0

I understand that highly integrated semiconductors enable physically diminutive devices. Microcontrollers with CPU, RAM, flash, power supply, a wide assortment of peripherals, all on a single piece of silicon are available for very cheap.

Why do our devices like laptops and phones still use multiple chips and so many passive components?

I understand capacitors and inductors are expensive to implement on silicon, then why not a single chip plus a bunch of those passives?

crossroad
  • 1,390
  • 7
  • 19
  • 1
    Not all things use the same silicon process and different manufacturers specialize in different things. Also, you can't take amortize expensive development costs if you have something too specialized; You might sell millions of CPUs and millions RAM chips, but not millions of phones. And the more you try to do the lower your yield. – DKNguyen Mar 04 '21 at 05:11
  • There are numerous reasons for this but the first that comes to mind is pure cost and material efficiency. Already, a high-end CPU or high-speed RAM chip is complex enough that it's considered great if even 10% of the dice work, but trying to put both of those on the same chip would make the yield far worse, as the chip wouldn't be usable if either of them was broken, even if the other worked flawlessly. – Hearth Mar 04 '21 at 05:13
  • If a company was large enough to implement such a design entirely by themselves, I suppose they may consider it. Of course, not all processes are the same or have the same number of steps to them. Pushing towards integration, it is expensive to create drivers and receivers for external pins that are connected to unknown sources and loads. It takes up die space as well as packaging costs. But the balance remains where it is at, today. On rare occasion, you may see some company implement a more integrated design. But the uniqueness of it will have to be supported by its volume and other factors. – jonk Mar 04 '21 at 05:14
  • In time, particularly with the advent of RISC-V, you may see more integration happening. Keep your eyes peeled. – jonk Mar 04 '21 at 05:17
  • If they could they would. The knowledge and ability to create all the parts is extensive. Modular designs are easier to implement. – Passerby Mar 04 '21 at 05:24
  • Regarding phones particularly, the fab process for flash and the fab process for a processor are pretty different, so it may just be more efficient to keep the flash and processor separate. The same could be said for power management. High current and or high voltage MOSFET's take up a lot of chip area. The fabs that make these parts are lower cost fabs compared to state of the art processor and flash memory fabs. It might be more costly to implement them on high cost silicon. – user57037 Mar 04 '21 at 05:25
  • 1
    Why don't you wear single piece of cloth (like [pink guy](https://filthy-frank.fandom.com/wiki/Pink_Guy)), but wear different accessories (like, t-shirt, shoes, pants, etc)? Our technology level must be high enough to make a single piece of dress enough to satisfy one person. – hkBattousai Mar 04 '21 at 05:35
  • In fact, for the reasons others mentioned, there is a trend to less than maximum integration by the use of chiplets- multiple dies within one package. – Spehro Pefhany Mar 04 '21 at 05:43
  • 1
    there are *many* kinds of silicon processes. The silicon for an opamp is different from the one needed for a cpu and also different for the one needed for flash memory (which is one of the reason flash is expensive on chip and cheap standalone). – Lorenzo Marcantonio Mar 04 '21 at 08:45

0 Answers0