When Apple presents the iPhone 18 and 18 Pro, the difference between the A20 and A20 Pro looks like an engineering chasm. In reality, it's often the same silicon crystal cut from a single manufacturing wafer. The difference is not made by a different design, but by a test that lasts just a few seconds.
Where does the variation on a single wafer come from?
The circuit is applied to a 300-millimeter silicon wafer layer by layer, and each pass is a chance for error: a speck of dust in the photoresist, a mask shift by fractions of a nanometer. At TSMC's N2 standards, where circuit elements measure just a few nanometers, such deviations become critical.
The larger the chip area, the higher the probability that at least one section will be defective. That's why even in the best production batches, some crystals reach the final stage with damage — and this is not a factory failure, but the physics of the process.
Probe testing determines chip class
Before the wafer is cut, microscopic needles are lowered onto it, touching the contact pads of each future chip. The test checks three things: whether all CPU, GPU, and neural block cores work correctly, at what voltage the crystal holds the specified frequency, and how much current leaks — that is, wasted without any benefit.
Crystals with a full set of working blocks and the best electrical characteristics go into the Pro version. The rest go into the base version. According to supply chain data, the iPhone 18 and 18 Pro will receive chips of the same crystal design — the difference arises at this sorting stage.
Hardware block disabling
Redundancy is built into the design from the start: cache memory has backup lines, and there are more physical cores on the crystal than the base configuration provides. A local defect then doesn't kill the entire chip — only the damaged area is disabled.
The disabling is irreversible: a current pulse burns out a micro-switch, and the defective block disappears forever from the chip's resource map. Apple has used this approach for years — the base MacBook Air with M1 had a 7-core GPU, the senior model had 8 cores, even though the crystal under the hood was physically identical.
The second sorting axis — voltage
Even fully functional crystals are not equal to each other. One needs low voltage for the target frequency, another needs more — and this decides everything, because power consumption grows proportionally to the square of voltage.
The chain is simple: lower voltage — less heat — longer retention of peak frequency without throttling. That's why Pro chips consistently work at higher frequencies: they have a thermal margin left where the base chip is already forced to reduce speed.
What this changes for the owner
The junior chip is not defective, but fully functional within the stated parameters. Binning allows you to use the entire wafer without waste — otherwise, electronics would cost significantly more, and the buyer would have to pay for it.
In practice, the difference is felt selectively: in messaging or automatic shooting, the absence of one graphics core is imperceptible; in demanding games, the senior chip holds a stable frame rate longer; and when editing video, the full set of GPU and NPU really accelerates rendering.
A question for the future: if the difference between "standard" and "Pro" is mostly the result of a factory test rather than separate engineering, are manufacturers ever ready to sell chips not by line brand, but by actual sorting class — the way it's already partially happening in the graphics card market?
