The GPU Price Index: how we spot real used GPU deals
Buying a used GPU quickly turns into a jungle
On the second-hand market, two listings for the same graphics card can be 80, sometimes 150 € apart. In that jungle, it is hard to tell whether you have found a bargain or are about to overpay — possibly as much as a brand-new card.
That is exactly the problem the GPU Price Index solves. The idea is simple: watch used listings continuously, compare them model by model, and surface the ones that genuinely deserve your attention. Here is how it works, with no black box.
One reference point: the median
For each card model (an RTX 4070 or an RX 7800 XT, say), we look at every asking price across recent listings, then compute the median: the price in the middle, the one that separates the cheaper half of the listings from the more expensive half.
Why the median rather than a plain average? Because one far-fetched listing at 1,200 € or a typo at 50 € would be enough to distort an average. The median ignores those extremes and reflects the price that genuinely represents the model at a given moment. It is a landmark: above it, you are paying on the high side; below it, things start to get interesting.
How a listing becomes a “deal”
Not every listing below the median is worth the same. We separate two cases:
- “Deal”: the listing is among the cheapest 10 % of listings for that model while staying within a credible range. This is where the real bargains live: a low price, but not an implausible one.
- “Worth checking”: the listing sits more than 40 % below the median. That is no cause for celebration. A gap that wide often hides something: a faulty card, a data-entry mistake, a misclassified bundle, or a listing too good to be true. We flag it so you examine it before getting carried away — a reflex worth applying to every listing, in truth.
The distinction is drawn automatically and in the same way for every model, so that “deal” always means the same thing from one card to the next.
We filter out the noise
A cheap listing is not automatically good news. Before anything is displayed, we set aside or mark:
- cards advertised as faulty: “dead”, “not working”, “for parts”, “won't power on” and equivalent wording are detected and kept out of the deals;
- bundles (several cards, or a complete PC): their price cannot be compared with that of a single card, so they never pollute a model's median.
The result: what surfaces as a “deal” is a single card, working as far as we can tell, at a price that is genuinely low for its model.
What it does for you, concretely
If you are buying:
- you know where the fair price sits for a model (the median) before you even start negotiating;
- you spot the genuine bargains of the moment at a glance, without watching the classifieds morning and night;
- you are warned when a listing is too good to be true (“worth checking”).
If you are selling:
- the estimator gives you a price consistent with the market, so you neither give your card away nor leave it sitting unsold because it is priced too high. (→ /gpu/estimer)
Either way, you save time and decide against a figure rather than on instinct.
What the GPU Price Index is not
In the interest of honesty, let us be clear about the limits:
- the figures shown are asking prices (what sellers request in their listings), not transaction prices
- the median is a landmark, not a verdict: a listing slightly above it may well be worth it depending on condition and accessories
- a listing that disappears has not necessarily been sold — so we draw no conclusion whatsoever about how “fast” the market moves.
The GPU Price Index is a decision aid, not a crystal ball. But between buying blind and buying with a median, a deal threshold and a “worth checking” signal, the search for your next graphics card becomes a great deal easier.
Take a look at today's deals
Listings change constantly. To see where the good buys sit today, set your budget and let the deals rise to the top: the GPU Price Index →. And if you are the one selling, start with the estimator →.