bryce@COGSCI.BERKELEY.EDU (Bryce Nesbitt) (04/24/87)
Rumors have it that the internal slot on the Amiga 500 will be able to accept 512 or so of SLOW FAST memory with no hope of splitting the bus to create FAST FAST memory. What does net-land think about this, and what does Commodore have to say in it's defense? #beginner_mode_on For those who do not know, the Amiga custom chips can be programmed to do operations like line draw and area fill in CHIP memory. Programs can also specify that extra colors are to be available on each video line. Both of these things can use memory slots that would otherwise be available for the processor. This is a ->feature<-. It is also a ->feature<- that (on the Amiga 1000) there is a separate bus that the processor can have all to itself. Memory that attaches here is called FAST memory. #beginner_mode_off The slot on the 500 will have the disadvantage of contention with with the bus used by the custom chips, yet not the advantage of being addressable by them. The real crime is that this memory will ->pretend<- to be FAST, but actually be CHIP, well not really 'cause... The hassles are potentially gigantic. The drop in potential is tragic, particularly for things like CAD that can be calculating in the background while the blitter is off doing it's thing. I feel the descision is unwise. I am aware of the cost related reasons (Hi Jeff!) for the descision. Still, I don't think any of the third party developers want to need to deal with SLOW FAST ram. In decreasing order of expense to produce the base machine I offer the following solutions: 1> Turn this into real FAST memory, buffered and isolated from CHIP memory. 2> Make this CHIP memory and poke a couple of holes in one of the ladies :-) to address it. 3> Since this is an EXPANSION board shuffle the cost of implementing item #1 onto the board. This is probably the best compromise of the three. Another point to consider is that this board will probably be cloned right off, as the internal 256k was. Commodore can make a half-hearted and possibly profitless attempt at the board, and leave it to the cloners to pull up the slack. ---------- One more A-500 comment: A clock chip on the above mentioned board is to be located at $dc00000 absolute with a reader on WB. This is is AGE OF AUTOCONFIG, the DECADE OF 'PLUG AND PLAY'. Cleaner for everybody except the poor sop who gets to design it would be a ROM that copies itself to RAM at config time and gets executed once DOS is up. It can then stay resident (say, for a SCSI driver) or de-allocate itself (our clock). ---------- -Bryce-
grr@cbmvax.cbm.UUCP (George Robbins) (04/26/87)
In article <8704240432.AA23509@cogsci.berkeley.edu> bryce@COGSCI.BERKELEY.EDU (Bryce Nesbitt) writes: > >Rumors have it that the internal slot on the Amiga 500 will be able to accept >512 or so of SLOW FAST memory with no hope of splitting the bus to create >FAST FAST memory. What does net-land think about this, and what does >Commodore have to say in it's defense? Yawn... Oops, that wasn't very political*, was it? The A500 *does* have an (electrically) A1000 compatible expansion connector on the side. Memory plugged in there *can* be "Fast". >The slot on the 500 will have the disadvantage of contention with with the >bus used by the custom chips, yet not the advantage of being addressable >by them. The real crime is that this memory will ->pretend<- to be FAST, >but actually be CHIP, well not really 'cause... Uh, silly question, but how slow is this "SLOW FAST" memory? Answer, precisely the same speed as the memory in an A1000! There are some software tuning issues involved, but but they are not complicated and can be addressed in the next kickstart release. If you're still not convinced, note that having the internal expansion be "pseudo" chip memory makes it a bit easier to turn it into "real" chip memory whenever we come up with some chipset enhancements.. > Another point to consider is that this board will probably be cloned > right off, as the internal 256k was. Commodore can make a half-hearted > and possibly profitless attempt at the board, and leave it to the cloners > to pull up the slack. That depends on the price of the expansion and it's availability / desirability at the point/time of sale. Ask any third party that tried to clone the A1000 256K memory expansion what happened to their sales once the dealers started selling the expansions as part of the initial system purchase. > One more A-500 comment: A clock chip on the above mentioned board is to be > located at $dc00000 absolute with a reader on WB. This is is AGE OF > AUTOCONFIG, the DECADE OF 'PLUG AND PLAY'. Cleaner for everybody except > the poor sop who gets to design it would be a ROM that copies itself to > RAM at config time and gets executed once DOS is up. It can then stay > resident (say, for a SCSI driver) or de-allocate itself (our clock). I don't follow this very well. The only way we could justify a battery backed up real-time clock (no volume product smaller than a PC/AT had one at the time) was to make it optional. You simply put a setclock command in your statup script, if the clock is present, it initializes the system date and time of day, if not it does nothing. Sounds kinda like plug-n-play to me... * Sorry, this slow memory/fast memory stuff expounded by people who don't really understand the trade-off's or who should know better tends to give me a red rash... -- George Robbins - now working for, uucp: {ihnp4|seismo|rutgers}!cbmvax!grr but no way officially representing arpa: cbmvax!grr@seismo.css.GOV Commodore, Engineering Department fone: 215-431-9255 (only by moonlite)
daveh@cbmvax.cbm.UUCP (Dave Haynie) (04/27/87)
in article <8704240432.AA23509@cogsci.berkeley.edu>, bryce@COGSCI.BERKELEY.EDU (Bryce Nesbitt) says: > Keywords: A-500 autoconfig clocks > > > Rumors have it that the internal slot on the Amiga 500 will be able to accept > 512 or so of SLOW FAST memory with no hope of splitting the bus to create > FAST FAST memory. What does net-land think about this, and what does > Commodore have to say in it's defense? The internal slot of the A500 is set up to take a card with 512K of non-chip memory that's subject to the same contentions that the CHIP memory is. That's what you call SLOW FAST memory. That's also why the list price of this add-in card, with a battery backed-up real-time clock, is about $150. You're perfectly welcome to buy an A500 and hand some auto-configuring real FAST memory off the expansion port, just like on an A1000, in which case that memory will run just as fast as it would on an A1000. This internal memory is a NEW FEATURE, no one's constrained to use it. > I feel the descision is unwise. I am aware of the cost related reasons > (Hi Jeff!) for the descision. Still, I don't think any of the third party > developers want to need to deal with SLOW FAST ram. There's no reason they should know or care if its there. Things will run just as fast on the A500 with built-in 1 meg as they will on the A500 with built-in 512K. -- Dave Haynie Commodore-Amiga Usenet: {ihnp4|caip|rutgers}!cbmvax!daveh "The A2000 Guy" BIX : hazy "These are the days of miracle and wonder" -P. Simon