Hard IP port effort estimator
Estimate what porting a hard macro to another process or metal stack involves, from its mix of custom, compiled and digital content and the change.
Digital and physical designFree to useHow it worksSourcesWhere it helps
Fills the inputs below with typical figures. Replace any of them with your own.
- Total effort
- 1,141 engineer-weeks
- About 24.8 engineer-years
- Schedule
- 24.5 months
- 106 weeks with a team of 20
- Total cost
- $6,453,837
- $5,133,453 labour, $570,384 EDA, $750,000 shuttle and qualification
- Design effort against a new design
- 100%
- A new design of 2.5 mm2 is 801 engineer-weeks
Check these
- The port costs about as much as a new design. Published accounts say analog circuits below about 180 nm often need redesign rather than migration.
- Circuit migration and re-optimisation215.4 engineer-weeks19%
- Layout migration or redesign326.9 engineer-weeks29%
- Verification, views and documentation261.5 engineer-weeks23%
- Characterisation49.88 engineer-weeks4%
- Test chip design10 engineer-weeks1%
- Test chip bring-up and measurement14 engineer-weeks1%
- Qualification8 engineer-weeks1%
- Project management106.3 engineer-weeks9%
- Contingency148.8 engineer-weeks13%
Show the data as a table
| Item | Engineer-weeks |
|---|---|
| Circuit migration and re-optimisation | 215.4 engineer-weeks |
| Layout migration or redesign | 326.9 engineer-weeks |
| Verification, views and documentation | 261.5 engineer-weeks |
| Characterisation | 49.88 engineer-weeks |
| Test chip design | 10 engineer-weeks |
| Test chip bring-up and measurement | 14 engineer-weeks |
| Qualification | 8 engineer-weeks |
| Project management | 106.3 engineer-weeks |
| Contingency | 148.8 engineer-weeks |
| Phase | Weeks | How it is set |
|---|---|---|
| Design | 56.97 | Circuit, layout, verification, tool set-up and test chip design, shared across the team |
| Characterisation | 3.491 | Runs while the test chip is in fabrication |
| Test chip fabrication | 16 | Fixed lead time |
| Bring-up and measurement | 5.227 | Limited to three people |
| Qualification | 14 | Fixed duration |
Sign in to save your inputs and come back to them later.
| Similarity | Multiplier | Effort (engineer-weeks) | Schedule (months) | Cost (USD) | Design effort against a new design |
|---|---|---|---|---|---|
| High | 0.75 | 753 | 18.9 | $4,512,626 | 63% |
| Medium | 1 | 968 | 22 | $5,591,077 | 84% |
| Low | 1.2 | 1,141 | 24.5 | $6,453,837 | 100% |
Show the data as a table
| Similarity multiplier (1 is medium) | Effort (engineer-weeks) |
|---|---|
| 0.5 | 537 |
| 0.75 | 753 |
| 1 | 968 |
| 1.25 | 1,184 |
| 1.5 | 1,400 |
| 2 | 1,831 |
| Multiplier | Effort (engineer-weeks) |
|---|---|
| 0.5 | 537 |
| 0.75 | 753 |
| 1 | 968 |
| 1.25 | 1,184 |
| 1.5 | 1,400 |
| 2 | 1,831 |
How it works
Porting a hard macro (an analog block, a PHY, a memory, a standard cell library) to another process or metal stack costs anything from a small fraction of a new design to as much as one. This tool prices it with a parametric effort model: a base effort from the macro's size and content mix, scaled by the kind of change, how similar the design rules are, the team's experience and whether a migration tool is available, plus the fixed costs of characterisation, a test chip and qualification. Every factor is a Hub editorial estimate that you should replace with your own project history.
