Why 10GHz CPUs Are Impossible Probably

#computerscience #cpu #processor
CPUs weren't always multi-core. In fact, it was only around 2005 that the industry shifted from single, high-speed processors to multi-core CPUs.
In this video, I go through the resons why CPU clock frequency couldn't go higher indefinitely. At some point, transistors - the basic building block of almost every electronic circuit in a computer - reached their physical and thermal limit. This limit is what forced the industry to move on and start designing multi-core CPUs.

SUMMARY
I this video, I explain the two principles that drove the computer industry to have such an incredible growth over the last half decade:
• Moore's Law: The improvements in manufacturing that led to the doubling of transistor count every couple of years.
• Dennard Scaling: The ability to lower the supply voltage of each transistor, as it shrunk in size.
Both of these principles allowed engineers to drastically reduce the power consumption of computer systems. Instead of creating ultra-low power computers, it was more beneficial to increase the frequency of the transistors and create much faster computers, without geenrating more heat.
This balancing trick broke down around 2005. Specifically, two things started to happen:
• Low Drive: The Supply Voltage had dropped so low that it almost matched the Threashold Voltage needed to switch a transistor between states. This causes them to become unstable.
• Leakage Current: Transistors were "leaking" electrons through the insulation because it was so thin (Quantum Tunneling). This caused them to waste energy as pure heat, even when the circuit isn't doing anything.
Both of these phenomena led to systems that had a hard time removing all the heat that was generated from the sillicon. This was the "Power Wal".
From this point on, CPUs were designed to be multi-core (multiple medium-speed processors packaged in a single CPU). The focus has shifted from single-processor latency to multi-core throughput.