IIGS Acceleration Idea
Greg Andrzejewski
I understand why you can't simply run a IIe machine at, say, 16 MHz.
Everything about the machine is so timing-dependent that if you try to
change the speed things get real ugly real fast. So the solution for most
IIe accelators was to use caching. All the computation could be done fast in
cache, but the chip would have to slow down when talking to the apple II
bus. Please correct me if I am mistaken.
Fast forward to the Apple IIgs. The same principle of getting a speed
increase through caching could be scaled up to work with the new hardware.
In the interest of getting a product to market in a timely fashion most of
the same concepts appeared in IIgs accelator cards.
When reading threads concerning cloning the TWGS, I thought after alomst
two decades of computer progress there has to be a better way.
A stock GS has a "slow" mode at about 1 Mhz and a "fast" mode at 2.8
MHz. What is preventing us from simply making this "fast" mode something
like 10 MHz? I don't know enough about the IIgs hardware to actually suggest
that this is possible, but I figure that you guys around here could give an
answer far better than any book or website.
Thoughts, ideas, ridicule?
-Greg
aiia...@gmail.com
a computer has to know when the bus is available to be used by the CPU.
the apple II computer uses the data bus to display video, because the
display is memory mapped.
the apple II computer uses the bus in this way:
1)load/save data to RAM
2)fetch next video byte
1)load/save data to RAM
2)fetch next video byte.
If you put in a faster oscillator, the CPU runs faster, so does the
video.
The original oscillator runs the motherboard at the correct speed to
make the video circuits output NTSC frequency.
NTSC frequency is required by your monitor (it takes that signal as
input and displays it on the glass)
if you speed up the motherboard, you get a signal from the video
circuits which isn't NTSC....
how about this: replace oscillator on motherboard, replace slow chips
with faster ones, and modify your monitor to work with the faster
speeds ?
:-P
PZ
vicinity of the 2.6mhz speed. Now what speed the logic can handle is
an unknown, probably not 10mhz, but possibly somewhere in the range of
3.5mhz - 5.2mhz. Ram would have to be upgraded for sure, but 50ns DRAM
are obtainable. A 5.2mhz (double 2.8mhz) cachless accelerator mod for
the masses wouln't be a bad idea. This is assuming we can trick the
VGC into stepping down the appropriate amount to synch with the stock
1.023mhz for "slow I/O", video, etc.
- Paul
Linards Ticmanis
And I think the 3.5" drive handling depends on the 2.8 MHz speed. If
this is so, you'd need at least three speeds to handle it all.
--
Linards Ticmanis
Greg Andrzejewski
> And I think the 3.5" drive handling depends on the 2.8 MHz speed. If
> this is so, you'd need at least three speeds to handle it all.
>
> --
> Linards Ticmanis
Then does the 3.5" floppy not work in slow mode? Or does the machine know to
speed up whenever the floppy is being accessed? It must be the latter, as
I'm pretty sure I've booted into GS/OS, forgetting to switch back to fast
after running some IIe stuff.
For the record, I think Appletalk depends on 2.8 MHz operation, too. But I'd
be willing to lose Appletalk if it meant a faster machine without a
costly/rare accelerator card.
-Greg
Ed Eastman
> A stock GS has a "slow" mode at about 1 Mhz and a "fast" mode at 2.8
> MHz. What is preventing us from simply making this "fast" mode something
> like 10 MHz?
In a word, the FPI/CYA chip. This is the thing that times and actually
does everything, from the ram refresh to the slot timing to the
mainboard and slot RAM addressing. To speed up the IIgs you'd have to
implement a new CYA chip to produce all the timing and address signals
for the main board plus the faster clock for the IIgs.
How to accelerate your IIgs without recreating the wheel:
1. Overdrive: Replace the CPU with a high speed CPU and simple 20x clock
circuit. Monitor the Speed softswitch, and the HIGH pulse of the CPU
clock along with the 'internal cycle' signals off the CPU. Turn those
signals into drive for overclock circuit. Each internal cycle will be
then 1/10 the normal time and be ready for the next activity at normal
speed. Shoudl be 100% compatible and break nothing.
