RTX 3090 Ti64.0% · 6,400Radeon RX 9070Radeon RX 907064.0% · 6,400GeForce RTX 5090GeForce RTX 5090100.0% · 10,000GeForce RTX 4090GeForce RTX 409089.0% · 8,900GeForce RTX 5080GeForce RTX 508082.0% · 8,200NVIDIA RTX 6000 AdaNVIDIA RTX 6000 Ada82.0% · 8,200GeForce RTX 4080 SuperGeForce RTX 4080 Super76.6% · 7,660GeForce RTX 4080GeForce RTX 408076.0% · 7,600Radeon RX 7900 XTXRadeon RX 7900 XTX74.0% · 7,400GeForce RTX 5070 TiGeForce RTX 5070 Ti71.0% · 7,100Radeon RX 9070 XTRadeon RX 9070 XT71.0% · 7,100Radeon RX 7900 XTRadeon RX 7900 XT68.2% · 6,820NVIDIA RTX 5000 AdaNVIDIA RTX 5000 Ada68.0% · 6,800GeForce RTX 4070 Ti SuperGeForce RTX 4070 Ti Super67.0% · 6,700Radeon Pro W7900Radeon Pro W790065.0% · 6,500GeForce RTX 3090 TiGeForce RTX 3090 Ti64.0% · 6,400Radeon RX 9070Radeon RX 907064.0% · 6,400

Table preview

First 100 of 212 rows; the file contains all rows and columns.

rankgpubrandsegmentrelease_yeararchitecturememory_gbmemory_typetdp_wbench_scorerelative_pct
1GeForce RTX 5090NVIDIADesktop2025Blackwell32GDDR757510,000100
2GeForce RTX 4090NVIDIADesktop2022Ada Lovelace24GDDR6X4508,90089
3GeForce RTX 5080NVIDIADesktop2025Blackwell16GDDR73608,20082
4NVIDIA RTX 6000 AdaNVIDIADesktop2024Workstation48GDDR63728,20082
5GeForce RTX 4080 SuperNVIDIADesktop2024Ada Lovelace16GDDR6X3207,66076.6
6GeForce RTX 4080NVIDIADesktop2022Ada Lovelace16GDDR6X3207,60076
7Radeon RX 7900 XTXAMDDesktop2022RDNA 324GDDR63557,40074
8GeForce RTX 5090 LaptopNVIDIALaptop2025Blackwell24GDDR71507,20072
9GeForce RTX 5070 TiNVIDIADesktop2025Blackwell16GDDR73007,10071
10Radeon RX 9070 XTAMDDesktop2025RDNA 416GDDR63047,10071
11Radeon RX 7900 XTAMDDesktop2022RDNA 320GDDR63006,82068.2
12NVIDIA RTX 5000 AdaNVIDIADesktop2023Workstation32GDDR63096,80068
13GeForce RTX 4070 Ti SuperNVIDIADesktop2024Ada Lovelace16GDDR6X2856,70067
14Radeon Pro W7900AMDDesktop2023Workstation48GDDR63256,50065
15GeForce RTX 3090 TiNVIDIADesktop2022Ampere24GDDR6X4506,40064
16Radeon RX 9070AMDDesktop2025RDNA 416GDDR62206,40064
17GeForce RTX 4070 TiNVIDIADesktop2023Ada Lovelace12GDDR6X2856,20062
18GeForce RTX 4090 LaptopNVIDIALaptop2023Ada Lovelace16GDDR61506,20062
19GeForce RTX 5070NVIDIADesktop2025Blackwell12GDDR72506,20062
20GeForce RTX 3090NVIDIADesktop2020Ampere24GDDR6X3506,00060
21GeForce RTX 4070 SuperNVIDIADesktop2024Ada Lovelace12GDDR6X2205,90059
22GeForce RTX 5080 LaptopNVIDIALaptop2025Blackwell16GDDR71505,80058
23Radeon RX 7900 GREAMDDesktop2024RDNA 316GDDR62605,80058
24Radeon RX 6950 XTAMDDesktop2022RDNA 216GDDR63355,78057.8
25Radeon RX 9070 GREAMDDesktop2025RDNA 412GDDR62205,61056.1
26GeForce RTX 3080 TiNVIDIADesktop2020Ampere12GDDR6X3505,60056
27NVIDIA RTX A6000NVIDIADesktop2020Workstation48GDDR62545,60056
28NVIDIA RTX 4500 AdaNVIDIADesktop2023Workstation24GDDR62505,50055
29GeForce RTX 4070NVIDIADesktop2023Ada Lovelace12GDDR6X2005,40054
30Radeon Pro W7800AMDDesktop2023Workstation32GDDR62705,40054
31Radeon RX 7900MAMDLaptop2023RDNA 3———5,40054
32Radeon RX 7800 XTAMDDesktop2022RDNA 316GDDR62635,30053
33GeForce RTX 3080NVIDIADesktop2020Ampere10GDDR6X3205,20052
34GeForce RTX 4080 LaptopNVIDIALaptop2023Ada Lovelace12GDDR61505,20052
35Radeon RX 6900 XTAMDDesktop2020RDNA 216GDDR63005,20052
36GeForce RTX 5060 TiNVIDIADesktop2025Blackwell16GDDR71804,90049
37GeForce RTX 5070 Ti LaptopNVIDIALaptop2025Blackwell12GDDR71154,90049
38Radeon RX 6800 XTAMDDesktop2020RDNA 216GDDR63004,90049
39NVIDIA RTX A5000NVIDIADesktop2021Workstation24GDDR62184,80048
40Radeon RX 7700 XTAMDDesktop2022RDNA 312GDDR62454,67046.7
41GeForce RTX 3070 TiNVIDIADesktop2021Ampere8GDDR6X2904,60046
42Radeon RX 6800AMDDesktop2020RDNA 216GDDR62504,60046
43Radeon RX 7700AMDDesktop2024RDNA 312GDDR62254,60046
44Radeon RX 7800MAMDLaptop2023RDNA 3———4,60046
45NVIDIA RTX 4000 AdaNVIDIADesktop2023Workstation20GDDR62044,50045
46GeForce RTX 4060 TiNVIDIADesktop2023Ada Lovelace8GDDR61604,40044
47GeForce RTX 4060 Ti 16GBNVIDIADesktop2023Ada Lovelace16GDDR61654,40044
48GeForce RTX 3070NVIDIADesktop2020Ampere8GDDR62204,30043
49GeForce RTX 3080 Ti LaptopNVIDIALaptop2021Ampere16GDDR61504,30043
50Radeon RX 9060 XTAMDDesktop2025RDNA 416GDDR61604,30043

