Originally posted by: Kevatl
Easy way to expalin it -
If my CPU can add 1+1+1=3 in one step
but it takes yours 2 steps i.e. 1+1=2 then +1=3
your CPU has to run twice a fast as mine to keep up.
case 1 - 1+1+1=3 (one step)
case 2 - 1+1=2 then +1=3 (two steps)
Now lets say you wanted to make both carry out the complete instruction in 1 sec.
Case 2 has to run twice as fast as Case 1 to do this.
AMD would be case 1, it does more per cycle, therefore requires less Ghz to equal or out perform a P4 for example.
DDR v DDR2
DDR2 high bandwidth but horrible latency.
DDR lower bandwidth much better latency.
Think of latency as delay.
Which is better -
A 4 lane highway where cars are released every 4 secs (DDR2)
or
a 3 lane highway where cars are released every 2 secs (DDR)
Both examples crude, but gives you an idea.
That's not my understanding of what happens.
Instruction-level computations are theoretical maximums; AMD's max is 9, Intel's is 6.
AMD chips are more efficient in parallelizing instruction-level calculations than Intel chips, so use a greater portion of their 9 per clock than Intel with their 6 per clock. As an example, let's say AMD averages use of 6 instructions for 66% efficiency; Intel uses maybe 3, 50% efficiency.
AMD chips excel for applications and programs with parallelized instructions. Intel chips excel best with serial, that is, instructions that must be performed one at a time and/or where instructions are dependant upon the previous instruction, such as certain types of data compression and encryption. Best-case scenario, one might expect Intel to have a 33% improvement over AMD in this area; however it's much closer to about 10-15% due to pipeline inefficiencies. This inefficiency similarly lets AMD chips to perform parallelized operations much better than the Intel chips.
EDIT:
I have no idea how some compression mechanisms are employed; however I suspect that while many are serial-heavy, there may be parallel isntruction components within this set of data manipulation that gives the AMD dual-cores a noticeable, in many cases significant improvements over their single-core counter-parts.