I cant understand your statement "thermal design power never should've been
allowed to get as high as it is regardless of how much performance could be
gained".
I mean, even if they could triple performance, they
shouldn't have released CPUs with that high a TDP to achieve
it. The matter should've been one of an upper limit to TDP,
not one of "what performance gain would there be".
Surely if the thermal design power was as high as it currently is
but (as an exaggeration to make things clearer) a machine equipped with such
a CPU could achieve benchmarks of twice what is currently available - i.e.
twice the performance, then wouldn't the fact that this could perhaps be
used to replace a server pool of 2 machines with only one, justify the high
power consumption?
It would be unusual and arbitrary to assume of the 2 servers
that the only significant bottleneck were the CPU. One
would think then that a single dual CPU server should be
used. Additionally there are server CPU that could have
different design goals, this is a desktop CPU.
It's performance per watt which really matters and not
just total power consumption.
Again I disagree, in the real world the total power
consumption does matter.
I do understand your reference to things such
as idle loops.
Quite simple, many people have their CPUs just sitting there
producing 70W plus, supposedly "idling" at the desktop.
Causes include viri, Yahoo message bars and other kinds of
little tray apps for scanners or who-knows-what. Bottom
line is, 70W continually if not higher, when if the TDP were
lower, so would this pseudo-idle power usage. We could
argue about what these users would do "in a perfect world"
but in this one, it happens and happens VERY often, might
almost be considered typical. Remember that Joe Average does
not get as "hands on" with their system as many do here,
that Joe Average is the majority of users.
Further, some uses do not have the CPU running OS with ACPI
halt cooling working or even that feature.
Just as CPUs can be inefficient, software can be designed
without efficiency placed as a high priority. I once had a piece of comms
software which took 25 seconds to initialise my modem ready for answering
machine use, I complained about it to the software publisher because I also
had a competing product which could do exactly the same thing on the same
hardware in about 3 seconds. The product which took 25 seconds is no longer
in production whereas the other (faster but uglier and less user friendly)
product is.
True, when it comes to performance there are other factors.
That diminishes the perceived need for utmost performance
regardless of TDP in many situations, doesn't it? Do we
keep buying faster/hotter CPUs even for tasks that shouldn't
need them, merely to combat poorly written software?
If soneone had a specific job(s) that need higher CPU
performance such that job length changes, we have a
different situation where higher TDP of CPU might actually
reduce overall power consumption by allowing system to run
shorter period of time, or fewer systems. This seems to be
the minority of uses for theses PC targeted CPUs or the
majority of systems.
