breemen@rulcvx.LeidenUniv.nl (E. van Breemen) (04/09/91)
There has been a lot of talk about the speed of the Custom chips
in the Amiga. It is clear that the Custom chips are currently slowing
down the Amiga. I am developing a accelerator board myself and I know
that such a board has to synchronize to the original 7 Mhz machine, because
the Amiga makes (too much in my opinion) assumptions about the signals in
different clock cycles. The best thing to do is to split up the Amiga in
to parts: 1) the calculation part with a 68030 on say 25/33 Mhz and
2) make a seperate part with all the slow stuff like video
io and sound at 7/14 Mhz.
In this way you can use (almost) twice as much buscycles for the video
part and the 68030 can run at full speed. In the video part, you can put
a 68000 or 68020 or even a customized graphicschip (i.e improved blitter).
The basic communication should be done by graphics.library (as supported
by Commodore). So good software (which doesn't poke in the screen) will
work directly. Software like games, can make use of the 68000 in the
video part for direct manipulation of video ram.
By separating the boards, one board is not slowed down by the other if
we doesn't speak of the software demands. In this way the Amiga can be
given more computing power by making a faster calculation board.
All these idea's are currently implemented by the Accelerator boards but
they are just a kind of a hardware patch. The Amiga isn't ( in my humble
opinion ) designed for upgrading. Just look at all the problems with
noise, dma etc.
I hope people will see this as a positive contribution to the Amiga
Future.
Erwin van Breemen
Orega Holland.daveh@cbmvax.commodore.com (Dave Haynie) (04/19/91)
In article <1991Apr9.091806.21523@rulway.LeidenUniv.nl> breemen@rulcvx.LeidenUniv.nl (E. van Breemen) writes: >The Amiga isn't ( in my humble opinion ) designed for upgrading. Just look at >all the problems with noise, dma etc. All Amigas were designed for expansion. In the proper ways. That means, Coprocessor boards go in coprocessor slots, memory boards go in expansion slots. No computer is designed to support such things wedged into CPU sockets, those you use at your own risk. A500s and A1000s were designed to support a single expansion box, which could be a Zorro II backplane or a stand-alone SOTS box like the A590. What problems? ANY upgradable system assumes proper design on the part of both the backplane/motherboard piece, and the add-in board piece. Commodore can certainly control the first piece, and certainly does. It can't do much of anything about the second part, the add-in board, except on its own add in boards. If you're having trouble getting your own designs to work, perhaps you need to learn some more about designing in the first place. No one said it should be real easy to design any plug in card. After all, if it was something that anyone fresh out of school could do perfectly, they wouldn't have to pay experts to do it here at Commodore. In general, we have very little trouble with either noise or DMA, these are simply things you must learn how to deal with. You can get some ideas on how to deal with noise by carefully examining some of the Commodore plug-in boards. You deal with noise in a design at both the design and the layout stages. Proper designs have plenty of bypass capacitors, logic line termination, and various other factors that keep such noise down. Problem is, a good many people igore this part of the design, and concentrate only on the digital end of things. Layout is even more critical, especially if you are designing a two layer board for the Zorro II bus (I don't believe any Zorro III board should be on a two layer board, period, end of discussion). Most EE types mess up their first PCB layout. DMA is significantly more difficult to do correctly than PIO. The only real problems with DMA that Commodore has ever had are with the A2090 controllers. The main reason these were a problem was their heritage -- the A2090 used a DMA controller designed for a different machines. And no one knew of the problems with bus acquisition you could run into back then on an Amiga. Since then (A2090 was the first Amiga DMA device), there are plenty of DMA devices that work just fine. Like the A2091. The Microbotics Hardframe (I got one of those in my office A2500 at the moment). There is no problem building a reliable DMA device if you know what you're doing. >Erwin van Breemen >Orega Holland. -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "That's me in the corner, that's me in the spotlight" -R.E.M.
hawk@pnet01.cts.com (John Anderson) (04/20/91)
Dave, I am having lots of trouble getting through to you by EMail. If Could you leave me mail telling if you never got the messages or if the headers were garbled or just what the situation might be. Thanks (sorry for posting publicly, Email seems to be fudged, so any net people wanting to flame ... *CHILL*)