What hurdles do we have left to get to system-on-silicon consumer devices as we already have system-on-chip today?
In: Technology
That’s a very interesting question.
A complete system needs many different silicon chips. A typical PC has a CPU chip, a GPU chip, lots of DRAM chips on the memory modules, and probably even more chips to make up the SSD.
The total area of all of those chips is enormous. Larger than what can be manufactured on one piece of silicon at this time. Basically, we can make chips that are about 800mm^2. A top of the line GPU would be about 650mm^2 which doesn’t leave a lot of room for all the other stuff.
That’s where packaging advances come into it. We can’t make one giant chip but we can combine smaller ones together in one large package or module. A Ryzen CPU is a good example of that. It has a number of CPU chips and also an I/O chip, all in one package. The CPU chips need to be fast so they are made an advanced process (i.e expensive) but the I/O chip doesn’t need to be so fast so it can be made using an older, less sophisticated, and cheaper process.
There are various ways to combine chips like that. You can place them side by side on another piece of silicon that has basically nanoscale wiring to connect things together (which is how Ryzens do it), or you can stack them on top of each other (which is a much more expensive and advanced option).
So to answer the question, I’d say the cost of combining all the chips in one package is the major hurdle. It’s a cutting edge technology and not many manufacturers can do it.
Latest Answers