> -----Original Message-----
> From: Frank_Gruendel_at_t-online.de [mailto:Frank_Gruendel_at_t-online.de]
> Sent: Wednesday, April 24, 2002 05:19p
> To: newtontalk_at_newtontalk.net
> Subject: [NTLK] Re - Overclocking
> Dave,
>
> I never thought I'd ever disagree with you. But please allow
> me to do so in this one case with all due respect.
Please Frank, be my guest. :)
> Although I'm 100% with you regarding the processor in the
> Newton, the previous discussion was of a more general
> nature. Apart from that, there are other highly integrated
> chips (ASIC's) in the Newton of which I doubt anybody
> on this list has a datasheet. I've been waiting for years
> for somebody to prove me wrong ;-))
I have everything except the Voyager chipset.
> During the year before developing
> hardware finally started to loose the challenge I need, I have
> developed ASIC's. At that time, fairly highly integrated ones
> with about 250.000 gates. And I was told that some of those
> were killed by accidently using the wrong clock generator.
Entirely possible. But how much was the difference? Was it 1MHz, 10MHz
100MHz etc. faster. If the part was designed for say, 50MHz, clocking with
60MHz would do no harm. Now it may not work but it would be ok. Try clocking
with orders of magnitude above that and all bets are off.
> As you correctly pointed out, processors from the same die
> are tested and labeled with the highest frequency at which
> they work reliably. This means of course that many processors
> were originally designed for a higher frequency and will
> work within their initial specification even if overclocked
> (up to a certain point).
In fact it is in the fabricators best interest to make just one piece of
silicon and have one assembly line.
> But (please be gentle and correct me if I'm wrong) this does
> also mean that there are *some* components that *do* work
> at the frequency they were initially designed for. If the
> production line is working as it should, I'd expect *most*
> of the components to fall into the latter category.
Not sure I understand what you are trying to say here, Frank.
But, a manufacturer makes a chip designed to the highest specifications
and the characterizes into set parameter bands. The bands are chosen to be
statistically distinct but there are always going to be overlap of those bands
in some devices. If you are lucky, you will be able to run your chip in the next
band.
> Although assuming
> that these tests allow for some safety margin even for those
> components, I'd expect this margin to be comparatively tight.
> If a 1 GHz processor would work reliably at 1.2 GHz, it would
> probably be sold as a 1.2 GHz processor.
Yes, but the difference between 1GHz and 1.2GHz is 200MHz! Hardly a small
increment!
> In my humble opinion one can't
> make a general statement when it comes to
> overclocking.
Sure can! Like anything else, it depends on what you are generalizing and
how generalized you want to be. (Does that make sense??)
> The outcome depends on many factors. Chip technology,
> gate count, physical gate size, availability
> or absence of automatic heat protection circuitry within the
> chip, redundancy in the external cooling system etc. etc. etc.
Yes, all these and more are factors in the equation but what tends to
happen in these discussions is that parallels with other unrelated systems
tend to cloud an already fuzzy issue. (Well, not for me).
> Clocking a processor specified with 1 Ghz
> with twice that frequency might be tempting fate. Clocking it
> with 100 MHz more might make no difference whatsoever.
Exactly. It is absolutely right that users question the potential
for problems that overclocking may present, but it is less than helpful
to the non-technical users to present compelling and well-meaning
psudoscience that they will take as gospel.
So, my 'in-a-nutshell' statement is this:
For the Newton, in my opinion, the act of overclocking the system will
not unduly stress any components within. The power usage will be measurably
higher but it is in no way comparable to the power hit seen in a larger machine.
The downside of overclocking is that the Newton may just lock-up at random
times YMMV.
Just for illustration purposes (they are not real parts)
Imagine you have 3 speed grades 100MHz, 200MHz and 300MHz
The manufacturer only makes 300MHz parts and rigorously tests each one.
The chips can do 3 functions A, B and C.
A chip must do all functions at a given frequency to pass.
A chip is tested at 300MHz and fails A and B.
It is tested at 200MHz and fails A.
It is tested at 100MHz and passes all three. It is stamped '100MHz'.
If I have this fitted in my unit it obviously works at 100Hz.
Now I overclock to 200MHz and because I never use function A, it works.
If I never use function B, I could overclock to 300MHz.
The above is a bit contrived but it serves to illustrate the point that
the dangers of overclocking (in the Newton arena) are largely limited
to more frequent pushes of the reset button than anything else.
This is my opinion based on some credible facts.
Of course, opinions are like a**holes;
Everyone's got one and everyone thinks the other persons' stinks!
Regards,
David Humphreys
-- Read the List FAQ/Etiquette: http://www.newtontalk.net/faq.html Read the Newton FAQ: http://www.guns-media.com/mirrors/newton/faq/ This is the NewtonTalk mailing list - http://www.newtontalk.net
This archive was generated by hypermail 2.1.2 : Sun May 05 2002 - 14:05:01 EDT