Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: Jose on May 13, 2004, 10:36:26 PM
-
Hey all. I'm defenitly not a hardware guy but I understand some concepts. Following some Zorro3 discussions in the past I got the impression that the only reason for the Zorro3 bus not to approach more the 100MB/s limit it's the Buster chip. Is this correct?
PCI must have some bus arbitration too right? So does that mean that PCI's "Buster" is much better, or the implementation is very different?
I'm asking this cause the theoretical speed limit of both isn't very different.
-
Would it be a case of the Buster being a 'passive' logic array, wearas a PCI controller is usually a small processor?
-
I love the way the ZorroIII figures go up every year ... by 2010 it'll be up to 10Gb/s ;-)
I think you'll find it is a case of PCI being a newer design/implementation than ZorroIII (late 80's early 90's vs late 90's)
There was "discussion" (actually fanatic ranting) about how the ZorroIII bus was better than PCI, think the conclusion was that ZorroIII could scale up better if a decent chipset was made.
In the end which bus technolgy won out ? Which has the best support ? ... not hard to answer.
Mark
-
The first retail PCI implementation went on sale here in May of '93, so it wasn't that far ahead. 10 years of serious development does wonders though (mostly to 8x AGP!).
-
Zorro3 is older than PCI. Dave Haynie, system architect at Commodore was planning to use PCI in the next Amiga chipset instead of Zorro3/4. PCI wasbuilt from the start as a faster thing, Zorro3 was a jump from 16bit (as in Zorro2) to 32bit access. PCI was defined with a 33MHz bus clock, Zorro3 is asynchronous meaning it has no bus clock. In theory this could allow better improvements, as asynchronous things can talk to each other as fast as the slower of the two devices can possibly go, so you aren't necessarily limited by a predefined clock speed. The problem is that Buster was never updated to go faster than the rev 11 does, so no Zorro3 expansion card can be used faster than Buster can go, as Buster sits between the card and the rest of the system. At the same time, I don't believe any existing Zorro3 cards are capable of going faster than Buster anyway, so there's almost no point in speeding up Buster anymore. The only real benefit we'd see at this point is possibly the PCI bus adaptors could be updated to allow faster PCI bus access from the Amiga system. That'd be cool, but probably not cost effective, and the existing Mediators/Prometheuses may not be able to be updated to significant faster Buster use anyway. I'd thought of tinkering around and making a replacement for Buster in an FPGA chip, but couldn't find a suitable verilog simulator that I could afford for debugging the logic design of what I would have done, and it went nowhere. I'd still buy such a thing if someone made one, but my interests have moved on to other things, plus I've ordered an AmigaOne, so I wont' need to spend the time to build it myself anymore.
The PCI equivalent of "Buster" was generally the "north bridge", but mroe recent things have the north bridge outputting a hypertransport bus that conencts to the "south bridge", and PCI then comes out of the south chip. Older south bridge chips talked across a PCI bus to the north bridge, and south had an ISA bus, and usually IDE, ethernet , serial/parallel, etc. things came from the south bridge. As time moves on, more and more of that stuff is being combined in withthe north bridge stuff to reduce number of chips, and thus make it cheaper to manufacture the motherboard.
-
The Zorro III Buster chip could be improved further with benefits to even present cards. This is not purely a timing issue. However it is way too expensive to even consider a Buster upgrade in terms of chip design, amount of sales against cost of production, etc..
Need for speed? Get an AmigaONE!
-
For a laugh someone could make a Zorro-Mediator for Amiga One :-P
-
Well, coudn't someone just make it as a hobby? There are a lot of projects like that that work in the end.
Kinda like the Coldfire accelerator someone is doing.
To improve the transfer of current PCI to Zorro cards would be cool.
I'd also love to see new modern ZIII cards but I know it's not worth it economically.