Kuzi
Senior member
We all know now that the route Intel took in glueing two CPUs together was the better choice. Had AMD done that, we would have probably seen quad core K10's (non native) maybe a year ago. I think when it comes to Barcelona/Phenom, AMD put a lot of time and effort on multi-core performance and also to have a "native" quad core.
That's why generally we see K10's efficiency in highly multithreaded apps higher than C2D, but single core performance is lower, and that leads to lower overal performance. I can't help but think that K10 was designed as more of a server chip than a desktop one, as it performs very nicely in the server space even at only say 2.2ghz.
Overall K10's improvement is pretty good over K8, but from the reviews I've seen, AMD still has some way to go when it comes to caches (latency/size/density compared to C2D). So if AMD can release 45nm Shanghai in 2008, that has tweaked caches with lower latency, 6MB L3 cache, and faster North Bridge, their performance can easily be 5-10% higher than 65nm K10. Which would basically allow it to perform on par with "current" C2D, and a bit slower than the new penryns. With the right price Shanghai can be a winner.
Yes AMD is very very late, and I'm not holding much hope for the B3 stepping K10s that should be released in a few months, cause even with higher frequencies, they still can't compete with Intel's 45nm Yorkfields on price/performance/power efficiency ratio.
Lets hope for AMD's sake they won't mess up their 45nm launch, no more delays or bugs, then they can be pretty competitive. Of course Intel will release Nehalem at the end of 2008 with integrated IMC and that will be a completely different story.
That's why generally we see K10's efficiency in highly multithreaded apps higher than C2D, but single core performance is lower, and that leads to lower overal performance. I can't help but think that K10 was designed as more of a server chip than a desktop one, as it performs very nicely in the server space even at only say 2.2ghz.
Overall K10's improvement is pretty good over K8, but from the reviews I've seen, AMD still has some way to go when it comes to caches (latency/size/density compared to C2D). So if AMD can release 45nm Shanghai in 2008, that has tweaked caches with lower latency, 6MB L3 cache, and faster North Bridge, their performance can easily be 5-10% higher than 65nm K10. Which would basically allow it to perform on par with "current" C2D, and a bit slower than the new penryns. With the right price Shanghai can be a winner.
Yes AMD is very very late, and I'm not holding much hope for the B3 stepping K10s that should be released in a few months, cause even with higher frequencies, they still can't compete with Intel's 45nm Yorkfields on price/performance/power efficiency ratio.
Lets hope for AMD's sake they won't mess up their 45nm launch, no more delays or bugs, then they can be pretty competitive. Of course Intel will release Nehalem at the end of 2008 with integrated IMC and that will be a completely different story.