How we calculate
BattleHub outputs FPS figures, prices and a score. All of these are model estimates, not measurements. This page sets out how the calculation works and where its limits are.
FPS estimate
Each of the 139 games carries a baseline: the average frame rate at 1440p on the High preset, measured on the reference system, an RTX 5090 with a Ryzen 7 9800X3D. For 36 graphics cards and processors we have measured figures from published tests; everything else is interpolated between those points. The model derives two limits from this, one for the card and one for the processor, and takes the lower. Resolution does not affect every card equally: an RTX 5090 drops to 66 percent at 4K, an RTX 4070 to 53.
What the model cannot do: it knows nothing about driver versions, background software, badly optimised patches or differences between individual scenes in a game. Ray tracing and upscaling are not included. It tells you what to expect from this class of hardware, not what your machine will deliver tonight.
Prices
Guide values are maintained for 195 products, aimed at the German market and the cheapest widely available model. Parts no longer sold new are marked as used-market prices. A product without a curated value is interpolated between its two nearest known neighbours.
Until July 2026 prices came from a formula over the score. That was wrong: for individual cards it was off by more than a factor of three, and it structurally could not express that the faster card is sometimes cheaper. Prices are snapshots and go stale, so check before buying.
BattleHub Gaming Score
A number from 0 to 100 combining raw performance, memory and value. It helps with sorting but does not replace a review. Two cards with the same score can differ noticeably in individual games.
Compatibility
We check socket against motherboard, memory generation against motherboard, power supply against the draw of CPU and graphics card, and cooling capacity against CPU heat output. We do not check whether the graphics card physically fits the case. Per-card length data does not exist yet, and checking against estimated dimensions would be worse than not checking, because it would look like it rests on data.
Unreleased games
6 titles in the model have not launched yet. For those, measurements cannot exist. Their baselines are derived from the studio's engine history and the level of detail shown in trailers, and are explicitly flagged as forecasts.
How we make money
Through affiliate links to Amazon. If someone buys through one, we earn a commission and the price stays the same for you. Recommendations follow the model and your budget, not the commission. Every purchase link is marked as such.
What the model is calibrated against
On 26 July 2026 the FPS model was recalibrated. Before that it capped system strength at 100 points, so even the best conceivable machine could never exceed a game's baseline. Against real measurements it was up to 2.4 times too low: the model put God of War Ragnarök at 110 frames, the measured figure was 268. A script re-checks these anchors on every change.
| Game | Resolution | Model | measured |
|---|---|---|---|
| God of War Ragnarök | 1440p | 264 | 268-1 % |
| God of War Ragnarök | 4K | 177 | 195-9 % |
| The Last of Us Part I | 1440p | 201 | 204-1 % |
| The Last of Us Part I | 4K | 135 | 125+8 % |
| Star Wars Jedi: Survivor | 1440p | 189 | 191-1 % |
| Star Wars Jedi: Survivor | 4K | 126 | 102+24 % |
| Marvel's Spider-Man Remastered | 1440p | 213 | 222-4 % |
| Starfield | 1440p | 123 | 125-2 % |
| S.T.A.L.K.E.R. 2: Heart of Chornobyl | 1440p | 93 | 94-1 % |
| S.T.A.L.K.E.R. 2: Heart of Chornobyl | 4K | 62 | 71-13 % |
| Delta Force | 4K | 160 | 1600 % |
To stay honest: 7 of 139 baselines come from a measurement. The rest were scaled up from the old figures with a fixed factor. That is better than before, but it is not a measurement, and individual games can be well off. Whenever we add a measurement, we correct the value.
Sources: Tom's Hardware, GPU Benchmarks Hierarchy (2026-07-26) · Tom's Hardware, CPU Benchmarks Hierarchy (2026-07-26) · TechSpot, Nvidia GeForce RTX 5090 Review (2026-07-26)
What to take away
Use the figures to compare orders of magnitude and to avoid obvious mistakes. Do not use them instead of a review of the specific card in the game you care about. If a figure here looks wrong to you, tell us and we will check it.