Just wondering about the release/ship date comparison between the 40MHz part and the 25/50MHz part.
SUCKER! There's no such thing as a "25/50mhz" part.
The 25/50:33/66:40/80:50/100 "thing" as applied to Motorola's 68040 CPUs was entirely sleezy marketing, as I mentioned above. One technical difference between the 80486 and 68040 CPUs is the Intel CPUs clocked their internal bits at the same rate as the input clock, while the Motorola CPUs internal cadence required a clock input 2x the desired speed. Therefore a "33mhz" 80486 uses a 33Mhz input clock while the equivalent 68040 uses a 66mhz clock. This doesn't mean squat in terms of performance as no part of a 68040 that actually performs calculations "runs" at the double speed; in fact, Intel's *previous* design, the 80386, also had a 2x relationship between the input clock and the effective ALU speed. So...
When Intel came out with the *genuine* clock-multiplying parts, IE, the "DX2" CPUs, Apple had nothing to counter that with because, again, Motorola NEVER MADE AN EQUIVALENT PART. However, some sleezy marketing guy happened to notice that "double-speed" input clock and decided to start pasting it into advertising literature, and likewise of course the upgrade makers latched onto it as well. IT'S PURE HOKUM, and everyone involved should be ashamed of themselves. A "25/50mhz" 68040 is just a 25mhz 68040, period, and is *not* anything like a 50mhz DX2, which *legitimately* runs important parts of the chip, IE, the ALU/FPU/cache/etc, faster than other parts, IE, the bus drivers.
If Motorola *had* made real double-clocked 68040s then the "Quad Doubler" wouldn't exist; note that it has circuitry between the CPU and the socket that supplies the clock-doubling and bus arbitration functions. Similar socket upgrades for Intel 486s usually had nothing on them but voltage regulators, and that was only necessary if the upgrade incorporated something like a clock-tripling DX4 CPU. (Which used a 3.3v line for internal logic instead of 5v.)