About best-gpu.com
An independent tool for GPU price-to-performance comparison.
Why this site exists
Buying a GPU is surprisingly hard. Benchmark sites show raw scores without prices. Retailer pages show prices without normalized performance. Review sites are often months out of date. And GPU pricing on Amazon moves daily — a card that was bad value last week can become the best buy after a $50 price drop.
best-gpu.com was built to solve this: a single table where every GPU is ranked by a composite price-to-performance metric that updates whenever Amazon prices change. The tool was created by an independent developer with no affiliation to any GPU manufacturer or retailer. There is no team, no investors, and no sponsored content — just a tool that one person built to answer their own question when buying a GPU.
What this site does
best-gpu.com tracks 483+ graphics cards listed on Amazon US, updated daily. Every GPU in the table has a calculated Value Score — a price-to-performance metric that combines real-world gaming benchmark data (3DMark TimeSpy) with raw compute throughput (FP32 TFLOPS). The goal is a single number that answers: how much performance are you getting per dollar?
The GPU Upgrade Calculator extends this by letting you enter your current GPU and instantly see which upgrades offer the best performance gain per dollar above what you already have. The upgrade-from pages provide quick, shareable links for common upgrade paths.
How the data pipeline works
Prices are collected daily via a custom Chrome extension that scrapes Amazon search results and product pages. The extension runs through a curated list of GPU search queries (RTX 5090, RTX 4080, RX 7900 XTX, Arc B580, etc.), visits each product page, and records the current buy-box price and listing title. It matches GPU listings using regex patterns on the product title — matching NVIDIA RTX/GTX/Quadro/Tesla, AMD RX/Radeon, and Intel Arc naming conventions — and filters out non-GPU results like laptop GPUs, cables, brackets, and accessories.
Deduplication is done by ASIN (Amazon's unique product identifier), so the same physical card listed by different sellers at different prices appears as separate rows — each with its own price and affiliate link. Products not seen in a scrape for more than 7 days are marked unavailable and hidden from the default view.
Each scrape run is logged to a database audit table. Only price changes are written to the price history table — if a GPU's price hasn't moved, the history row is skipped to keep the dataset clean. The deals page uses this delta history to surface the biggest price drops.
Performance data (TFLOPS, TimeSpy scores, VRAM) is sourced from manufacturer specifications and community benchmark databases, curated manually in a lookup table. The Value Score and Performance Index are recomputed on every page load based on the current price.
Why 70% TimeSpy / 30% TFLOPS?
The 70/30 weighting was chosen because 3DMark TimeSpy is the most widely used and reproducible cross-platform gaming benchmark — it captures the full graphics pipeline (geometry, rasterization, shading) in a way that raw TFLOPS does not. However, FP32 throughput is still important for AI inference, video rendering, and compute workloads, so it earns a 30% weight to ensure compute-heavy cards (Quadro, professional GPUs) aren't unfairly penalized.
Both components normalize to 100 at the RTX 4090 reference point (36,000 TimeSpy / 165 TFLOPS). GPUs without TimeSpy data (many professional and datacenter cards) use compute throughput only.
What GPUs are tracked
The site tracks consumer gaming GPUs from NVIDIA (RTX 5000 series, RTX 4000 series, RTX 3000 series, GTX 1000/1600 series), AMD (RX 9000 series, RX 7000 series, RX 6000 series), and Intel (Arc B-series, Arc A-series). Professional and workstation GPUs (Quadro, Tesla, Radeon Pro) are included where Amazon listings exist.
What this site doesn't do
best-gpu.com only tracks Amazon US listings. It does not scrape Newegg, B&H Photo, Micro Center, Best Buy, or any other retailer — so prices at those stores may differ significantly. It does not track used listings from eBay or other secondary markets. It does not cover laptop GPUs. Pricing outside the United States is not tracked.
Affiliate disclosure
All product links are Amazon Associates affiliate links (tag: bestgpu-20). If you purchase a GPU after clicking a link on this site, we may earn a small commission from Amazon at no extra cost to you. This commission funds the site's hosting and development costs.
The affiliate relationship has no influence on rankings, which GPUs are included, or how Value Scores are calculated. No GPU manufacturer or retailer pays to appear in results or to be ranked higher. Every listing is ranked purely by its calculated Value Score based on live Amazon prices and published performance data.
Contact
Questions, corrections, or feedback: contact@best-gpu.com