jm7e+@ANDREW.CMU.EDU ("Jeremy G. Mereness") (07/31/88)
It has been talked about (in Open-Apple for one) that the gs could be sped up tremendously by some simple(?) hardware restructuring. Thinking about this, I reasoned that one of the biggest bottlenecks would be the slow RAM dedicated to the display, as even in 16-bit mode, the processor must slow down to access it. Well, wait a minute. I understand that in 8-bit mode, it would be necessary for compatibility to slow down the clock for display access, but why should this be in 16-bit mode? I guess the real question is how much of a difference a RAM chip's speed makes to 65xxx processors (I have read on this board something implying that on the 68000 series, RAM speed doesn't matter so much)? If the slow RAM could be filled with fast RAM chips, what would happen (besides the price going up... damned boycott; where are the American companies to take advantage of this)? If RAM could be accessed at any speed so long as it is not beyond its capability (yes/no?), then why not fill the slow RAM with fast chips and arrange for the 65816 to slow down to 1 MHz only when in 8-bit mode? Would this work? Would it trash every 16-bit program that's been written so far? I wanna know.... Capt. Albatross jm7e+@andrew.cmu.edu ============ Looking forward to a GS+... disclaimer: These opinions are mine and will remain so until more intelligent or insightful or informed people are kind enough to show me the error of my ways because in the barbecue of life, a mind is a terrible thing to baste.
whitney@think.COM (David Whitney) (08/01/88)
In article <UWwZ6xy00Uo140PUUb@andrew.cmu.edu> jm7e+@ANDREW.CMU.EDU ("Jeremy G. Mereness") writes: > >It has been talked about (in Open-Apple for one) that the gs could be sped up >tremendously by some simple(?) hardware restructuring. > >Thinking about this, I reasoned that one of the biggest bottlenecks would be >the slow RAM dedicated to the display, as even in 16-bit mode, the processor >must slow down to access it. > ... >If the slow RAM could be filled with fast RAM chips, what would happen (besides >the price going up... damned boycott; where are the American companies to take >advantage of this)? > >If RAM could be accessed at any speed so long as it is not beyond its >capability (yes/no?), then why not fill the slow RAM with fast chips and >arrange for the 65816 to slow down to 1 MHz only when in 8-bit mode? Would this >work? Would it trash every 16-bit program that's been written so far? I think the reason for slowing down when accessing video RAM was to keep timing with the video circuitry simple. That's what I heard was the real trouble. It makes for a pretty good bottleneck, as there tends to be a lot of video activity in any useful program. Note also that the processor slows when the 5.25" disk is on. Try this: call-151, C0E9<ret> (drive turns on - make sure it doesn't have disk inside!), 300L - note the printout speed - C0E8<ret> (drive turns off), 300L - zing! much faster. It may also slow down for the 3.5" but I don't think so (have absolutely no idea, really). As for the 2.8 MHz speed in general, it turns out that 150ns RAM limits the processor speed to 5.6 MHz. Since I haven't yet taken the appropriate class, I can't tell you the formula used to compute this, but a friend who TA'd the class tells me 5.6 MHz is the top. Now, 150ns RAM is relatively cheap, so the clock speed restriction basically kept the price of the machine down. Also, since the max speed can be 5.6MHz, that's why the two accelerators promised sometime soon (AE's, which was demoed at the Boston Applefest, and one from MDIdeas that I've read about in The Apple //GS Buyer's Guide) promise a speed of 5.6 MHz. If they were to go faster, they would need faster onboard RAM which would cost a whole lot. If a GS+ (or some such) *does* go at nice dreamy speeds of 7.x MHz or more, then it will need faster RAM which will cost a whole lot more unless the machine uses wait states (like the Mac family) which effectively slows down the computer. Oh well... David Whitney, MIT '90 DISCLAIMER: Nobody knows what I'm up {out there}!harvard!think!whitney to. Don't blame them for my actions whitney@think.com nor me for theirs. ^^^^^ will be changing before 1989 is here. Don't depend on it after 1/1/89.
sshankar@umn-cs.cs.umn.edu (Subash Shankar) (08/02/88)
In article <24698@think.UUCP> whitney@godot.think.com.UUCP (David Whitney) writes: > >As for the 2.8 MHz speed in general, it turns out that 150ns RAM >limits the processor speed to 5.6 MHz. Since I haven't yet taken the Thanks. I was hoping that somebody would mention the memory bottleneck instead of blaming Apple for not producing a 25 MHz GS. In actuality, I believe that the maximum speed would be less then 5.6 MHz due to bus delays. (note 150 ns = 5.6 MHz). I've heard estimates of around 4 Mhz. > >If a GS+ (or some such) *does* go at nice dreamy speeds of 7.x MHz or >more, then it will need faster RAM which will cost a whole lot more >unless the machine uses wait states (like the Mac family) which >effectively slows down the computer. Oh well... As if wait states on MSDOS machines weren't bad enough, they would be absolutely meaningless on the 6502 family. The 6502, unlike other processors accesses memory on every cycle either for instructions or data. The addition of wait states for higher clock rates thus has absolutely no effect on performance. In contrast the 68XXX and 80XXX have several instructions that can proceed without memory access, and these instruction (mainly arithmetic) benefit from the higher clock rates. Also, their four cycle memory access is slowed down less by a wait state (20%) when compared to the single-cycle memory access 6502 which slows down by 50%.