Model register
Every calculator on the Hub is a model in this register, with its purpose, its inputs and the ranges they hold for, its outputs, the formula, its uncertainty and its known limits. A formula change is published as a new version, and a pinned result keeps the version it was made with. Where a model follows a method published elsewhere, its note names the source and says where the Hub's version differs and why.
Calculators
Die per waferv1Published method, with stated departures
Estimates how many whole rectangular die fit on a round wafer.
Formula
gross die = floor(pi x (D/2)^2 / A - pi x D / sqrt(2 x A)), where D = wafer diameter - 2 x edge exclusion and A = die width x die height
| Input | Unit | Range it holds for | Comes from |
|---|---|---|---|
| Wafer diameter | mm | 50 to 450 | Your figure |
| Die width | mm | 0.1 to 60 | Your figure |
| Die height | mm | 0.1 to 60 | Your figure |
| Edge exclusion | mm | 0 to 10 | Your figure |
Outputs: Gross die per wafer (die).
Assumptions
- Rectangular die on a full round wafer.
- The scribe lane is not added: enter the die size including it if you want it counted.
Uncertainty: An approximation of a placement count. An exact layout can place a few more or fewer die, and the difference grows as the die gets large relative to the wafer.
Known limits
- No flat or notch.
- No partial-die harvesting or reticle stitching.
- Gross die only: apply a yield model for good die.
Where the method comes from
Follows AnySilicon, die per wafer formula (opens in a new tab). The platform's version departs from it as follows.
- The diameter is reduced by twice the edge exclusion before both terms are computed. Why: Die in the excluded ring cannot be used; leaving it in overstates the count. With an edge exclusion of 0 the arithmetic is the same as the published refined formula.
- The result is rounded down to a whole die. Why: A fraction of a die is not a die.
Worked example from the source: 300 mm wafer, 10 x 10 mm die. The article takes about 706 die by area alone and subtracts about 5 percent for edge loss, giving roughly 670. Source result: about 670 gross die. This model, same inputs: 640.
Reference example, checked by test: Worked example in this method note: 300 mm wafer, 10 x 10 mm die, 3 mm edge exclusion. Gross die per wafer 613.
Versions
- v1 (2026-10-17): Entered in the model register. The formula is the one the tool already used; nothing about the result changed.
Yield estimatev1Published method, same arithmetic
Estimates the share of die free of random defects, and the good die per wafer that leaves.
Formula
yield = e^(-D0 x A), with A = die area in cm2; good die = floor(gross die x yield), with gross die from the die per wafer model at a 3 mm edge exclusion
| Input | Unit | Range it holds for | Comes from |
|---|---|---|---|
| Defect density | defects per cm2 | 0 to 10 | Your figure |
| Die width | mm | 0.1 to 60 | Your figure |
| Die height | mm | 0.1 to 60 | Your figure |
| Wafer diameter | mm | 50 to 450 | Your figure |
Outputs: Die yield (fraction); Gross die per wafer (die); Good die per wafer (die).
Assumptions
- Defects are randomly and evenly spread (no clustering).
- Edge exclusion is fixed at 3 mm.
Uncertainty: The Poisson model ignores defect clustering, so it tends to understate yield for large die. Treat the result as a lower estimate.
Known limits
- Random defects only: no parametric or systematic yield loss.
- One defect density for the whole die.
Where the method comes from
Follows Poisson yield model, as set out in EESemi, Test yield models (opens in a new tab). The arithmetic here is the same; the platform claims no improvement on it.
Worked example from the source: Die area 1 cm2 and defect density 1 per cm2, so D0 x A = 1. Source result: yield = e^-1, about 0.37. This model, same inputs: 0.367879.
Reference example, checked by test: Worked example in this method note: D0 = 0.5 per cm2, 10 x 10 mm die, 300 mm wafer. Die yield 0.606531; Gross die per wafer 613; Good die per wafer 371.
Versions
- v1 (2026-10-17): Entered in the model register. The formula is the one the tool already used; nothing about the result changed.
Package / ASIC costv1Hub's own arithmetic
A parametric per-unit cost: silicon over good die, plus package, test and amortized engineering cost.
Formula
silicon per unit = wafer cost / (gross die x Poisson yield); NRE per unit = NRE / volume; total = silicon + package + test + NRE per unit
| Input | Unit | Range it holds for | Comes from |
|---|---|---|---|
| Wafer cost | USD | 0 to 1000000 | Your figure |
| Wafer diameter | mm | 50 to 450 | Your figure |
| Die width | mm | 0.1 to 60 | Your figure |
| Die height | mm | 0.1 to 60 | Your figure |
| Defect density | defects per cm2 | 0 to 10 | Your figure |
| Package cost per unit | USD | 0 to 100000 | Your figure |
| Test cost per unit | USD | 0 to 100000 | Your figure |
| Total NRE | USD | 0 to 10000000000 | Your figure |
| Volume | units | 1 to 10000000000 | Your figure |
Outputs: Gross die per wafer (die); Die yield (fraction); Silicon cost per unit (USD); NRE per unit (USD); Total per unit (USD).
Assumptions
- Gross die from the die per wafer model at a 3 mm edge exclusion.
- Yield from the Poisson model.
- Package and test costs are per unit as entered, with no further yield loss.
Uncertainty: As good as the figures entered. Every input is the user's assumption; the tool asserts no market price.
Known limits
- Assembly and final-test yield are not applied: the founder cost primitives stage them.
- No minimum order value.
Where the method comes from
The platform's own cost roll-up. It reuses the die per wafer and Poisson yield models above and adds plain sums; it borrows no other published method.
Reference example, checked by test: Worked example in this method note: USD 4,000 wafer, 300 mm, 10 x 10 mm, D0 0.5, USD 0.80 package, USD 0.30 test, USD 100,000 NRE over 100,000 units. Gross die per wafer 613; Silicon cost per unit 10.7584; NRE per unit 1; Total per unit 12.8584.
Versions
- v1 (2026-10-17): Entered in the model register. The formula is the one the tool already used; nothing about the result changed.
84 more models for registered members.
Register or sign in to use this tool
The working tool, its inputs and its results are for registered members. Registration is free and takes a minute. You come back to this page afterwards.