2. Freewheeling 65816: A little more complex, 100% compatible. Add fast
CPU, clock, Ram and refresh circuit. Additional circuit to monitor the
low clock pulse to 'freeze' the CPU to obey the FPI/CYA. Obviously
circuit to route and clock Bank 0/1 (E0/E1) and Rom and Softswitches to
main board, run the rest in the CPU RAM.
3. Reinvent the wheel, er IIgs mode: Build the same thing but instead of
freezing the CPU on low clock from main board, ignore the main board
completely except for: video ram updates and slot access. Monitor speed
softswitch, Rom access, Softswitch access, drive activity, downshift/
sync to main board clock as necessary. aybe load ROMs to RAM and
eliminate rom slowdown.
Want a block diagram? Want a wiring diagram? Can your read minds? THen
nevermind.
Thankx,
Ed
Bryan Parkoff
"Ed Eastman" <no...@nowhere.net> wrote in message
news:drl9kf$7rr$1...@unlnews.unl.edu...
Ed,
I have to say -- do 1 GHz (GigaHertz) crystal osc chip exist? If so,
you will have to redesign Apple //e or Apple IIgs motherboard. 65816 CPU
will run at 1 GHz. One GHz crystal osc is divided by two or four to handle
RAM as DIMM (533 MHz or beyond?). Then 1 GHz is divided by 75 (Yes, it is
seventy-five) to handle video at 14MHz to support NTSC TV and monitor.
You have to find some ways to work with sound, floppy drive, etc that
they only handle at 1MHz while 3.5" floppy drive and other devices handle at
2.5MHz. You would need to use DMA to prevent from CPU and video scanner
sharing the same RAM like real Apple //e and Apple IIgs have. If it does
not work, you will need to design a special chip which it allows to delay
video and other devices until CPU is processing million cycles (less than 1
GHz) before video can be resumed to process in turn.
Michael Mahon and other folks said that it is impossible, but you have
to understand how timing generator works. It is what "Understanding the
Apple //e" manual explains everything how timing generator works.
I wish that it would WORK when you reinvent Apple IIgs' design of
motherboard. If it does work, I can say that it is faster than Pentium IV
2-4 GHz.
Bryan Parkoff
Michael J. Mahon
I was going to let this pass, but since you "quote" me I'll respond.
Nothing that doesn't violate the laws of physics is impossible. What
I said was that it is not possible to just speed up the clock and have
anything workable.
I would regard a 1GHz 65816, if it were built, as a curiosity--and a
ridiculous use of transistors.
What are you trying to do? Do you regard is as necessary to "destroy
the Apple II in order to save it"?
Of course an Apple IIgs could be completely redesigned to run at
gigahertz-speed, but that doesn't mean that it would be useful,
or that it should be done, or--and this is the critical point--
that it ever will be done.
There is no existing Apple software that does not run "acceptably"
with a conventional accelerator, and there will never be any software
for the Apple that requires gigahertz speed. This is a fantasy.
(BTW, gigahertz clocks are generated by phase-locking to a lower
speed crystal.)
-michael
Music synthesis for 8-bit Apple II's!
Home page: http://members.aol.com/MJMahon/
"The wastebasket is our most important design
tool--and it is seriously underused."
Eric Smith
> I have to say -- do 1 GHz (GigaHertz) crystal osc chip exist?
Possibly, but it wouldn't have a 3.3V or 5V CMOS logic compatible output.
Chips that internally run over 200 MHz use an internal PLL to multiply
a lower-frequency external clock.
> If so, you will have to redesign Apple //e or Apple IIgs motherboard.
> 65816 CPU will run at 1 GHz.
Not bloody likely. You'd be very lucky to get reliable 50 MHz operation
out of those.
And even if you could run it a 1 GHz, you can't get memory suitable for
use as a cache at that speed.
If you designed your own chip in a 90nm process, you could perhaps make
a 500 MHz 65816 with some on-chip cache. You could expect to spend over
$500K on NRE. If you were able to sell 1000 chips, you'd probably have
to sell them for about $750 each to break even.
