Wafer Probe
Wafer probe testing electrically tests every single die on a wafer before it's ever diced or packaged — pressing an array of tiny needles directly onto each die's bond pads, running it through a battery of electrical tests, and marking it pass or fail before the wafer ever moves to the far more expensive packaging step. It's the single quality gate that decides which die are even worth packaging in the first place, and getting it right has an outsized economic impact precisely because packaging a defective die wastes real assembly cost on a part that was never going to work anyway.
A probe card is the physical interface that makes wafer probe possible at all, pressing a precise array of tiny needle-like contacts directly onto a die's bond pads with enough simultaneous, reliable electrical contact to actually run real tests. Different probe technologies suit different pad geometries: cantilever probes are the traditional, flexible approach that's been in use for decades; MEMS-based probes pack in much higher contact density, suited to the finer pad pitches modern dies increasingly use; vertical probes handle flip-chip bump arrays, where the contact geometry differs from a traditional wire-bond pad layout entirely. Choosing the right probe technology for a given die's pad layout is itself a real engineering decision, not an afterthought.
The test sequence at wafer probe covers several genuinely different categories of measurement, each catching a different class of potential defect. DC tests measure fundamental electrical parameters like leakage current, drive strength, and threshold voltage — catching gross process or doping problems before anything more complex even gets tried. Functional tests apply scan patterns to verify the die's logic actually behaves correctly. Speed tests (often called shmoo testing) find the maximum frequency a die can actually operate at reliably, directly feeding into speed-grade binning decisions made later. Memory built-in self-test exercises every on-chip SRAM array using dedicated on-die test circuitry, and analog tests check things like ADC/DAC linearity or PLL lock range that a purely digital functional test would never catch.
Testing a die costs real money, but the economics overwhelmingly favor catching a defect at wafer probe rather than letting it slip through to packaging. A single die's probe test might cost a tiny fraction of a cent, while packaging that same die — if it turns out to be defective — can cost many times more, since packaging involves real materials, equipment time, and process steps that wafer probe simply doesn't. Catching even a modest fraction of defective die at the probe stage, before that packaging cost is ever spent, routinely pays for the entire test operation many times over — which is exactly why wafer probe exists as a dedicated, mandatory step rather than something skipped to save time.
Multi-site probing — testing several die simultaneously rather than one at a time — exists specifically because test time itself is a meaningful cost driver, not just the test equipment and probe card. A probe card and tester configured to contact and test four, eight, or more die in parallel directly multiplies test throughput for roughly the same per-wafer test time, which matters enormously at high production volumes where total test cost scales directly with however many individual test cycles a wafer requires. This parallelism is one of the main levers fabs pull to keep test cost manageable as die counts per wafer, and total production volume, both keep climbing.
Wafer probe results feed directly into everything that happens downstream — packaging decisions, yield tracking, and binning — making it the single data source that connects fabrication quality to the eventual finished, shippable product. Every die's pass/fail status (and, for passing die, its specific performance characteristics) gets recorded, whether as a traditional physical ink mark or, far more commonly today, in an electronic wafer map that downstream equipment reads automatically. This record is what lets a fab track yield trends over time, diagnose systematic process problems through spatial patterns across the wafer, and route each individual die to exactly the packaging and binning path its own test results actually earned it.