AMD Zen2,一个巨大 规模的巨无霸。采用的是先进的7nm工艺。
而一个更矮的6T的标准单元库,导致的更高的逻辑密度和实现难度。
相对于14nm,7nm导致线上电阻的增大。因此花了很多力气在net的优化上,例如,先择高层金属用于关键net的绕线(线宽减少主要是底层金属)。用via ladder降低线上的电阻以及解决EM的问题。
阅读本文,感受一下超大型芯片以及先进工艺所带来的挑战
Last year AMD introduced the Zen 2 microarchitecture, the company’s second major iteration of its Zen design. Zen 2 brought significant architectural changes at both the core level and the full SoC level. Undoubtedly, a big part of this success can be directly attributed to the company’s decision to dive headfirst into TSMC’s leading-edge 7-nanometer node. The move to TSMC 7-nanometer is a pretty big aggressive move, not only because it’s such a large technology shift, but also because of the new physical challenges it brings along with it – challenges that required the entire ecosystem – such as the foundry and CAD and EDA vendors – to collaborate on resolving those issues. At the recent ISSCC 2020 conference, AMD delved into some of those challenges in a little more detail.
Compared to AMD’s original Zen process which was the 14-nanometer FinFET process by GlobalFoundries, TSMC 7-nanometer represents a more-than-one generational leap in terms of density which provided the designers with significantly higher transistor budget per fixed area of silicon. Additionally, AMD switched from using a high-performance cell library to using TSMC 6T library which brought a whole new set of challenges especially when it comes to routing a high-performance SoC. We have covered both processes extensively on WikiChip so for brevity’s sake a short comparison is provided below.
Perhaps the biggest difference is the switch from a fairly tall cell to a high-density short cell. Those of you who are a bit more familiar with the previously-reported Zen numbers may notice that there are a couple of discrepanci