Head-to-head component benchmarks for your next purchase.
Comparing by catalog specs (cores, GHz, TFLOPs) is misleading, because they don't predict real gaming performance well. Here we pit two processors or two graphics cards against each other by their gaming score, built from averages of real 1080p benchmarks, and give you the verdict as a percentage plus estimated FPS in five reference games.
To isolate each component, we pair the graphics cards with a top-end processor (and vice versa) and 32 GB of RAM, so the result reflects the power of the component you are comparing and not a bottleneck from the rest of the build. In very light or competitive games both options tie: that is normal, because the limit is set by the engine or the processor, not the graphics card.
Use the comparison to decide between two specific models before buying, always factoring in price: sometimes a card that is 5-8% slower costs considerably less and is the smart buy. Figures carry a ±10-15% margin.
Comparing two parts by their product name is misleading more often than not. Vendors reuse numbers across generations, and a higher-numbered card from the previous generation can land below a current mid-range one. That is why this comparison works with normalised in-game performance scores rather than isolated specs like clock speed or core count, which are only comparable within the same architecture.
Spec-sheet numbers explain the why, not the how much. More VRAM does not make a card faster, but it stops it from choking on textures at high resolutions. More cores will not speed up a game that only uses four threads, but they change everything if you also stream or edit video. Power draw, in turn, decides whether your PSU copes and how much heat you need to move out of the case.
When comparing processors, look at the difference in the right context. At 1080p with a fast graphics card the gap between two CPUs is visible; at 4K, two processors five years apart can deliver practically the same frame rate because the graphics card is the limit in both cases. A CPU comparison that does not state the test resolution does not say much.
The percentage difference we show is an average across games. In practice the spread is wide: the same pair of cards can be 10% apart in one engine and 30% apart in another, depending on how each one exploits the architecture, cache or memory bandwidth. Use it to choose between two options, not as a promise of what you will see in your favourite game.
And one idea that saves money: past a certain point, paying 40% more for 12% more performance rarely pays off, especially if your monitor cannot display those extra frames. Before moving up a tier, check your display's resolution and refresh rate.
Because TFLOPs are not comparable across architectures or brands. An RTX performs differently than a Radeon with the same TFLOPs; real benchmarks capture that difference.
Yes, that is its purpose: to see at a glance how much more one option performs over another in games and at each resolution. Always cross-check with the current price.
No: only CPU vs CPU or GPU vs GPU. To check whether a processor and a graphics card are balanced, use the CPU/GPU Pairing tool.
From in-game performance averages collected from public reviews, normalised onto a common scale so parts from different generations can be compared. The method is explained on our methodology page.
The database focuses on desktop parts. Laptop versions with the same name usually perform noticeably worse because of their power limits.
It is reviewed when new CPU and GPU generations launch, incorporating the published figures from the independent testing available at that time.