# Why do PC bus frequencies often end in multiples of 33MHz?

It seems like there are a lot of frequencies for different system buses that end in 33 or 66MHz. Is there a particular reason for this?

Once you have one subsystem operating at a multiple of a particular crystal frequency, it's convenient to use the same basis crystal for generating other frequencies. This changes the line of inquiry to "what was the first use of 33MHz in computers?", which may be some 386 processors or the PCI bus.

It's possible that the first supply of 33MHz crystals was for some radio purpose, but googling is not suggesting a ready answer.

• This would suggest that either fsb133 is really 132 or 133.33333 – PlasmaHH Nov 23 '15 at 17:40
• Quite likely it is.. – pjc50 Nov 23 '15 at 20:11
• crystals ar mainly used a a reference clocks are usually come from a PLL oscilatior, thus overclocking is just a mteer of sendin some commands over the system management bus. – Jasen Nov 24 '15 at 7:51

The exact number is not 33.0000MHz, it is 2^25 Hz, in lower power CPUs, it is 2^15. It is easier to implement a power of 2 counter than other arbitrary number.

Edit: After digging into the PCI spec and the 486 datasheets, I didn't find any hard requirement for clock accuracy. The PCI spec allows operation from DC to 33MHz and the 486 requires only 0.1% stability from its clock source.

• Oh, awesome! Where does the 25 come from then? – zymhan Nov 23 '15 at 16:33
• Do you have a reference for that? – jippie Nov 23 '15 at 16:45
• 2^15 is real, though rarely for CPUs per se, but for real-time clocks. 2^25Hz I have never seen, though Farnell certainly list 33.0000MHz crystals... – user_1818839 Nov 23 '15 at 16:50
• The clock source is usually a quarz crystal, and 33.0 MHz is a common frequency where cheap crystals are available. – Turbo J Nov 23 '15 at 17:15
• Multiply by 4, round up for marketing. – Lior Bilia Nov 23 '15 at 18:55