Column dictionary

Column names in the files are in English and stay the same between versions; new columns are only appended at the end.

ColumnTypeUnitMeaning
rankinteger—Rank (1 = highest score)
gputext—Graphics card model
gpu_slugtext—GPU identifier in FPSHQ URLs
brandtext—Manufacturer
segmenttext—Segment: desktop, laptop, integrated (GPU) or desktop, mobile (CPU)
release_yearintegeryearRelease year
architecturetext—GPU architecture
memory_gbintegerGBVideo memory size
memory_typetext—Video memory type
tdp_wintegerWTypical board power per manufacturer specs
bench_scoreintegerpointsFPSHQ Bench Score — gaming performance score of the GPU
relative_pctnumber%Score as a percentage of the ranking leader
urllink—Page on fpshq.com

How to cite this dataset

The CC BY 4.0 license lets you copy, modify and use the data for any purpose, including commercial, as long as you give attribution.

Attribution line (required)

Source: FPSHQ — “FPSHQ GPU ranking” (https://fpshq.com/en/data/gpu-rating/), license CC BY 4.0

Reference

FPSHQ (2026). FPSHQ GPU ranking [Data set], version 2026.10.13. https://fpshq.com/en/data/gpu-rating/. License CC BY 4.0. Accessed 2026-10-15.

BibTeX
@misc{fpshq_gpu_rating_2026,
  author       = {{FPSHQ}},
  title        = {FPSHQ GPU ranking},
  year         = {2026},
  version      = {2026.10.13},
  howpublished = {\url{https://fpshq.com/en/data/gpu-rating/}},
  note         = {Dataset, CC BY 4.0}
}

A link next to the chart, table or text that uses the data is enough. If you changed the data, say so (“based on FPSHQ data, recalculated by the author”). Call model estimates estimates, not measurements.

Version log

Version = snapshot date; a new version appears only when the file content changes.

VersionDateRowsChangesPeriod
2026.10.13212 +0 added, −3 removed, 199 changed 2012-01-01 — 2025-12-31
2026.10.10215 +0 added, −0 removed, 9 changed 2012-01-01 — 2025-12-31
2026.10.09215 +0 added, −0 removed, 14 changed 2012-01-01 — 2025-12-31
2026.10.06215 +0 added, −0 removed, 0 changed 2012-01-01 — 2025-12-31

Open data FAQ

How is Bench Score different from FPS?
Bench Score is a relative index of gaming performance. FPS in a specific game also depends on the game requirements, resolution, preset and CPU — see the “Can it run” pages and the “FPSHQ FPS model” dataset for that.
Why aren’t ASUS ROG or MSI Gaming X cards in the ranking?
Board-partner versions differ from the reference model by a few percent, so the ranking merges them into one row per model (for example GeForce RTX 4070). This way the same GPU is not listed dozens of times.
Can I use FPSHQ data in a commercial project?
Yes. The dataset is licensed under CC BY 4.0: you may copy, change and use the data for any purpose, including commercial. The only condition is attribution: “Source: FPSHQ” with a link to the dataset page next to the data.
How do I cite the “FPSHQ GPU ranking” dataset?
Use the attribution line “Source: FPSHQ — “FPSHQ GPU ranking” (https://fpshq.com/en/data/gpu-rating/), license CC BY 4.0”. For research and analytical texts use this reference: FPSHQ (2026). FPSHQ GPU ranking [Data set], version 2026.10.13. https://fpshq.com/en/data/gpu-rating/. License CC BY 4.0. Accessed 2026-10-15.
How often is the dataset updated?
The snapshot is rebuilt every 24 h. When the content changes, a new dated version appears (now 2026.10.13 from Oct 13, 2026), and the version log shows how many rows were added, removed and changed.
Do the files contain personal data?
No. The files contain no IP addresses, device IDs, accounts or exact timestamps. Gamer hardware statistics are published only as shares and only above the k-anonymity threshold; community benchmarks only as medians per “game × GPU × resolution × preset” cell.

Methodology and sources

The dataset includes every published GPU in the FPSHQ catalog with a non-zero Bench Score, except board-partner versions (ASUS, MSI, Gigabyte etc.), which repeat the reference model. Bench Score follows the FPSHQ methodology (review benchmarks, architecture, clocks, memory); relative_pct = score ÷ leader score × 100. Segment: desktop, laptop or integrated. Year, memory and TDP come from manufacturer specifications. The dataset snapshot is built automatically from the FPSHQ database every 24 h; a new version appears only when the content changes. Game and hardware names follow the FPSHQ catalog; links lead to site pages. More about the methodology →

Sources:
FPSHQ Team · Data updated
Measure FPS on your PCThe FPSHQ Windows app adds to community benchmarksFPS in TelegramThe FPSHQ bot answers from the same data