[Coco] What about the double speed upgrade?

Joel Ewy jcewy at swbell.net
Sat Jan 6 11:52:30 EST 2007


This thread was originally about the 6309 and Sockmaster's 4MHz
circuit.  Here's a summary as I see it:

1.  The 63B09s Cloud9 sells are compatible with Sockmaster's circuit. 
Though they (and the 68B09) are really only rated for 2 MHz, they will
very likely work.  Kowalski's upgrade does overclock the CPU.  Even a
63C09, which is rated for 3MHz, would probably be run somewhat out of
spec with this circuit.  But that's the nature of overclocking. 
Sockmaster's upgrade however, does not run the CPU at double speed most
of the time.  It just inserts extra, faster clock pulses during periods
when the CPU is not doing an external access.  So the extra heat
generated by overclocking would have more time to dissipate.  This is a
smart, and, I would suspect, relatively safe form of overclocking.  I
think that if I were to build this circuit I might add a heat sink to
the 6309.  But otherwise, I wouldn't be afraid to try it out.

2.  The 6809 benefits more from the 4MHz upgrade than the 6309 (running
in native mode) since the latter already has fewer non-memory clock
cycles, and it is only those that get the speed-up.  But Kowalski gave
us some benchmark numbers that show the 6309 still benefiting quite
significantly.  The combination of the clock-doubler and the 6309 gives
the biggest performance boost.  So the 6309 + 4MHz upgrade is the best
of both worlds.  The amount of improvement you see will really depend on
the mix of instructions in the program you're running.  Generally,
number crunching will benefit more, and I/O less.  But John K. mentions
that even the 6309's memory block transfer instruction gets a 50% boost.

3.  Since the 4MHz circuit only speeds up internal processing and not
accesses to the GIME, RAM, or I/O devices, the other areas of the
computer effectively run at the 2MHz rate.  So it should not prove
incompatible with the hardware, unless there is a device that responds
so slowly that it needs those extra clock cycles between accesses.  This
isn't a likely problem.  Software timing loops would be affected, but
they would also be affected with a 6309 running in native mode, or by
anything else that speeds up the CPU.  To compare with the Amiga (as
mentioned below), there was a DIY accelerator for that machine called
the Little Ugly Cheap Accelerator System (LUCAS) that ran a 68020 w/
optional 68881 FPU (and had an optional RAM expansion board on a local
bus).  It ran at 16MHz, while the rest of the system ran at 7MHz. 
Sockmaster's upgrade is, in effect, the same thing, except that it
doesn't have a separate, asynchronous clock.  It's a clever, much
simpler solution that also provides a faster CPU while leaving the rest
of the system alone.

JCE

jdaggett at gate.net wrote:
> PCs have gone beyond just clock multiplying alone. AMD/Intel with their more 
> modern processors now have two busses. One to access ram and slower 
> peripheral and the other bus to communicate with the rest of the chipset,  ie 
> graphic controller.
>
> james
>
> On 6 Jan 2007 at 0:17, Diego Barizo wrote:
>
>   
>> I might be wrong, but... Isn't that the same that PCs do?
>> A 400 MHz CPU and the bus at 100 MHz...
>> If so, does it have the same implications for any computer? 2 times
>> the CPU speed, same bus, just a 30% real speed increase?
>>
>> Diego
>>
>>
>> Richard E. Crislip wrote:
>>     
>>> On Thursday 04 January 2007 17:58, jdaggett at gate.net wrote:
>>>   
>>>       
>>>> SPeed compatibility issues will exist with either processor when
>>>> doing anything outside the processor. All the existing peripheral
>>>> devices would have to be of equal speed grade with the E clock.
>>>>     
>>>>         
>>> <snip>
>>>
>>> That was the same problem the Amiga guys ran into when they tried to
>>> accelerate the Amiga beyond 7mhz. The supportring chips would not
>>> tolerate being sped up, so everything had to go around them and only
>>> bring them in at the correct moment.
>>>
>>>   
>>>       
>> -- 
>> Coco mailing list
>> Coco at maltedmedia.com
>> http://five.pairlist.net/mailman/listinfo/coco
>>
>>
>> -- 
>> No virus found in this incoming message.
>> Checked by AVG Free Edition.
>> Version: 7.5.432 / Virus Database: 268.16.6/617 - Release Date:
>> 1/5/2007 11:11 AM
>>
>>     
>
>
>
>   




More information about the Coco mailing list