I have come across many people who just can’t figure out exactly how is the multicore architecture better than the single core. A more important question among them is, how is a single core 2.55GHz single core [e.g. pentium 4] processor is not better than a 1.6GHz Dual-core, Core2Duo etc. processors.
Well it’s certainly a very strong point that, the clock frequency ultimately decides the processor speed. Higher the clock frequency, better the processor performance. But still we moved to the multicore architecture even if the clock frequency is not necessarily greater. WHY?
The reason is simple.
The higher the frequency, the faster the processor.
The faster the processor, greater the data speed.
The greater the data speed, more the heat generated.
As the generation of fabrication technology advances, the transistors become smaller the current processors are built over 32nm (µ=32nm >Sandy bridge & λ=32nm > in R&D) fabrication technology. The smaller the transistor get, the less power is required and the less heat they can sustain.
So, we can’t just keep increasing the clock frequency. As the generation of fabrication advances, the transistor density increases in the processor. The processors become smarter and more capable. But the supported clock frequency may not increase relatively.
So far, it is clear that processor clock frequency can’t just be increased to achieve higher efficiency. Now, comes the multicore architecture. The most common idea about multicore processors is that, they consist of multiple processors cores in the same board, thus gives twice the performance than single core. Well it’s not entirely wrong. Lets take a Dual Core processor for understanding. In a Dual Core processor, there are 2 cores embedded on the single core environment. The cores share the same BUS system. Thus if one core performs at its 100% the other can at max achieve 80% of its processing capability (as per the processor architecture).
This is an example shown here. In the environment, there are 2 cores each of 3.6GHz core speed, sharing the same bus system. The BUS structure is optimized to a level to achieve 80% of max performance while the other core delivers 100%.
If we increase the number of cores, the effective performance will be relatively lesser than an architecture with lesser number of cores, though the overall performance increases.
Consider a system with 4 cores. A Core 2 Duo processor. In this type of architecture 2 dual core processors are scaled together to give a 4 core based processor system. The dual cores a are designed in the fashion described earlier. Here dual core fabrication is just a part of the process. The dual cores achieved 80% SOSS, resulting in an 180% processing power compared to a single core. The 2 cores are embedded together in a single processor architecture. Here the SOSS remains 80% but the COCS is 70% for both the dual cores. If the number of cores are increased under the same architecture, with each layer of scalability the efficiency will get relatively reduced. Standard Core2 Quad systems were built in an architecture similar to Core 2 Duo, but the Xeon Quad cores are designed in a more efficient way.
So this was the general architecture of multicore environment processors. Though one must be aware that, multi core is not all that makes the new generation processors so much efficient.
I will post, in the upcoming weeks, about the latest Core i series processors, 2nd generation intel processors and various technologies that help increase performance dramatically, e.g. HTT, AVE, DFT, ITB etc.
Thank you for reading.