Ok, so here we are again, another speed bump, another article and everything you read in here is going to be so predictable. Well, think again. True, there will be quite a bit of material that is essentially the benchmarks that everybody wants to see and, since there has not been anything in terms of changes under the hood of the Phenom II X4 965 BE, there will be no big surprises, nor upsets. Unless one counts the victory of a Phenom II X4 965 BE over a Core i7 965 an upset, after all, they use the same moniker, so the performance should be similar … dream on fan boys!

So what’s it going to be, the surprise effect of this article, the stuff nobody has yet read about? If I could put it into a single sentence, I wouldn’t have to write an entire review to get to the point but sometimes, it is necessary to go through the motions and finally ask yourself what the hell it was that you have been doing anyway and finally start trying to make some sense out of the results. In other words, sorry, we won’t spare you the agony of at least looking at the table of contents and trying to figure out where the juicy stuff is – unless you read it on the forums first. That could be some other forums as well, though.

Back to the story line: AMD has just released its fastest CPU ever, until the next one comes along. At 3.4 GHz, there is no holding back in core speed, the Black Edition moniker also indicates that the NB/IMC will be capable of running at least 2.4 GHz, meaning that the new Phenom II X4 965 BE is the fastest officially released quad core processor ever, besting even Intel’s Core i7 975 by a whopping 66 MHz.

Featuring the two independent memory controllers, configurable to run in ganged or unganged mode, the latest AMD flagship screams performance all over the place but, it does come with a drawback: the thermal design power is 140 W, up 15 W from the earlier Phenom II X4 955 sitting at 125W. What exactly does that mean? According to the verbatim meaning of the letters, it means that in order to be qualified to run this CPU, the motherboard manufacturers have to ascertain that their voltage regulator modules have to be able to deliver sustained 140 W of power to the CPU and the cooler has to be able to dissipate that same exact thermal load. That’s all.

What 140W TDP does not mean is that the CPU will draw 140W. In fact, no matter what we threw at it, we have not been able to even get close to that number, neither before nor after the VRM. Let’s jump ahead of ourselves a bit here. The operational word here is p-states, that is processor states in which the CPU is capable of operating. Sometimes, the CPU load is high, sometimes it is low and we know from Intel’s EIST (Enhanced Intel Speedstep Technology) that a CPU can ramp up its core speed and voltage over several small steps for a smooth transition from full speed to idle throttle. Having one full and one idle speed is a good thing, but it is certainly better to have some additional steps in between to match the computational requirements of the work load. All of this has been around with AMD’s CPUs to a level where each clock was regulated separately for each core, leading to the infamous CnQ performance hit. So where is this going? I guess, we have finally stopped overlooking the obvious and analyzed how some applications translate into p-states and actual power consumption.