If you are incredibly fortunate, you might find you already have an 8372AB fitted or 8372B.
The 8372AB is the 8372B, it's just the circuitry of the 8372B inside the packaging of the 8372A, so they separately suffixed a standalone "B" to signify it's a 8372B.
However, it's impossible to encounter such an Agnus in a production A2000. The 8372(A)B Agnus offers a single RAS line while the RAM design of the A2000 requires 2 RAS lines from the Agnus.
On the A3000, they solve this by recreating the separate RAS lines by combining the information from the RAS line and the two highest order address bits. But such circuitry doesn't exist in the A2000.
Even if the VA2000/CX does work with a PAL megachip, external NTSC equipment will not work, even if the machine is soft switched into NTSC mode (because then the megachip has issues switching an 8375 from PAL to NTSC, which is why there are 2 versions).
If this isn't accurate, then please tell me why there are 2 different versions of the megachip, PAL and NTSC. Just default timing settings?
Yes! No Agnus has the capability to auto-detect the master oscillator frequency! So designing a "plug and play" universal megachip isn't possible.
If it could autodetect whether the motherboard is equipped with a 28.63636 MHz (NTSC) oscillator or a 28.37516 MHz (PAL) one, it could of course also be designed to automatically adjust its bootup timings accordingly. But this would require an independent internal clock inside the Agnus and extra comparator circuitry, which is an unnecessary complexity.
So when you set an Agnus to PAL timings, you're simply more or less just telling it "look, I assure you that the master clock is 28.37516 MHz and I want PAL timings, so each scanline will be 1816 ticks long, even fields will be 313 scanlines high and odd fields will be 312 scanlines high, VBLANK time will be 25/24 scanlines and I want hardware enables/stops at ticks #X and #Y for each scanline"
Similarly, when you set an Agnus to NTSC timings, you're simply telling it "there's a 28.63636 MHz master clock so I want NTSC timings, so even fields will be 263 scanlines high and odd fields will be 262 scanlines high and scanline length will alternate between 1816 and 1824 ticks etc etc"
The above instruction set for example correctly produces the NTSC interlaced field pattern:
-first field (#0) is even, so it has 313 scanlines (odd number), so it begins and ends with a short line (1816)
-second field (#1) is odd, so it has 312 scanlines (even number), it begins with a long line (as field #0 ended with a short line) and thus ends with a short line
-third field (#2) is even, so it has 313 scanlines (odd number), it begins with a long line (as field #1 ended with a short line) and thus ends with a long line
-fourth field (#3) is odd, so it has 312 scanlines (even number), it begins with a short line (as field #2 ended with a long line) and thus ends with a long line
... and the sequence repeats, i.e. fifth field (#4) is identical to #0 etc, so we get the periodic sequence:
- long field ending on short line
- short field ending on short line
- long field ending on long line
- short field ending on long line
Given all that, it's impossible for an Agnus to 'refuse' to be set to PAL timings if an NTSC oscillator is presence, and vice versa.
This leads to the 1% faster "nearly PAL" (when oscillator = NTSC but Agnus timings are PAL)
and 1% slower "nearly NTSC" (when oscillator = PAL but Agnus timings are NTSC) timings, which may well be video-hardware incompatible.
But these are the only video-illegal modes.
Any NTSC video hardware finding an Amiga with an NTSC oscillator and an NTSC-timed Agnus (regardless if it was set via jumpers or soft-switched via the KS 3.x menu) will work just fine.
