What do think will be the speed difference in the Prometheus after the DMA update is complete (between cards and to processor cards) ?
Your questions are hard to answer, but I'll try.
First of all: the upgrade will mainly cover DMA issues, not speed issues, as the Prometheus design is already fast enough.
For PCI-PCI card speed, you can take the ElBugs phantasy value: 32bit x 33.33MHz = 128MByte/s.
PCI cards will never reach this limit, as you will only get it with a burst of infinitive length and no address phase. In reality it will be lower, as bursts are limited in length by the bus arbiter, and PCI cards will use different access speeds (DEVSEL# timing, and TRDY# timings).
Not to mention that PCI cards can ask the master for RETRY, which will also cost bandwidth.
For Zorro III - PCI speed: depends on the PCI card addressed. Memory on gfx cards is fast by writing but slow on reading (not done usually on gfx cards).
Most important is: the new CPLD designs are mixing PCI-PCI DMA and Zorro III accesses in a reliable way as fast as possible. You can have Picasso96 running (yielding in many Zorro III - PCI accesses to the gfx memory), with heavy PCI-PCI DMA at the same time. The DMA arbiter will even use the small time gap between two consecutive Zorro III accesses (originating in the Buster) for PCI-PCI transactions. ;-)
Michael
