FAQ

You enter your current GPU. We look up every GPU currently listed on Amazon for your market, calculate the performance gain over your card using a unified Perf Index, and rank them by % gain per CA$100 (higher is better). The best-value option becomes your verdict.

Yes — use "For a game or app" instead of the general upgrade search. Pick a specific title (or the closest category if we don't track yours by name yet) and we'll check whether your current GPU already clears the requirement, and if not, show only upgrades that actually get you there.

A single number that blends two benchmarks: 3DMark Time Spy (70%) and FP32 throughput in teraflops (30%). Formula:

0.70 × (TimeSpy ÷ 360) + 0.30 × (TFLOPs ÷ 1.65)

Both inputs are from publicly available benchmark databases (3DMark and TechPowerUp). No marketing claims.

It measures how much faster you get per CA$100 spent. Formula:

(% faster than your current card ÷ listing price) × 100

A score of 18+ is excellent value. 6–18 is reasonable. Under 6 usually means waiting for a price drop or next-gen release is the better call.

Under 25% faster is rarely a step you will feel in day-to-day use, so upgrades below that threshold aren't scored here. 25%+ is a clear step up; 50%+ feels like a generational leap.

Enter your current GPU, switch to "For a game or app," and pick a 1440p target — the requirement bar and rankings adjust to that resolution automatically. As a rough guide, most current 1440p titles need at least 10–12GB VRAM and a Time Spy score around 14,000+.

Prices are scraped from Amazon daily. Listings that haven't been seen in 7 days are removed from results. New and used conditions are tracked separately.

Currently US (amazon.com), UK (amazon.co.uk), France (amazon.fr) and Germany (amazon.de). Canada and Australia are in progress.

Verdicts are priced independently per marketplace — Amazon UK and Amazon US carry different listings at different prices, so the best-value pick (and the % gain per £100 vs CA$100) can differ even for the same GPU.

Yes. When you click "Buy on Amazon" and make a purchase, we may earn a small commission at no extra cost to you. This is how the site stays free and independent.

Affiliate commissions do not influence rankings. Cards are ranked purely by performance gain per dollar, and we never promote a listing because it pays more.

We track 184+ GPU models. If yours is missing, it's likely because it predates our benchmark database or uses a very uncommon name variant. We add models regularly.

Yes, and they're flagged clearly. Used cards are often excellent value; a used RTX 4070 can outperform a new RTX 3080 at a lower price. The condition badge (New / Used) is shown on every listing.

Over 18 is a steal, 6–18 is reasonable depending on budget, and under 6 means wait for a price drop or next-gen release. These thresholds are what drive the Yes / Maybe / No verdict on every GPU page. They are set in US dollars and converted to each store's currency at fixed exchange rates, so the same real price gets the same verdict in every store.

Check the recommended power supply wattage on the GPU's product page or manufacturer spec sheet against your current PSU's rated wattage, leaving headroom for the rest of your system. We don't currently factor PSU compatibility into the verdict — treat it as a separate check before buying.

We source specs from 3DMark's official database and TechPowerUp GPU database. Time Spy is a synthetic rasterization benchmark, which correlates well with gaming performance at 1440p but won't reflect ray tracing or compute-heavy workloads equally. Real-world results always depend on your specific game, resolution, and CPU.

upgrade-gpu.com is an independent tool built and maintained by a single developer. It's a sister site to best-gpu.com, which shows the full GPU price/performance table for all markets.