One Wafer, Two Price Tags: How Chip Binning Turns Defects Into Product Lines

Two processors sit on a shelf at very different prices. They came off the same wafer, made by the same process, on the same day. One has cores disabled that work perfectly well.

This is binning, and once you understand it, a lot of otherwise strange behaviour in the chip market stops being strange.

The Wafer Is the Unit of Cost

Chips are not manufactured individually. A wafer — a disc of silicon — goes through a long sequence of steps, and the cost of that sequence is essentially fixed regardless of how many usable chips come out the other end.

The wafer is then cut into dies, each of which becomes a chip. So the economics are: a large, roughly constant cost, divided by however many dies turn out to be good.

That denominator is where everything interesting happens.

Defects Land at Random

Manufacturing at this scale is not perfect. Contamination, lithography variation and material flaws mean some areas of a wafer contain defects, and those defects fall more or less at random across the surface.

Now consider what that does to dies of different sizes. If defects are randomly distributed, a larger die presents a larger target. Doubling the die area more than doubles the chance that at least one defect lands inside it, because you need every part of the die to be good.

This is why yield falls faster than die area grows, and why the largest chips are disproportionately expensive rather than proportionately so. A big die is not twice the cost of a die half its size — it is considerably more, because a much larger share of them fail.

Harvesting: Selling the Damage

Here is the elegant part. A defect usually does not ruin a whole chip. It ruins a region.

Modern processors are built from repeated blocks — cores, cache slices, memory controllers, graphics units. If a defect lands in one core and the rest of the die is fine, the manufacturer can disable that core and sell the chip as a lower-tier part with fewer cores.

A die that would have been scrap becomes a product. The manufacturer recovers revenue from silicon it has already paid to produce, and the customer gets a cheaper part that works correctly.

This is why product lines so often look like the same chip with different amounts switched off. That is exactly what they are, and it is a feature of the economics rather than an accident.

Speed Binning: The Same Chip, Sorted

The second axis is performance rather than defects.

Even among dies with no defects at all, small variations in the manufacturing process mean some run stably at higher clock speeds, or hit the same clock at lower voltage, than others. Every chip is tested and sorted by what it can actually do.

The ones that clock highest become the top parts. The rest are rated lower and sold accordingly. Nothing is wrong with any of them; they simply landed in different places on a distribution.

This is the origin of the “silicon lottery” that enthusiasts talk about — two chips with identical model numbers can behave differently at the margins, because the rating is a guarantee of a minimum rather than a description of the exact part.

Disabling Good Silicon on Purpose

The part that sounds like a scandal and is not, once you see the constraint.

Demand does not arrive in the same proportions as the yield distribution. A manufacturer may have plenty of orders for a mid-tier part and a surplus of fully-working dies. Rather than leaving demand unmet or the inventory idle, it disables working blocks on good dies and sells them as the lower tier.

The alternative is not that everyone gets the better chip cheaply. The alternative is that the cheaper part is unavailable, or that the whole line is priced against the yield of the best dies — which would make everything more expensive.

It is also why unlocking disabled blocks occasionally worked in the past: sometimes the silicon really was fine.

Why Tier Pricing Ignores Cost

The payoff. Once you know that a flagship and a mid-range part are frequently the same physical die, the pricing gap becomes impossible to read as a manufacturing cost difference — because there largely isn’t one.

Tier pricing is segmentation. It sorts customers by willingness to pay, using performance as the sorting mechanism. The cost basis for the top and middle of a product line is far closer than the prices suggest, and the gap is a commercial decision.

That is not a criticism. Recovering the cost of a fixed, enormous manufacturing investment requires charging different customers different amounts, and binning is a genuinely clever way to do that while wasting almost nothing. But it does mean the intuition that a chip costing three times as much cost three times as much to make is simply wrong.