Quote:
Originally Posted by roy bates
could you please explain a syncronous disign,do you mean cpu ,memory,or something else.or both?
i will understand if you explain fully.thanks.
|
"explain fully" would probably take a few months, so only a few spots of the "full explanation":
Memory and CPU on an A1200 accel are always syncronous. It could be done async, but that's more complicated and would have worse performance. I doubt that there was any accelerator for the A1200 where memory and CPU were async - I don't see why anyone would spend more money and get lower performance out of the board.
"sync design" of an A1200 accel always looks at the CPU clock vs. the A1200 mainboard clock: If an accelerator does not have it's own oscillator, but derives it's clock from the A1200 mainboard, it's a syncronous design. This has the advantage of not losing any time when syncing the "fast" CPU to the motherboard for example for Chipram access. ACA1230 is a sync design.
In an async design, the accelerator has it's own clock generator which runs totally independent from the A1200 mainboard clock. This loses a little time on chipram (and other mainboard) access, but it has the advantage of freedom for the CPU and memory clock speeds.
The ACA1231 and ACA1232 are async designs: The ACA1231 benefits from maximum memory clock speed (83,325MHz), which gives the shortest-possible first-access-penalty for a fastmem access. The ACA1232 perfectly matches multiple different CPU clock speeds, which gives the highest speed that the CPU was rated for by it's vendor.
You can't say that one or the other is "the perfect approach" - both have their advantages.
Jens