Get email delivery of the Cadence blog featured here
Over at Cadence's Industry Insights blog by Richard Goering, he has a great writeup of a panel at the Virtual SoC Conference entitled "Are SoC Development Costs Significantly Underrated?" In it, there was a great analogy comparing a chip design project to a home remodeling project. This is an analogy I see lots of potential for. But let's first focus on that cost estimation.
The panel's focus was on development costs, so I won't spend any more time on that. But what about materials cost? You wouldn't embark on a home remodeling project without calculating that. In our case this is the cost of silicon and packaging. This is even more significant for chip design because it is a variable cost that is realized across larger and larger volumes.
Most companies do not ignore it. They calculate it with some kind of usually complex spreadsheet. These efforts are admirable and represent a large accumulation of knowledge, even if the people who provided that knowledge are no longer with the company. Yet there are some drawbacks to this method:
Spreadsheets are backward-looking. This is by-nature - the knowledge that has gone into the data is based on past experience. If you are moving to a new process, foundry, IP vendor, etc, you are in uncharted territory.
The knowledge is limited to your scope of experience. Again, this is by-nature. How do you know that there is not a more well-suited process flavor or memory architecture for your chip's goals?
The model is manual and static. If you want to see what happens to power if you shut down a power domain, your spreadsheet won't ask you if you are forgetting to add isolation and state retention, and then calculate the area overhead associated with these (if it does, you might be spending way too many resources on Visual Basic development).
Spreadsheets can't seed the design and implementation process. This may seem like a small detail, but when you spend time coming up with a power architecture, it helps prevent human error when you don't have to manually re-generate your intent.
This is why we see more and more customers adopting our InCyte Chip Estimator. Yes, there is a lot of institutional knowledge and spent costs associated with the spreadsheets, and that is not lost as you move to a more automated method. After all, your typing classes on a typewriter are still useful today with word processing!