- Set the base effort of a new design. Each content type has an effort rate in engineer-weeks per mm2 for building it from scratch: custom analog, custom digital, compiled memory and standard cell digital. The base is the size times the mix-weighted rate. A library priced by cell count uses engineer-weeks per cell instead of the standard cell area share.
base = area x (analog share x analog rate + custom digital share x custom digital rate + memory share x memory rate + standard cell share x standard cell rate) + cells x rate per cell - Split it into activities. The base is divided between circuit work (schematic and pre-layout simulation), layout, and verification with views and documentation. The default split is 28%, 40% and 32%.
activity base = base x activity split - Scale for the change. The kind of change decides how much of each activity is repeated: a new node repeats most of it, a metal stack change or a PDK version change very little. The design rule similarity, the team's experience and the migration tool then scale it. A migration tool helps the circuit and layout activities only; it does not help verification.
activity effort = activity base x change factor x similarity factor x skill factor x tool factor (circuit and layout only) - Add characterisation. Characterisation scales with the number of corners. Each corner has a fixed cost plus a cost per mm2 for each content type, or per cell for a library, then the change-type factor.
characterisation = corners x (fixed + area x sum of share x rate per mm2 + cells x rate per cell) x change factor - Add the fixed items. A test chip adds design and bring-up effort, and its shuttle cost. Qualification adds effort and a cost. A migration tool adds set-up effort. Project management is a percentage of everything, and contingency a percentage of that.
total = (activities + characterisation + test chip + qualification + tool set-up) x (1 + project management) x (1 + contingency) - Make the schedule and cost. The design and characterisation phases are shared across the team at the stated efficiency. Characterisation runs while the test chip is in fabrication, bring-up is limited to three people, and qualification takes a fixed time. Cost is engineer-weeks times the loaded rate plus the EDA and compute rate, plus the shuttle and qualification costs.
design weeks = effort / (team x efficiency); schedule = design + max(characterisation, fabrication) + bring-up + qualification, all x (1 + contingency) - Test the similarity. The sensitivity view repeats the whole estimate at high, medium and low design rule similarity, and the chart sweeps the similarity factor, because similarity is the least certain input at the start of a port.
same model with the similarity factor set to 0.75, 1 and 1.2
Assumptions
- Every rate and factor is a Hub editorial estimate. The public sources describe what is hard about porting, not how many engineer-weeks it takes. Calibrate the rates on a past port of your own before relying on the result.
- Effort is in engineer-weeks; a loaded rate covers salary, overhead and facilities, and a separate rate covers EDA licences and compute.
- Size is the area of the finished macro or a representative set of instances. A device count is converted to an area with an average devices-per-mm2 figure.
- The kinds of change are from the same foundry family or different foundries; the model does not separate a licensed re-use of IP from a clean re-implementation.
Limits and warnings
- A parametric model is a first estimate to within a factor of two. It cannot see an unusual specification, a new circuit topology or a missing model.
- Below about 180 nm and at FinFET nodes, analog circuits often need redesign rather than migration; a low similarity setting moves the estimate toward a redesign but cannot replace a design review.
- Migration tool factors are kept modest (circuit work at 0.7 and layout work at 0.85 of the unaided effort) because the published accounts say tool set-up and manual layout and simulation work remove much of the gain. Tool vendors claim more.
- Licence fees for the target process IP, foundry NRE, reticle sets beyond the shuttle cost, customer qualification and a production test programme are not included.
Sources
The methods and default figures rest on these public sources. The wording here is our own; read the source for its full text.
- Analog Migration Equals Redesign (opens in a new tab), Semiconductor Engineering (Ann Mutschler), 24 May 2018. Why analog migration across nodes is far harder than digital reuse, and why FinFET effects push teams to redesign.
- Accelerating Analog Design Migration (opens in a new tab), Synopsys blog (Yuval Shay), first published in Semiconductor Engineering, 12 October 2023. The migration steps and why layout migration and simulation are harder than schematic migration.
- Reuse of Analog Mixed Signal IP for SoC Design: Progress Report (opens in a new tab), Cadence Design Systems (Tim Henricks), published on Design and Reuse, 8 November 2004. Tool-assisted porting that took about as much effort as laying out from scratch; it pays off for many similar blocks.
Where it helps
The lifecycle stages this tool serves: 1 stage in 1 lifecycle.
More digital and physical design tools
Method notev1Formula, ranges, sources and limits
Estimates the engineer-weeks, schedule and cost of porting a hard macro to another process or metal stack from its size and content mix, the kind of change, design rule similarity, team experience, tooling, characterisation corners, test chip and qualification.
