There are many factors that go into an overall performance result. If we just look at the CPU (Central Processing Unit, or primary processor chip), the first measurement is the number of times (or cycles) that the processor "fetches" - that is, it gets its instructions from memory to run. This cycle speed is measured in number per second, or Hertz. Kilohertz is 1000 per second, Mega million, and Giga billion, really simple. True, today's processors run, for example, at 2.6 GHz, and at first cut 2600 MHz/4.77 MHz=545. If that were the only factor, that would be accurate.
However, those old Intel and Motorola processors ran at 8 bits wide, while modern ones are 64 bits wide. That is 8 x the throughput per cycle.
But wait - today's processors are hyperthreaded with multiple cores. That is, they run as if there are multiple processors, because they are so fast that they outpace the ability of the system to feed data to them, so they are split up into cores that run in parallel. An Intel Core i5 processor has 4 cores, thus is running 4 of those. So we have 545 x 8 x 4 = 17,440 times as fast as that little 8086 Intel running in a single core at 8 bits per cycle. [EDIT: Corrected number of cores and calculations]
Still, I have this gut feeling that I am leaving out some other multipliers. Sure, all the supporting systems (memory bus, disk bus, etc.) have to be upgraded to keep up with the processors. But I think it is actually perhaps one or two more multipliers here, so it could be an even larger factor.
No comments:
Post a Comment