Guesses on M4 Pro base RAM configurations?

[mr_roboto]

Thank you all for your responses. I was expecting more 224 * 33.33 = 7,465.92. However, I believe the 7,500 figure is merely a rounded and marketable approximation.

Perhaps, but please note that 33.33 is definitely a rounded/marketable number - in reality it's far more likely to be 33.3333333333333... MHz. Or, if you like, 33 and 1/3 MHz. Once you calculate with the precise fraction, a multiplier of 224 results in 7466 and 2/3 MHz, and 225 resutls in 7500.0 exactly, no fraction.

You might think 33 1/3 MHz is a strange reference clock frequency, but flip it on its head (literally) - that frequency corresponds to a clock period of exactly 30ns. Since clock period is the inverse of frequency, many frequencies that must be expressed with an infinite string of fractional digits are simple integer periods, and vice versa. I mentioned 24 MHz as another common reference clock frequency, and it illustrates this - 24 MHz clocks have a period of 41 and 2/3ns.

 

[theorist9]

Perhaps, but please note that 33.33 is definitely a rounded/marketable number - in reality it's far more likely to be 33.3333333333333... MHz. Or, if you like, 33 and 1/3 MHz. Once you calculate with the precise fraction, a multiplier of 224 results in 7466 and 2/3 MHz, and 225 resutls in 7500.0 exactly, no fraction.

You might think 33 1/3 MHz is a strange reference clock frequency, but flip it on its head (literally) - that frequency corresponds to a clock period of exactly 30ns. Since clock period is the inverse of frequency, many frequencies that must be expressed with an infinite string of fractional digits are simple integer periods, and vice versa. I mentioned 24 MHz as another common reference clock frequency, and it illustrates this - 24 MHz clocks have a period of 41 and 2/3ns.

Of course, it also doesn't make sense to express it as 33.3333333333333... MHz, since that indicates a precision in the clocks that they don't have. When discussing this area specifically, it's probably better to say that 30 ns is the nominal period, and 33 1/3 MHz is corresponding the nominal clock speed. The actual period and clock speed varies due to jitter. For example:

Memory Output Clock Jitter Specifications

www.intel.com

 

[mr_roboto]

Of course, it also doesn't make sense to express it as 33.3333333333333... MHz, since that indicates a precision in the clocks that they don't have. When discussing this area specifically, it's probably better to say that 30 ns is the nominal period, and 33 1/3 MHz is corresponding the nominal clock speed. The actual period and clock speed varies due to jitter. For example:

Yup. In fact, it's incredibly common to implement something called "spread-spectrum clocking", where you add frequency modulation to your PLL. With this feature, the PLL's output frequency is still centered on nominal (or sometimes, nominal is treated as an upper limit), but the instantaneous frequency wanders around within some band (say, +/-10 MHz).

This is done to help systems pass the EMI (electromagnetic interference) tests mandated by national bodies like the US FCC. When you use a normal clock that doesn't wander, you get large, narrow-band EMI spike at the clock frequency (and smaller ones at each of its harmonics). Spread-spectrum clocking doesn't reduce the total radiated energy, but by spreading it out over a wider frequency band, the peak is much shorter, and that helps pass the test.