Since the launch of the Ryzen 3000 processor last year, AMD has a 7nm Zen2 architecture that has a lot to do in the desktop, server and notebook markets. Other than that, on the core number alone, AMD can do 64 core 128 threads, the opponent side has only 28 core 56 threads.
What’s more, AMD not only doubles the core number, the price is much lower than the opponent, 64 nuclear EPYC dragon price is not more than 7000 dollars, (desktop version of 64 nuclear as long as 29999 yuan, less than 4000 dollars) less than half of the opponent, calculated that this is almost four times the price, how to do?
In AMD’s Zen2 architecture, AMD in the original MCM multi-chip design further, to use the chiplets small chip design, simply put, the CPU core and IO core separation, respectively, using different processes, the former is TSMC 7nm process, the latter is the GF 14/12nm process.
So what are the benefits of this design? At the recent ISSCC conference, AMD released data that compared 7nm Zen2 cost data in different core configurations.
First look at the desktop version, if the 16-core 32-threaded Ryzen 3 generation is the 100% benchmark, then the cost of a 16-core processor with native core sits will be more than 2, at least twice the cost.
If it is the EPYC Dragon processor, then the more core number, the more the cost advantage is obvious, 64 core 7nm rye dragon as the benchmark, then the cost of 48 core is 0.9, and the cost of the original 48 core design is at least 1.9, is still double the level.
Of course, with the reduction of the core number, the cost benefit will also decline, the cost of desktop 8 core is similar to the native core design, 16 core and above will have a more obvious cost reduction.
The reason is also very simple, because AMD’s small chip design, there is the IO core is 14nm process production, this part of the cost is more fixed, will not increase or decrease with the core number, so the less CPU core, the cost advantage is not obvious, the more core more obvious.