LoadWB,
I'm not sure which CPU boards you're refering to exactly. But, if you're refering to the Dragon (one board, not many, made by Elbox, more info on their site) which uses the Coldfire CPU, well then from my understanding it's sitting on a PCI card. You take this card and plug it into an existing "PCI solution" (Elbox Mediator 4000Di for example, or Prometheus or G-Rex, even though I don't know if the last two will be supported by Elbox), which itself is sitting on top of the existing existing Zorro bus of your Amiga. Therefore performance between native chips (OCS, ECS, AGA and the Coldfire CPU won't be so hot, I think ZII caps at 10MB/sec?), but performance between the Coldfire and the other *PCI* cards will be as fast as if it was on a typical PC system (ex: 32bit PCI @ 33Mhz = 4bytes*33Mhz=132MB/sec max theoretical transfer rate, which you won't ever achieve
...unless there's some burst or double-clocking magic that these PCI/CPU gurus can pull off? or perhaps with faster PCI like 66Mhz or 64bit).
Anyhow, the end result is you get backwards compatibility with your old, make that "original", Amiga hardware and new capabilities with endless PCI hardware. Of course no AGP and no PCI-X
So we still can't catch up with present day PeeCeez and Macs.
A couple of questions of my own for gurus:
Many modern day PCs rely on a "southbridge" chipset to do the PCI handling, meaning that there is a bottleneck between the PCI cards and the CPU or more importantly (for DMA) memory! Does having a Dragon with a Coldfire on the PCI bus mean that we can "skip" the southbridge chip completely and let the PCI devices access memory (on the Dragon, I assume, or another PCI card), at much higher speeds? Although now that I think about it, I don't see how, as the PCI itself is limited (ex: 32bit x 33Mhz or 64bit x 33Mhz or 32bit x 66Mhz, etc), while the CPU to memory interface is MUCH faster: 64bit x 133Mhz (for PC133 FSBs) or 64bit x 400Mhz (for DDR-200Mhz double triggered FSBs) or 64bit x 667Mhz (for quad-pumped FSBs), and if you through interleaved memory access or dual-channel (NVidia motherboards) you double from 64bit to 128bit, which is MANY GB/sec, not a measly 132-528MB/sec for PCI.
Clues anyone?