For our viewers that often retains monitor of Intel’s product portfolio, it will be arduous to overlook that the naming technique of Intel’s course of node applied sciences is a little bit of a large number. To some, these phrases are themselves an understatement, as Intel has shifted its naming technique 2-3 occasions for the reason that launch of Intel’s first 10nm merchandise. Not solely that, even Intel’s numerous departments internally have a tough time holding monitor of ‘what is this manufacturing process being called today’ when the press like AnandTech ask for particulars on the newest upcoming merchandise.
Knowing this, and realizing what points Intel has been having, I wished to demystify Intel’s manufacturing course of naming scheme such that customers and engineers alike, even when they’re inside Intel, can perceive what’s what but in addition importantly why. The why is the essential issue.
If you are on the lookout for a useful decoder ring for Intel’s 10nm Products, it is right here in web page 3.
Why Do We Have Multiple Versions of a Process?
With Intel’s 14nm, we had been invited to 14nm, 14+, 14++, 14+++, and should you consider Intel’s personal slides, there have been variants that went past this ++++ naming scheme. Each a type of further + factors on the tip of the title signified a change within the course of know-how – often to help for rising efficiency or effectivity.
Each one among these + factors is an replace to the BKM, or Best Known Methods.
While an engineer can draw {an electrical} layouts for part of a processor, reminiscent of an addition circuit, really making use of that design to a silicon floorplan for manufacturing is a distinct ability altogether. Transistor libraries are designed to reap the benefits of a given course of, and when a floorplan is optimized for a course of, it will probably then be pasted and repeated as obligatory – on prime of this, simulation on thermals, energy, and present density are utilized to make sure that there aren’t any hotspots or that important paths contained in the design have as few bottlenecks as potential.
When an replace to the BKM happens, two issues can occur. Normally we see the replace on the extent of the transistor library that’s modified – if the gap between two fins on a transistor will increase for instance, the transistor library and the macros could also be made greater, after which the floorplan could be redesigned to take account for this. As for any course of node design, there are 100 completely different controls, and enhancing one may make three different controls worse, so it’s a advantageous balancing act. Not solely this, however the BKM must be validated on the manufacturing degree. The BKM replace might apply to the metallic stack as nicely, which in of itself can regulate the efficiency.
In the lengthy gone, BKM updates had been by no means marketed externally. If Intel or TSMC or one other foundry found a method to enhance the efficiency, or lower the voltage, or enhance the yield, the replace was silently rolled into the design and nothing a lot was fabricated from it. Sometimes processors could be listed as ‘1.0 volts to 1.35 volts’, and it will simply be a roll of the cube if a consumer obtained one of many decrease voltage fashions.
However, as time between completely different course of node updates has elongated, these BKM updates have began to be recognized and successfully monetized by the semiconductor firms. An replace to a course of that improves the voltage by 50 millivolts and will increase frequency by 200 MHz instantly turns into a productizable occasion, and merchandise constructed on these updates could be provided for more cash over the standard. Or, relying on the speed of updates, the entire subsequent era of merchandise could possibly be constructed on the replace.
So we by no means noticed BKM updates formally introduced at Intel’s 45nm, 32nm, or 22nm course of nodes. These updates had been quick sufficient that…
