There are many factors responsible for AMD’s recent success in the consumer market. But, the chiplet or MCM design (Multi-chip Module) is easily at the heart of it. It allows AMD to increase the core counts to never-before-seen figures in the consumer market. The Ryzen 9 3950X features 16 cores while the Threadripper flagship, the 3990X boasts an insane 64 cores, same as the Epyc Rome chip. This is all thanks to the chiplet approach. It allows the BOM to scale linearly, not exponentially (as they do with Intel CPUs) with core counts. This means that, at any given price-point, AMD can deliver more cores, more threads, and better multi-threaded performance, than Intel can, even after a number of Team Blue price cuts.
More on this here: Intel vs AMD Processor Architectures: What’s the Difference between them?
How does Ryzen’s modular design work: CCX and CCD
These two functional units lie at the heart of AMD’s modular approach to Ryzen. The basic unit of a Ryzen processor is a CCX or Core Complex, a quad-core CPU model with a shared L3 cache. In newer Ryzen 3000 parts, the amount of L3 is higher and it’s referred to as “Gamecache.”
There are pros and cons to having the CCX be Ryzen’s basic functional unit. A negative is that the baseline cost of manufacturing can be somewhat on the high end since AMD needs to pay up for a minimum of four cores. However, this is offset by the fact that Team Red salvages partially functional CCDs with, say, two or three functional cores, to create different SKUs. For example, the Ryzen 5 3600 features two CCXs, each of which has one core disabled, for a total of 6 functional cores.
However, while CCXs are the basic unit of silicon dabbed, at an architectural level, CCDs or Core Chiplet Dies are your lowest level of abstraction. A CCD consists of two CCXs paired together using the Infinity Fabric Interconnect. All Ryzen parts, even quad-core parts, ship with at least one CCD. They just have a differing number of cores disabled per CCX.
Keeping this in mind, it’s a lot easier to see the biggest benefit: scaling. Intel uses what’s called a monolithic approach to CPU design. Every CPU it makes has a discrete design, with a certain number of cores. When manufacturing takes place, all cores in a particular design need to be fully functional. Intel simply throws away partially functional parts. For smaller dual and quad-core processors, this makes a lot of sense: it costs less to manufacture a dual-core processor.
However, since silicon yields are never 100 percent, costs increase exponentially with larger designs. When you get to 10 cores and more, it’s virtually guaranteed that for every functional CPU, Intel discards at least one defective part: keep in mind, only one core needs to be non-functional for the part to be scrapped. In practice, this means that parts in Intel’s Xeon HPC lineup exceed five figures. It also means that there’s an upper limit to the number of cores Intel can viably implement in a single design: Xeontop outs at around 28 cores while AMD offers as many as 64 cores with Epyc Rome.