GPU value rankings built on real pricing data and normalized benchmark performance.
I started this because I was trying to pick a GPU for my own build and kept running into the same problem — review sites rank cards by raw performance, not by what you actually get for your money. A card that scores 20% higher but costs 60% more isn't a better buy, it's just a more expensive one. So I built the index I wanted to read.
Takes benchmark FPS data, normalizes it across generations so old and new cards are comparable, then divides by current market price. The result is a single number: cost per FPS. Lower is better. No weighting, no opinion, just math.
Current index: 59 GPUs at 1440p Ultra — updated monthly.
Benchmark data is sourced from Tom's Hardware's GPU Hierarchy — one consistent test suite across all cards.
The normalization problem: Tom's updated their benchmark suite in 2022. Modern cards are tested with heavier, more demanding titles, so a legacy card's raw FPS score is inflated relative to a current card tested under the same conditions. Comparing them directly would be misleading.
The fix: I identified ~20 GPUs that appear in both the old and new benchmark suites and computed the ratio between their scores. The trimmed mean of those ratios gives a scale factor (~0.703) that converts legacy scores to modern equivalents. If a card appears in both suites, the modern score is used as-is.
Pricing is pulled from Newegg monthly — lowest available new price from a reputable AIB, non-OC model. No gray market, no flash sales.
Cost per FPS = current_price / normalized_fps
data/
raw/benchmarks/ # source benchmark CSVs (Tom's Hardware)
reference/ # GPU master list with TDP, VRAM, architecture
processed/ # normalized benchmarks, cleaned pricing
outputs/ # dated rankings CSVs, charts
src/
processing/ # normalization pipeline
scoring/ # cost-per-FPS calculation
visualization/ # chart generation (PNG + HTML)
pip install pandas matplotlib
# Normalize benchmarks + compute rankings
python src/processing/build_benchmarks.py
python src/scoring/cost_per_frame.py
# Generate charts
python src/visualization/generate_charts.pyOutputs land in outputs/ dated to today.
- 1080p tier (budget card focus)
- Used market pricing (Post 2)
- Buy/Wait signal based on historical price distribution
- Automated price refresh
- RAM and SSD value indices
- Benchmarks: Tom's Hardware GPU Hierarchy
- Pricing: Newegg
