How FPSVibe estimates FPS

1. The inputs

• Graphics card index for 162 graphics chips and processor index for 119 processors. See "The performance scale" below for how these are built.
• For each of the 102 games, two numbers: how much graphics power it needs for about 60 FPS at 1080p High (GPU need), and how much processor work each frame takes (FPS per CPU point). Plus the RAM below which the game starts to stutter. The GPU need is fitted to the game's official requirements (see below).
• Your choice of resolution and quality preset.

2. The performance scale

• Graphics index: relative rasterization gaming performance at 1440p High/Ultra, RTX 4090 = 100. Newer cards can go above 100 (the RTX 5090 is 112). Anchors that keep the scale consistent: RTX 3060 12GB = 36.9 and RX 6600 = 30.6.
• 4K graphics index: a separate scale for 4K, RTX 4090 = 100 at 4K (the RTX 5090 is 124). The 1440p index is used for 1080p and 1440p, the 4K index for 4K. A single scale would keep the gap between two cards the same at every resolution, but at 4K fast cards pull further ahead (in reviews the RTX 5090 leads the 4090 by about 12-17% at 1440p and about 25% at 4K) and cards with 8 GB of video memory fall behind. Sources: the 4K results in Tom's Hardware GPU hierarchy tables and TechSpot reviews, mapped onto one scale through cards tested in both. Cards without a 4K test (109 of 162, marked rough) use their 1440p index times the typical 4K/1440p ratio for their video memory: 16 GB or more × 0.82, 10-12 GB × 0.75, 8 GB × 0.70, 6 GB × 0.68, 4 GB or less × 0.55, integrated × 0.5.
• Processor index: relative CPU-bound gaming performance at 1080p, Ryzen 7 9800X3D = 100. Anchors: Core i5-12400F = 54.6 and Ryzen 5 5600 = 53.6.
• Each value blends published measurements from more than one review where available. Different test suites are rescaled through hardware that appears in both, then averaged. No source's table is copied as is. Where no direct test exists, the value is interpolated from related hardware and marked as an estimate.
• Every chip has a confidence level. High: direct review data. Medium: estimated from related tests. Low: interpolated with no direct modern test; results for these parts carry a "Rough estimate" note.
• Integrated graphics values are a ranking scale, not real 1440p results, and assume dual-channel memory. A single RAM stick can cost 30–50%.

Performance index compiled by FPSVibe from published reviews by Tom's Hardware, TechSpot, Notebookcheck and others, and from manufacturer and Wikipedia specifications.

3. Fitting each game to its official requirements

A game's GPU need is set from the graphics cards its publisher lists, using two rules:

• The recommended card should get about 60 FPS at 1080p High.
• The minimum card should get about 30 FPS at 1080p Low.

When a tier lists several cards ("GTX 1060 or RX 580"), we average them. Publishers' minimum specs are usually generous, so the two rules rarely agree exactly. We weight the recommended rule 75% and the minimum rule 25%, so the recommended card lands close to 60 FPS and the minimum card usually lands a bit above 30. Games whose requirements name no card we can match (for example "DirectX 11 card with 1 GB") keep a hand-set value, and a game whose listed card has only a rough index can get a documented manual value (currently only GTA V). The processor and RAM inputs are not refitted.

4. The formula

GPU limit = 60 × (your GPU index ÷ game's GPU need) × resolution × preset
            GPU index: the 4K index at 4K, the 1440p index at 1080p and 1440p
CPU limit = your CPU index × game's FPS per CPU point (the same at every resolution)
FPS       = the lower of the two
if your RAM < game's RAM need:  FPS × 0.7
never above the game's own frame cap (e.g. 60 FPS in Elden Ring)

The graphics side scales with resolution and quality; the processor side does not. Whichever is lower is your bottleneck. FPS never rises with resolution: a 1440p result is never above the 1080p one, and 4K never above 1440p.

5. Labels and advice

• Smooth: 60 FPS or more. Playable: 30–59 FPS. Choppy: under 30 FPS.
• The "best setting" is the highest preset and resolution that still holds 60 FPS (or 30 FPS if 60 is out of reach).
• The card "for 60 FPS" is the weakest dedicated card whose index reaches the game's GPU need.
• GPU vs GPU (the tool and the comparison pages): "% faster" is the geometric mean of the per-game FPS ratios over every game. At 1440p and 4K it uses the graphics side of the formula only (no processor limit, no RAM limit, no frame caps), so the gap is the cards' own; otherwise a processor would cap light games and both cards would tie there. At 1080p it uses a fixed reference PC (Ryzen 7 9800X3D, 32 GB RAM, each game's frame cap), because fast cards really are held back by the processor at 1080p and reviews show a smaller gap there. Every FPS number shown in a comparison comes from that reference PC, never above a game's frame cap, and above 500 it reads "500+". The headline uses 1440p, or 4K when both cards score 75 or more on our index. Under 4% we call it "about the same". Game pages, graphics card pages and the FPS tools use your own processor, because they answer "what will I get on this PC?".
• The PC gaming score is (GPU index × 0.72 + CPU index × 0.28) × 10, out of 1000.

6. Data sources

• Game requirements: the public PC requirements on each game's Steam store page, fetched politely (cached, one request every 1.5 seconds). For games not on Steam (for example Valorant, Fortnite, Minecraft) we use the publisher's official page. Each game page links to its source and shows when it was last checked.
• Hardware specs (memory, power, cores, clocks, release date, launch price): manufacturer pages from NVIDIA, AMD and Intel and Wikipedia hardware lists. Each graphics card and processor page lists its own sources.
• Performance indexes: FPSVibe's blend of published reviews (Tom's Hardware, TechSpot / Hardware Unboxed, Notebookcheck, PCWorld, KitGuru and others), as described above. Game model inputs: fitted by FPSVibe and flagged as estimates.

7. Limits (please read)

• One average number per setting. No 1% lows, stutter or frame-time data.
• Some games cap the frame rate themselves (for example Elden Ring and Sekiro at 60 FPS). Results for those games stop at the cap, and the tools say so. Mods that remove caps are not modelled.
• Integrated graphics are compared on the same index as graphics cards. In demanding games our numbers for integrated graphics tend to be on the low side; results assume dual-channel memory.
• No upscaling (DLSS, FSR, XeSS), frame generation or ray tracing. These can change results a lot.
• 4K results for cards with 8 GB or less video memory are often limited by that memory (stutter, lower textures). For integrated graphics and cards with 4 GB or less, tables show 4K as "Not recommended" instead of a number.
• Laptop versions of desktop cards usually run slower than the desktop card with the same name.
• Game patches and drivers change performance over time. Expect real results within roughly ±25% for typical PCs.
• Very old processors (4 cores, no Hyper-Threading) may do worse in new games than the model suggests.

8. How often data updates

Game requirements can be refreshed weekly from the Steam store by an automated job that opens a change for review. Hardware is added as new cards and processors launch. Every record stores its source and the date it was last checked.

Spotted a number that looks off? Tell us.