More Cores the Better Right?
Actually, it’s a bit more complicated than that…
You most likely have noticed that GPU performance have increased dramatically (and continues to do) year in and year out. However, the same cannot be said about CPU a lately. So why is that? Let me give a quick rundown.
Unlike GPU, CPU parallelism is hard to program for and not automatic. Where GPU manufacturers can increase the core count and get better performance all the time without any further programing or modifications. CPU processes need to be assigned to specific threads, which can get very complicated and will actually hurt performance if thread allocation is not optimized.
Overtime, software will get more and more multi-threaded. However, the problem is it is much harder or sophisticated when handling the CPU cores efficiently as opposed to GPUs.
So from where will performance come from for the CPUs? Performance improvements will most likely come from architectural improvements. With the improvements, the CPUs can squeeze out more performance for clock (hz). Increasing the clock frequency is also a guaranteed improvement, however due to the reality of physics, more clock is essentially means more oscillations which in turn means more heat 🔥.