Why is Silicon getting “Dark”?

Dark silicon refers to an area that cannot be turned on by supplying power on a silicon chip due to power consumption restrictions. It is an area that cannot be turned on at the same time as the Central Processing Unit (CPU), and it becomes a wasteful area that cannot be used without doing anything. [1] As the process of miniaturization progresses, a dark silicon area will be generated on the die (semiconductor body), and the area will increase with further miniaturization. The reason is that the power consumption of the circuits on the chip does not scale down as much as the miniaturization of the process. [1], [2] Even if the CPU core becomes smaller with miniaturization, the power does not become smaller enough to match it.

Therefore, the number of CPU cores that can be mounted on the chip gradually decreases. This is the dark silicon problem. The word dark silicon has been one of the buzzwords in the semiconductor industry past decade.

It is reminiscent of the "dark matter" of astrophysics, but the darkness is about power. At the 2010 processor conference "Hot Chips," the term was used when the University of California San Diego announced the "Green Droid" chip. Widely used was the paper entitled "Dark Silicon and the End of Multicore Scaling" at the computer architecture symposium "International Symposium on Computer Architecture (ISCA '11)."[3] As you can see from the title of the ISCA paper, dark silicon stands in the way of multi-core scaling.

In other words, the dark silicon problem makes it impossible to increase the number of CPU cores with multiple cores. At least bring the end of the symmetric multicore era. Performance PCs and CPUs for servers have been increasing the number of cores, but that era is about to end.