Are round CPU/memory frequencies less stable?

[geraltofrivia]

My physics professor today said that you rarely see round frequencies like 4000Mhz in CPU or memory speeds because of potential interference. He used the example of the bclk being 100.1MHz on the computer in the lecture hall instead of purely 100.0Mhz. Is there any truth at all to this? It sounds unbelievable yet my bclk is indeed 100.1Mhz and my VRAM defaults to 6008Mhz, not 6000Mhz.

 

[defaultluser]

My physics professor today said that you rarely see round frequencies like 4000Mhz in CPU or memory speeds because of potential interference. He used the example of the bclk being 100.1MHz on the computer in the lecture hall instead of purely 100.0Mhz. Is there any truth at all to this? It sounds unbelievable yet my bclk is indeed 100.1Mhz and my VRAM defaults to 6008Mhz, not 6000Mhz.

You don't see round frequencies because of build variations from one chip/motherboard to another. The source clock for most devices is multiplied many dozens of times (in some cases), which can result in speeds being off by several MHz. There are specifications detailing how much drift is acceptable, and as long as you hit within spec, it's fine.

The reason they multiply a base clock is because it is much easier to transmit a CLEAN clock signal at a lower speed and then use a PLL for accurate multiplication of the source. This is why your BCLK is only off by .1 MHz (that is a reference clock), but your processor clock will be 3603.6, +/- additional error from the PLL.

As for harmonics (what your professor was leading on to), they are part of life when you're dealing with digital circuits. Every clock you have is sourced from a base clock, so you're going to have to deal with the harmonics anyway. But designers get around this inherent problem by using Spread Spectrum for communications.