Wafer Sorting / Binning
Wafer sorting and binning is what happens to each die's wafer-probe test results after the test itself is finished — classifying every individual die into a specific bin based on exactly how it performed, not just whether it passed or failed. A die that passes every functional test but runs faster than another otherwise-identical die doesn't necessarily become the same product; binning is the mechanism that turns a spread of real manufacturing variation into a deliberate lineup of distinct products sold at different price points.
Speed binning is the clearest, most consequential example of binning's real business impact, because it turns ordinary manufacturing variation into deliberate product segmentation. Every die from the same mask set and same process doesn't run at exactly the same maximum frequency — normal process variation across a wafer, and across different wafers, naturally produces a spread of real performance. Rather than treating that spread as noise to be averaged away, speed binning sorts die into explicit performance grades (fast, typical, slow) based on their actual tested maximum frequency, and those different grades frequently become entirely different products sold at different price points — the fastest bin commanding a premium, slower bins sold as lower-tier parts, even though every die started from an identical design and mask set.
A die's test program, run automatically during wafer probe, is what actually applies each bin's specific pass/fail criteria and assigns the die to its correct category. The test program applies the full set of test vectors, measures the resulting electrical responses, and compares those measurements against the specification limits that define each bin — a die exceeding spec on every dimension and hitting the fastest speed grade's threshold goes to the top bin; a die that passes basic functionality but can't hit the fastest grade's frequency requirement goes to a lower bin; a die failing even basic functional criteria goes to the fail bin and never leaves the fab as a product at all. This automated, criteria-driven sorting is what makes binning consistent and repeatable at the volume a modern fab actually produces.
Parametric measurements collected during binning — threshold voltage, saturation drain current, leakage current, and similar electrical parameters — serve a second purpose beyond simply sorting the current die: statistical process monitoring across the whole fab. These same measurements, aggregated across many die and many lots, reveal process trends that a simple pass/fail binning decision alone would never surface — a parameter drifting gradually across successive lots, even while every individual die still passes its bin's criteria comfortably, can be an early signal of a developing process issue worth investigating before it actually starts moving die into lower bins or outright failure.
Marking which bin a die belongs to has moved almost entirely from physical ink marks to electronic wafer maps, which is a genuine process-flow improvement rather than just a cosmetic change. Traditional ink-dot marking physically stamped bad die so they could be visually identified and discarded later in the flow — functional, but limited to a simple pass/fail distinction and vulnerable to physical handling issues. Modern electronic wafer maps record each die's exact bin assignment digitally, letting downstream equipment (dicing, pick-and-place, packaging routing) read a die's bin directly and route it automatically to the correct process path — fail bins scrapped, different speed bins potentially routed to different packaging or product lines — without relying on a physical mark surviving intact through every subsequent handling step.
Total wafer sort testing cost scales directly with test time per die multiplied by the number of die on the wafer, which makes test efficiency a real, ongoing economic concern rather than a fixed cost to simply accept. A longer, more thorough test program catches more marginal die and sorts bins more precisely, but it also costs more test time per die — a genuine tradeoff between test thoroughness and test cost that every fab has to balance deliberately. This is exactly why multi-site probing (testing several die in parallel) and careful test-program optimization (running only the tests actually needed to make a confident bin decision, rather than exhaustively testing every parameter on every die) are both treated as serious, ongoing engineering disciplines rather than one-time setup decisions made once and never revisited.