> Michael Mahon and other folks said that it is impossible,
And they're correct. You'd not just have to design your own processor,
but your own memory controller, and an interface to the "legacy" Apple II
hardware. This isn't something you can do by minor modifications to
your existing Apple IIgs logic board.
If you disagree with Mr. Mahon and want to prove him wrong, the way to
do it isn't by making up wild claims, but rather by actually doing what
is claimed to be impossible.
> I wish that it would WORK when you reinvent Apple IIgs' design of
> motherboard. If it does work, I can say that it is faster than Pentium IV
> 2-4 GHz.
No, a 1 GHz or even 2 GHz 65816 would NOT be faster than a 2 GHz Pentium 4.
bie...@terra.es
>
> I have to say -- do 1 GHz (GigaHertz) crystal osc chip exist? If so,
> you will have to redesign Apple //e or Apple IIgs motherboard. 65816 CPU
> will run at 1 GHz.
Bryan,
Is this year the a2's 30th anniversary...?
An a2@1GHz==1000x faster, would be a nice way to celebrate it !!
but...
65816@1GHz... where have you seen it ? The last time I checked wdc they
had 20MHz parts.?
Even if you had the most optimized better available 65816 ip core I
think there's no way to sopc either as no current fpga will eat it
@1GHz.?
And, also, you would need a 2GHz memory bus for a 1GHz cpu clock !! (do
memory dimms this fast exist yet ?) or about 10 or so cpu wait cycles
(for a 100MHz memory bus) and then, it's of little (or no) use to run a
65xx cpu at 10X the memory bus speed...
:-(
Bryan Parkoff
<bie...@terra.es> wrote in message
news:1138661356.9...@g47g2000cwa.googlegroups.com...
It is just our imaginary to come true which it would happen. The fact
is that WDC provides 65816 CPU which it limits to 14MHz while it can extend
to maximum 20MHz. If you want to run 1 GHz on 65816 CPU, you will need to
redesign 65816 CPU to support 1 GHz.
I do not need to say to prove if Michael Mahon is wrong, but try to
imagine some ideas which it has never happened. Try to say if Pentium 4 2-4
GHz is faster, but it has worse performance because Intel can't outperform
up to 20 GHz so they decide to redesign CPU to find other ways how to
outperform. It is the same example that 65816 CPU should be redesigned to
outperform up to 100MHz and then try to do 1GHz. Remember, it is only your
imaginary
Bryan Parkoff
Michael J. Mahon
It sounds like you are saying that what the Apple II community is
short on is imagination.
I disagree. We have plenty (maybe too much) collective imagination.
What we are lacking is sufficient execution!
Henry
Henry S. Courbis
www.GSE-Reactive.com
Apple II Series Legacy Hardware - Come take a look at what we have to offer!
Michael J. Mahon
I wrote (>>):
>>It sounds like you are saying that what the Apple II community is
>>short on is imagination.
>>
>>I disagree. We have plenty (maybe too much) collective imagination.
>>What we are lacking is sufficient execution!
>
> I'm working on that. ;)
Way to go!
logjam
TRYING to stop me and focus my energy on an accelerator.
Possibly af 65816 can be made out of TTL logic. NOW we're thinking...
Is it cheating to call something a TTL computer if it has ethernet and
ram?
Michael J. Mahon
> I have too much imagination and way too much execution. Henry is
> TRYING to stop me and focus my energy on an accelerator.
Ha! ;-)
> Possibly af 65816 can be made out of TTL logic. NOW we're thinking...
>
> Is it cheating to call something a TTL computer if it has ethernet and
> ram?
Not at all--the very first ethernet-connected computers were TTL.
And all the TTL machines since the late 1970s used DRAM (unless
core memory was a mission requirement).
Henry
then that and he's just all over the place.
Henry S. Courbis
www.GSE-Reactive.com
Apple II Series Legacy Hardware - Come take a look at what we have to offer!
"logjam" <gr...@cmosxray.com> wrote in message
news:1138788775....@g44g2000cwa.googlegroups.com...