End of Moore's law?

[me-5yx05kfkO/aqeI1yJSURBw at public.gmane.org (Matthew Gillen)]

On 7/10/2010 8:32 PM, Mark Komarinski wrote:
> On 7/10/2010 3:19 PM, Mark Woodward wrote:
>> Any thoughts?
>>    
> 
> Start off by reading what Moore's Law actually is:
> 
> http://en.wikipedia.org/wiki/Moore%27s_law^
> 
> Moore's law say nothing about performance.  It merely says that the 
> number of transistors that can be inexpensively put on an IC doubles 
> every 2 years.  Let's see what improvements have appeared in the last 
> few years:
> 
> - went from a single core per CPU (hyperthreading doesn't count) to 8, 
> with 12 around the corner (and hypertheading is back in the mix now, so 
> 16-24 possible cores, depending on if you're using threads)
> - moved memory controller and PCI-E controller into the CPU and removing 
> the northbridge
> - went to a 32nm die from 90nm in 2002
> 
> There's likely others, but I don't follow CPU design that closely.  But 
> to say that Moore's law is dead ignores a lot of what's been going on, 
> especially on the server side.

I understand what you're saying, but the practical implication of the
more 'conventional wisdom' interpretation of Moore's Law (that single
processor speeds double) was that MBA types could mindlessly upgrade
every couple years and get better performance.

I think it's a major (and long-term) boon to computer-science related
professions that the world is going to multi-core as a way to boost
performance, instead of increasing single-processor speeds.  Getting a
24-core monster is no longer a trivial decision to make when you've got
to boost performance.  You've got to actually plan how to utilize
something like that.  Which is good for technologists.

Matt