Comment Re:Try to build a comparable Dell for $3000 (Score 1) 875
Would I dare ask what sort of processing you've seen this happen on?
P4s may easily outperform G4s in situations where the P4 doesn't suffer from penalties due to guessing the wrong instructions that will be coming up next.
Every time the P4 guesses wrong, it has to flush the pipieline, which makees it simply stall for 20 pipeline stages! This happens ALL the time in day to day processing.
Also, the P4 is not able to do different kinds of processing in between, like integer, floating and vector processing. The P4 has to flush the pipieline when it needs to process different kinds of instructions. More stalling!
In everyday processing the P4s AND Atholn XPs do a LOT of this pipeline flushing, resulting in a lot of dead stalls. Have you ever felt that Windows isn't operating as smooth as you had hoped? :)
Using benchmarks pipieline flushing can be avoided, resulting in very good benchmark results. Benchmark results are not very much usefull in the real life, though...
P4s may easily outperform G4s in situations where the P4 doesn't suffer from penalties due to guessing the wrong instructions that will be coming up next.
Every time the P4 guesses wrong, it has to flush the pipieline, which makees it simply stall for 20 pipeline stages! This happens ALL the time in day to day processing.
Also, the P4 is not able to do different kinds of processing in between, like integer, floating and vector processing. The P4 has to flush the pipieline when it needs to process different kinds of instructions. More stalling!
In everyday processing the P4s AND Atholn XPs do a LOT of this pipeline flushing, resulting in a lot of dead stalls. Have you ever felt that Windows isn't operating as smooth as you had hoped?
Using benchmarks pipieline flushing can be avoided, resulting in very good benchmark results. Benchmark results are not very much usefull in the real life, though...