Formula
base = area x (analog share x analog rate + custom digital share x custom digital rate + memory share x memory rate + standard cell share x standard cell rate) + cells x rate per cell; activity base = base x activity split; activity effort = activity base x change factor x similarity factor x skill factor x tool factor (circuit and layout only); characterisation = corners x (fixed + area x sum of share x rate per mm2 + cells x rate per cell) x change factor; total = (activities + characterisation + test chip + qualification + tool set-up) x (1 + project management) x (1 + contingency); design weeks = effort / (team x efficiency); schedule = design + max(characterisation, fabrication) + bring-up + qualification, all x (1 + contingency); same model with the similarity factor set to 0.75, 1 and 1.2
| Input | Unit | Range it holds for | Comes from |
|---|---|---|---|
| Size given as | choice | area, devices | Your figure |
| Macro area | mm2 | from 0 | Your figure |
| Device count | devices | from 0 | Your figure |
| Average devices per mm2 | devices/mm2 | from 1000 | Your figure |
| Custom analog share | % | 0 to 100 | Your figure |
| Custom digital share | % | 0 to 100 | Your figure |
| Compiled memory share | % | 0 to 100 | Your figure |
| Standard cell digital share | % | 0 to 100 | Your figure |
| Library cells to port, 0 to use the area share | cells | from 0 | Your figure |
| Kind of change | choice | new-node, other-foundry, shrink, metal-stack, pdk-version | Your figure |
| Circuit work repeated | fraction | 0 to 1.5 | Your figure |
| Layout work repeated | fraction | 0 to 1.5 | Your figure |
| Verification work repeated | fraction | 0 to 1.5 | Your figure |
| Characterisation repeated | fraction | 0 to 1.5 | Your figure |
| Design rule similarity | choice | high, medium, low | Your figure |
| Similarity factor | x | 0.3 to 3 | Your figure |
| Migration tool available | choice | yes, no | Your figure |
| Tool factor on circuit work | x | 0.2 to 1 | Your figure |
| Tool factor on layout work | x | 0.2 to 1 | Your figure |
| Tool set-up | engineer-weeks | from 0 | Your figure |
| Team experience on the target process | choice | experienced, mixed, new | Your figure |
| Experience factor | x | 0.5 to 2 | Your figure |
| Characterisation corners | corners | 1 to 300 | Your figure |
| Test chip | choice | yes, no | Your figure |
| Test chip design | engineer-weeks | from 0 | Your figure |
| Test chip bring-up and measurement | engineer-weeks | from 0 | Your figure |
| Shuttle or mask cost | USD | from 0 | Your figure |
| Test chip fabrication time | weeks | from 0 | Your figure |
| Qualification | choice | yes, no | Your figure |
| Qualification effort | engineer-weeks | from 0 | Your figure |
| Qualification cost | USD | from 0 | Your figure |
| Qualification duration | weeks | from 0 | Your figure |
| Circuit share of the base | % | 0 to 100 | Your figure |
| Layout share of the base | % | 0 to 100 | Your figure |
| Verification share of the base | % | 0 to 100 | Your figure |
| Custom analog rate | engineer-weeks/mm2 | from 0 | Your figure |
| Custom digital rate | engineer-weeks/mm2 | from 0 | Your figure |
| Compiled memory rate | engineer-weeks/mm2 | from 0 | Your figure |
| Standard cell rate | engineer-weeks/mm2 | from 0 | Your figure |
| Rate per library cell | engineer-weeks/cell | from 0 | Your figure |
| Fixed cost per corner | engineer-weeks | from 0 | Your figure |
| Analog cost per corner | engineer-weeks/mm2 | from 0 | Your figure |
| Custom digital cost per corner | engineer-weeks/mm2 | from 0 | Your figure |
| Memory cost per corner | engineer-weeks/mm2 | from 0 | Your figure |
| Standard cell cost per corner | engineer-weeks/mm2 | from 0 | Your figure |
| Cost per cell per corner | engineer-weeks | from 0 | Your figure |
| Team size | engineers | 1 to 200 | Your figure |
| Team efficiency | % | 30 to 100 | Your figure |
| Loaded cost per engineer-week | USD | from 0 | Your figure |
| EDA and compute cost per engineer-week | USD | from 0 | Your figure |
| Project management | % | 0 to 50 | Your figure |
| Contingency | % | 0 to 100 | Your figure |
Outputs: Total effort (engineer-weeks); Schedule (months); Total cost (USD); Design effort as a share of a new design (fraction).
