Was the Gekko influenced by the Intel 486?

[Lumina333]

I wonder if the designers of the Gekko were fans of the Intel 486, because the Gekko is clocked at 486 MHz...

Maybe they were being sneaky trying to hide the fact that they were being influenced by x86 at the time, because they were working on extending a PowerPC processor, so they hide this little easier egg in the clock speed...

Or, it's just a fun little coincidence... Why else choose that exact clock speed?

[Extrems]

It's a coincidence. The bus clock (162 MHz) is 3x the 54 MHz video clock, and then the core clock is 3x the bus clock.

[Lumina333]

Awww.... Do they just like /3 clock dividers instead?

[Papy.G]

It probably is a choice for convenience or maybe optimisation. In the power PC line, there used to be all kind of dividers, on Macs, for example, from 2 to 7 and even more on third-party accelerator G3 cards, I think I 've even read about non-integer divider on the 604ev daughterboards, don't ask me how it worked.

[Extrems]

Awww.... Do they just like /3 clock dividers instead?

Well, the alternatives were 2x (324 MHz) which is probably too slow, and 4x (648 MHz) which would probably be marginal on this silicon.

[Lumina333]

You mean the extra speed of 648 MHz wasn't worth the power/cooling requirements? I guess. The GameCube isn't a silent console.

[Extrems]

648 MHz might not work reliably on all chips.

[Lumina333]

Ok, so it's more like: bad idea to overclock it because problems can happen.

[Lumina333]

One official source that says 485 Mhz:

https://www.nintendo.com/en-gb/Hardware ... 27134.html

That could matter for TAS verification/accurate emulation - dunno if anyone has ever measured it to more significant digits.

[Extrems]

It's just plain wrong.

[Lumina333]

I'd have to agree - also at the bottom of page 1-15 of the Gekko User's Manual (under 1.2.9 Clocking):

The feedback in the PLL guarantees that the processor clock is phase locked to the bus clock, regardless of process
variations, temperature changes, or parasitic capacitances.

That's good enough for me - it would go out of phase with a multiplier of slightly less than 3.

[Lumina333]

I'm finding more places that say 485 but I think we have bigger fish to fry than chasing down all the authors.

[Papy.G]

Ok, so it's more like: bad idea to overclock it because problems can happen.

You have to see it the other way: the CPU can support a said frequency, the question is: what's the cost/benefit ratio of getting slower Ram with higher divider, wether getting faster Ram with lower divider. If most of the instructions take three cycles to complete, you'll be wrong with 2x divider, you'll lose 1cycle waiting for the next fetch, with 4x, you'll lose 1 cycle too.
That's a rough example, ignoring cache works, pre-fetch, and predictive computing (which last was ripped off the Gecko as well as the contemporary G4s, AFAIK).
You have to keep in mind that memory cost has always been very critical in computer manufacturing, It is even more true in consoles where the RAM is usually fixed with no expansion possibility.

One official source that says 485 Mhz:

https://www.nintendo.com/en-gb/Hardware ... 27134.html

That could matter for TAS verification/accurate emulation - dunno if anyone has ever measured it to more significant digits.

That's not as if they made the Super GameBoy not run at the DMG's actual speed.
I don't know if speedrunners would see a 0.2% difference anyways. Still if you have the base crystal speed, then the multiplying rate of the different stages, you'll be able to say the speed of the system, no need to scope anything.

That's good enough for me - it would go out of phase with a multiplier of slightly less than 3.

No, it's just that everything is tied to the same sync base, so you won't have to check and wait for an operation to complete on one of the chips in the system, you can just predict it and deliver data or commands just in time.