[Coco] What about the double speed upgrade?
John Kowalski
sock at axess.com
Thu Jan 4 23:41:00 EST 2007
At 06:04 PM 04/01/2007 -0600, L. Curtis Boyle wrote:
>On Thu, 04 Jan 2007 17:29:36 -0600, Joel Ewy <jcewy at swbell.net> wrote:
>
>> It's been a while since I looked at Sockmaster's 4MHz upgrade, but IIRC,
>> it increases the clock speed for the CPU only. E remains at 2MHz for
>> the rest of the system.
>>
>> However I also seem to remember that much of the speed increase would be
>> lost on the 6309. I don't recall the exact details right now, but
...
> Sock's (and Bob Puppo's before him) clock doublers only double the
>speed when the CPU is not accessing RAM. So, you are right, in 6309 native
>mode, where the chip caches the instruction byte, you don't see as much of
>a gain (except on many cycle CPU instructions that don't touch RAM, like
>DIVx and MULD instructions). On things like TFM, it won't speed up at all.
Actually, TFM runs 50% faster with the clock doubler because TFM executes a
non-memory cycle for each byte that is transfered. Without clock doubler it
takes 3 cycles per transferred byte. With clock doubler, it takes 2 cycles
per byte.
Some operations see more benefit than others - MUL and DIV run nearly twice
as fast, while some cases like ADDA #immediate run at the same speed.
The averaged clocked increases work out as:
Clock doubled 6809 runs 34% faster.
Clock doubled 6309 in native mode runs 21% faster.
For a side-by-side comparison, here's an index table I had put together -
all CoCo configurations were running the exact same complex program when
measuring speed:
1 = stock CoCo 1 running at 0.89 Mhz
1.14 = CoCo 1 with 6309 in native mode (but still running 6809 code)
2.00 = stock CoCo 3 double speed poke (1.79 Mhz)
2.28 = CoCo 3 with 6309 in native mode (1.79 Mhz)
2.68 = clock doubled ""pseudo 3.58 Mhz"" 6809
2.75 = clock doubled 6309 in native mode (running 6809 code)
John Kowalski (Sock Master)
http://www.axess.com/twilight/sock/
More information about the Coco
mailing list