2023-10-26 18:32:38
A speed gain is what everyone expects when purchasing a new computer or a new phone. But how do Apple and other manufacturers speed up their processors year following year, or at least try to? After explaining the challenge of architectures in the first article in our series “The race for CPU power”, let’s take a look at the manufacturing of chips themselves.
The race for CPU power ????️ A history of architectures Developments engraved in silicon Apple’s resounding success (to come)
Engraving is seen by some as the Holy Grail in developments, but this is rarely the case. Let’s go back to the basics: when we talk regarding the fineness of the engraving, we give a value in nanometers (nm) which is a marketing indication of the size of the transistors. The lower the value, the smaller the transistors and the more transistors it is possible to put on the same surface. Furthermore, even if the result is less linear, an improvement in the engraving finesse generally allows lower consumption at the same frequency, and this is important.
The size of wafers (the surface on which the chips are etched) rarely scales, so improved finesse means more transistors on the same surface.
A simplistic thought would be to think that CPU designers are taking advantage of improved finesse to reduce chip size and power consumption. But as we will see below, it is the exception rather than the rule. In practice, manufacturers go in the opposite direction, taking advantage of additional transistors. To take the case of Apple, the surface area of classic A chips (those of the iPhone) is between 85 and 110 mm² in general (the A10 Fusion of 125 mm² is an exception) while the engraving has gone from 32 to 3 nm between the A6 and A17 Pro. What are these additional transistors for? There are many answers.
The possible paths depend on the CPU manufacturers but also on the evolution of the engraving fineness, which tends to slow down over time. When engraving progressed steadily, Intel put forward its model Tick / Tock. In what Intel called a Tick, the idea was to keep the same architecture and take advantage of the available transistors to add cache memory or optimize a possible GPU, for example. Some also took the opportunity to add cores to the chip, while keeping approximately the same physical size. A Tick was seen as a minor development, even if the modifications might bring interesting gains.
This article is reserved for members of the iGen Club
The iGen Club is:
Exclusive articles No The Kernel Panic podcast Unparalleled reading comfort
This article is reserved for members of the iGen Club!
Join the largest French-speaking Apple community and support the work of an independent editorial staff. The iGen Club is:
quality items just for you! an exclusive podcast! a dedicated discord! no ! a dedicated site!
This article is reserved for members of the iGen Club
The iGen Club is: |
Support independent writing
This article is reserved for members of the iGen Club
The iGen Club is:
Exclusive articles No The Kernel Panic podcast Unparalleled reading comfort
1698351214
#race #CPU #power #developments #engraved #silicon
