AVR-Chat

Thomas Keller wrote:
> On Thu, 2005-10-06 at 11:00 +0800, Kathy Quinlan wrote:
> 
>>Don wrote:
>>
>>>Eduardo Munoz wrote:
>>>
>>>>I tried it with At76c713DK and JtagIce MkII.  
>>
>>A 48Mhz AVR now that is nice (I wonder how high our clock will go, 
>>without liquid nitrogen cooling ;)
> 
>    Actually, heat isn't the limting factor.  While higher clock
> frequencies will,  of course, increase heat generation, the real
> problems are elsewhere.  Issues such as leakage currents which rise
> dramatically as device geometries scale down for high dspeed, etc.
> 
>    If a Pentium or an Athlon can run at 3 Ghz and more without cryogenic
> cooling,m an AVR can, and probably with considerably less mechanical
> (e.g., fans) cooling, as well.

The speed is limited by the architecture. The IO (pins, busses, etc)
need to be separated using a lower speed controller because they can't
run at the GHz core speed. Things like on-chip core caches are needed
for this kind of architecture.

 > The real question is, what would one
> reaonably *DO* with a 3 Ghz AVR?  The primnary reason that embedded
> controllers run at such relatively slow clock speeds is economics.
> Embedded controllers must be inexpensive.  Mutli-Ghz clocked chips are
> NOT inexpensive, at any level of complexity.  Otherwise, why would wqe
> all be salviating over 48 Mhz AVRs, when Intel and AMD and others
> produce multi Ghz processors?  (hint: complexity and expense)

Most micros that have a cache have a core that runs faster than the
IO pins and busses (ARM, etc).

>    Have you looked at what it takes to get a Pentium or an Athlon to do
> anything useful, in terms of external support chips and circuitry (not
> to mention the mechanical and cooling support required)?  Now compare
> that to an AVR,m which can run completely stanbalone, withbout any
> external chips at all.
> 
>    And there I go in lecture mode again.  Sowwy.
> 
> tom