Jeremy_Sawatzky
Well-known member
- Joined
- Jun 4, 2009
- Messages
- 354
- Reaction score
- 0
- Points
- 0
- Age
- 36
Actually, is is VERY possible that they are in fact the same sensors and speed binned on how fast they can operate.
There is typically quite a bit of other support circuitry on the sensor die, and if they are really pushing the silicon speed, the EPIC S35 needs might be only a very small percentage of what they get out of fab. Typically extremely high speed chips target a single process corner to get the fastest operation there, and then the slower corners are used to produce a slower model.
It's not just the support circuitry in the rest of the system.
Chip design is done using statistical analysis of process variation. They typically take a 6-sigma approach around a typical result. This means there are probabilities involved in what you will get back on each wafer AND across each wafer. The distribution is a Gaussian distribution. The harder the requirements, the smaller section of that curve will produce a usable die. There is always a balance in chip design between wrangling in variance in your calculations or just taking a hit on product since if you factor in ALL variation possibilities, it would be impossible to design a chip that always worked in every die at high speeds.
So if their sensors were like 1 epic produced for every 10 scarlet, you can see that the epic, on cost value, is 10x since you have to produce 10x as many to get that number that work for your application.
i like how you put this. it makes alot of sense. its the same deal with cpus, i just built a new i7 rig and il tell you, i got one of the corner chips.