You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
What kinds of effects do oyu think this could have on the mcu market(and general electronic design) , both from the technical perspective, and from the business perspective ?
That's an interesting article about the startup named Indie, yes I do think that they have an opportunity to differentiate themselves in the MCU market by using a multi-die packaging approach (aka More than Moore). They can be more flexible in their offerings because of how they arrange pre-verified chip features, and my only question is on costs as a function of volume.
If their marketing message is unique, and their value proposition is accepted, then success must follow.
Daniel: Regarding costs: In their site, they talk about a use case with mcu that costs less than $0.2(high volume) and includes sensing and display . Sounds too good to be true.
I suppose it boils down to whether a one-chip SoC in 40 nm (for example) would not be a better deal than two dies (digital+mixed signal) on different (e.g., 90-180 nm) processes. You would get better power/performance on 40 nm. Supplier management (including second-sourcing) would also be simpler if all the functions came from one foundry.
Fred , i think the value is customization. For example, unless you have large volumes(probably more the 5M), you cannot design each mcu in 40nm, for ex.
But say you need a standard 40nm mcu+peripherials , but some unique analog(done in the best analog process) and some unique peripherials , for mid volume design, SIP would work better. That could be a decent share of the mcu market. And there are performance benefits of optimizing different processes separately, so maybe this will fits the large volume market
I cannot dispute the general appeal of SIP. But some combinations of functions are just better done as SoC. The best or most familiar example is logic and memory. Taking memory off chip would make the logic wait longer. Extending to analog, maybe that's the particular focus here, but it seems it will depend on how fast the particular analog signals would be coming to be processed by the digital logic.
