You are talking about overclocking existing parts. I was talking about
Intel/AMD learning from the overclockers and manufacturing parts that
consistently run at the speeds that overclockers can already achieve.
If you have some special insight on what it would take to
change their manufacturing process to achieve this, I'm sure
they'd love to hear from you.
Remember that any good engineer, engineers in as much margin
as possible when the environment of use is unknown. A
processor that might overclock well, might instead be put
into a poor generic case, wear a poor heatsink, be ran in a
fairly warm environment. AMD/Intel know, as do the
overclockers, that keeping the temperature below a certain
level is important. If they spec'd a CPU for the highest it
could run in *ideal* conditions, it would be a terrible
mistake as the vast majority of CPUs are not ran in ideal
conditions. Remember that some people don't even open up
their systems to clean the dust out, ever, rather taking it
to a shop in a few years saying "it's broken/crashes, must
be a virus".
My old
Athlon 2400+ would happily run undervolted at its stock speed.
I too have undervolted processors, but remember that it may
not run like this for the "potential" life of the system, 10
years give or take. As the motherboard capacitors and PSU
age they will have more ripple and the low voltage threshold
for stable operation will tend to rise. Even so, someone
familiar with aging and able to periodically recheck
stability might have a good effect undervolting, but this is
a bit opposite of the previously desired goal of higher
speed.
Also remember that AMD and Intel have been in the biz for
quite a few years, I'm sure they have some consideration of
what the public perception is when they can't supply parts
to meet demand when their process won't allow enough of part
X and speed Y to fill orders. They don't want a public
perception of instability either by allowing too little
margin.
Do you recall the Pentium 3 @ 1.13GHz? In stock
configuration it wasn't stable and that became a bit
notorious, looked bad for Intel. If one cooled them better
and raised voltage, they could be ran stabily, some even
past 2GHz but the industry in general was not ready for this
amount of attention put into cooling. Educating the system
integrators in smaller shops as well as end users is a slow
process, some are not so keen on the finer details of
implementation, rather slapping the system together and
installing windows as fast as possible.
Mobile
processors run at significantly lower voltages, so why not start there -
Because they cost more and the market has already shown it
wants low cost more than max performance per power. What
sells in high volume, has an additional edge when it comes
to driving down prices by concentrating on that market and
devoting manufacturing to that core... and then later rating
them to fill the orders per part, not necessarily based on
max speed any one specimen could support. They have to meet
X # of orders for part Y and differentiate the price between
part Y and Z, even when the two parts might be physcially
identical except for being set to a different multiplier for
those willing to pay more.
take low voltage processor technology and develop it so that with a small
increase in voltage, a large increase in GHz could be achieved and still
stay within the heat spec of the higher voltaged part, thereby still
producing a chip that can be cooled with the stock HSF.
This is already the case, have you bought these mobile parts
and done this? Your Athlon 2400 for example, in a mobile
Barton form did tend to allow it.
That's marketing for you!
See my point following 1.
Yes, but it still doesn't resolve system aging of
capacitors, dust buildup, or less than ideal environment.
Be glad there is so much margin instead of trying to wring
every last bit out. It does tend to make computers more
reliable in general and if a system isn't reliable, stable,
it becomes little more than a toy.
Yes, faster speeds *contribute* to higher heat, but as you know there are
other ways around the heat problem.
You're trying to get elaborate, which always tends to cause
more problems once the end users become part of the
equation. If Intel only made sealed systems with the
processors installed, it could be more viable but they are
not interested in doing this. What is possible through
attention to detail and what is practical when we can't
assume there will be this level of attention to detail are
always opposing factors. Likewise you can wring the last
foot pound of torque out of a car but ran more
conservatively the public gets better reliability long-term.
The general public doesn't need faster processors as much as
they need their present system to keep running reliably.
Just the other day a friend of the family brought me her old
K6/2-500 based system which was (several years ago) built by
me for her out of parts I was ready to throw away at the
time. Even now years later all she wanted was for it to
work still (after she had basically created multiple
hardware configurations in windows, and seeing a prompt come
up at each boot she fiddled around trying to fix it until
she had managed to set the boot partition as inactive).
Overclockers are a small percentage of the market for CPU. They are actually
spoiling it for everyone else by selling chips that aren't set to run as
fast as they 'can'.
They're not spoiling anything. I've given several reasons
why the chips aren't engineered to so thin a margin they
won't work reliably a large % of the time or remain stable
for life of the system.
Intel and AMD don't take extra steps to "change" anything
for overclockers, when it comes to design decisions for
practical purposes overclockers are ignored as if they don't
exist. It is the basic requirements of being in THIS
business, this market, making the parts they do, that
dictate the decisions made.
My point was just that I would like a really fast single threaded/cored CPU
that will compile my work very quickly and play games very quickly. I don't
want to have to invest time and money in cooling so I can overclock and mess
about and I would prefer not to pay for a second core that I really don't
need! However, I have a Core 2 Duo as it was the fastest processor (on
paper) that I could afford at the time of my upgrade.
Why are you in such a *rush*? Gamers manage to use existing
processors and typically find the video card the primary
bottleneck. Businesses have ran fine for years with far
slower processors than we have today. Enjoy the tech
instead of finding fault, it is meant to make things easier
not wrought with hopes for something more.
I have no idea which architecture performs best at your (was
it Visual Studio?) app but seeking benchmarks would be the
place to start. Since you demonstrate the basic
understanding of overclocking, it is up to you whether to do
so. Intel and AMD can't guarantee what level of competence
you will have doing so, therefore they can't spec a part to
be guaranteed to run as fast as you might be able to run it.
Margin is a good thing.
