anyone check the sisandra memory benchmark of these phenom babes? it seems that it is slower than current K8. this is one of the main reason why phenom isnt phenomenal. The L3 cache is the culprit here, the latency is just so high that overall it might cancel out the advantage of having integrated memory controller.


"Phenom is much like the solemn relative, visiting during a time of great sorrow within the family" . quote from anandtech.

which translate into...less competitive CPU market for the next 3 years or more. . unless AMD can do something to remedy the slow memory bandwidth. But that was impossible unless a architecture update is being planned.

This post has been edited by sparrowed: Nov 20 2007, 12:37 AM

from what i see, K10 is a risky bet from AMD. They introduce native quad core architecture while having uncertainty of having benefits over non native quad core in real world application, they introduce the slow L3 cache which dramatically slow down the memory bandwidth even slower compare to K8 and hence overall performance was reduced, and K8/K10 benefit alot of having fast memory bandwidth(not the case for C2D).

AMD design of K10 is good, but they took a wrong direction of having those risky bets. IMO, if AMD can somehow fix its slow L3 cache and doubled the memory bandwidth compare to K8, im sure they can at least match intel's Core cpu.

the power saving feature on phenom suppose to turn non-working cores down and even the entire HT lane connecting form the cpu to the chipset, (correct me if im wrong), but somehow this feature wont work. look at the power consumption of phenoms during idle, it is even higher than Athlon X2. - X2 cant turn off one of its core, while phenom suppose to turn down 3 of its cores.