This 2022 is the year we will finally see, if there is no delay, the launch of the 5nm AMD Zen 4. However, this is not a new crafting node, as Apple has been using it for quite some time. Why is AMD not using a more advanced fabrication node for its Zen 4?
Lisa Su’s announcement of the Ryzen 7000 based on the Zen 4 architecture has us biting our nails, since we’ve been wanting to get our hands on AMD’s next big CPU architecture for a long time. At the moment we have to wait a few months, but little by little new details are leaking.
New information on the AMD Zen 4
During the last CES, AMD in its conference confirmed to us that its Ryzen 7000 desktop processors will be released in the second half of this year and will be based on the Zen 4 architecture, which will bring, among various new features, a significant increase in performance thanks to to improvements over the architecture that sought to achieve a higher ratio of instructions resolved per cycle.
Something we’re very clear on is that AMD will be repeating the design in chiplets that started with the Zen 2 based Ryzen 3000 and continued with the Zen 3 architecture based Ryzen 5000. So it shouldn’t there would be no change in this aspect and therefore the tradition of the last two Zen generations will continue, i.e. each processor would be a 2.5DIC configuration with one or more CCD chips which include the cores and a IOD for the rest of the processor. All of this mounted on an interposer with up to 2 CCDs for the Ryzen and up to 8 and a more advanced IOD for the EPYC and Threadripper models coming out by 2023 for the server and HEDT markets.
Another thing we’ve seen confirmed by AMD’s CEO is that of the 3DIC models
Why isn’t AMD using the 3nm node in Zen 4?
CCD Chiplets are of a size that does not even reach 100 mm2 On average, due to their dimensions, they could be built using the most advanced manufacturing nodes and since the Ryzen 7000 launch will coincide with Apple’s premiere in TSMC’s 3nm, the question is why is AMD using he Zen? 4 to 5nm instead of a more advanced node?
The reason for this is very easy to understand, it is not the inherent limitations of the 3nm node, but rather each node manufactured by TSMC has two different variants. One optimized for high-performance systems and the other focused on energy efficiency. Although AMD may use the 3nm node that Apple will use to build Zen 4, it would not get a processor with the same performance as using the 5nm node they chose. And it is that the nanometers are an approximation of the number of transistors per area, but not of the performance that can be obtained.
Let’s not forget that one of TSMC’s key partners for the 5nm node is AMD, which collaborated to build the Full Custom Libraries for the design and subsequent construction of its processors. Remember that the whole process of designing, validating and manufacturing a processor takes many years and that changing the process can mean a fatal delay for the processor.