Assumptions
- Every rate and factor is a Hub editorial estimate. The public sources describe what is hard about porting, not how many engineer-weeks it takes. Calibrate the rates on a past port of your own before relying on the result.
- Effort is in engineer-weeks; a loaded rate covers salary, overhead and facilities, and a separate rate covers EDA licences and compute.
- Size is the area of the finished macro or a representative set of instances. A device count is converted to an area with an average devices-per-mm2 figure.
- The kinds of change are from the same foundry family or different foundries; the model does not separate a licensed re-use of IP from a clean re-implementation.
Uncertainty: The arithmetic is exact, but every factor is a Hub editorial estimate. Expect the real effort to differ by a factor of two either way until the rates are calibrated on one of your own ports.
Known limits
- A parametric model is a first estimate to within a factor of two. It cannot see an unusual specification, a new circuit topology or a missing model.
- Below about 180 nm and at FinFET nodes, analog circuits often need redesign rather than migration; a low similarity setting moves the estimate toward a redesign but cannot replace a design review.
- Migration tool factors are kept modest (circuit work at 0.7 and layout work at 0.85 of the unaided effort) because the published accounts say tool set-up and manual layout and simulation work remove much of the gain. Tool vendors claim more.
- Licence fees for the target process IP, foundry NRE, reticle sets beyond the shuttle cost, customer qualification and a production test programme are not included.
Where the method comes from
The Hub's own parametric effort model. Its factors follow the shape described in the sources (schematic mapping is the cheap part; layout, simulation and verification are not) but are editorial; no public source gives engineer-weeks for a port.
Reference example, checked by test: Worked example in this method note: the LDO and bandgap preset (0.15 mm2, 85% analog and 15% custom digital, another foundry, medium similarity, a migration tool, 12 corners, a test chip, no qualification, a team of 3), hand-worked in the tool's unit test. Total effort 65.8652; Schedule 8.59898; Total cost 409,326; Design effort as a share of a new design 0.5208.
Versions
- v1 (2026-10-25): First version, built from the tool idea.
Terms on this page
- Hard IP
- IP delivered as a fixed physical layout for one specific process, such as a PHY, PLL or memory compiler instance.
- Metal stack
- The set of metal and via layers a process offers for wiring, with their number, thicknesses, pitches and materials.
- PHY
- The block that converts digital data into signals on the medium and back, comprising the PCS, PMA and PMD sublayers plus clocking and analogue front end.
- Standard cell library
- A characterised collection of small fixed-height logic cells, flip-flops and buffers designed to abut in rows and be assembled automatically by synthesis and place and route.
- Design rule
- A geometric constraint a layout must meet to be manufacturable with good yield on a given process, such as minimum width, spacing, enclosure or density.
- Characterisation
- Systematic measurement of a device's real electrical and functional behaviour across voltage, temperature, frequency and process corners to establish the limits published in the datasheet.
- Test chip
- A chip made to prove designs, IP blocks or process features in silicon rather than to be sold.
- Standard cell
- A small, pre-designed logic cell, such as a gate, flip-flop or buffer, with a fixed height so that cells can be placed in rows and connected automatically.
- PDK
- The foundry-supplied package that makes a process usable by EDA tools.
- Bring-up
- The first phase of post-silicon work, in which a new part is powered on for the first time and taken step by step through power sequencing, clocking, reset, boot ROM execution and basic interface t...
- Shuttle
- A scheduled multi-project wafer run that a foundry or broker offers on a fixed calendar, with hard data submission deadlines.
- EDA
- The software industry and tool category used to design, simulate, verify and lay out integrated circuits.
Tell us what you're after and leave your email: we'll connect you with the right person on the Hub.