I think the people in the CPU section really need to learn a few things:
1. Dont argue with JF AMD about AMD CPU's, design, and naming. He knows infinitely more than you do considering...well none of us know anything :lol: Personally, i wouldnt argue with Jimmy either...just a suggestion :lol: He's extremely knowledgable when it comes to CPU's.
2. Will people please stop bringing up the whole fanboy thing? This is the real world boys and girls, not high school. Theres no need to call someone a fanboy or whatnot if they like one company or another. They can be afilliated with whatever company they want. As long as they arent blatantly lying or spewing nonsense, just ignore and move on.
3. If you want to argue, argue with facts, not personal threats/insults/slander/...
4. The math here could use some work :lol:
Anyway, to contribute somehow to this thread...the 3 most important letters that determine the success of Zambezi are
IPC. I think in my quick read through of this thread, someone mentioned that current AMD 6 cores arent all that good. Those 3 letters are why. CPU's are a lot like cars. So you have 6 great Ford pickup trucks. And theres a track with 2 lanes thats 1 mile long. And lets say 600 pounds of goods you want to get from one side to the other. Each truck can go 100 MPH. But, each truck only has a cab, nothing in the back, no cargo bay. So each truck puts 50 pounds of...lets go with potatoes...in the cab, and 3 of the trucks race down the track, they get to the end in around 35 seconds, offload 100 pounds of potatoes, and take 35 seconds to get back. Two more trucks race down the track, 35 seconds down, offload 100 pounds, 35 seconds back, and then 2 more, 2 more, 2 more, all the way to 600 pounds of potatoes. For this purpose, those trucks took 7 minutes to carry 600 pounds of potatoes 1 mile going 100 MPH. This is the AMD 6 core approach, not necessarily by choice.
Now, onto Intel. Same scenario. 2 lane 1 mile racetrack, 600 pounds of potatoes. But this time, Intel only has 4 pick up trucks. Good news is, each truck has a cargo bay that can hold 100 pounds, and the cab that can hold 50 pounds, for a total of 150 pounds per truck. 2 of the 40 trucks race down the track going at 85 MPH, just because they have a cargo bay. The 2 trucks make it down there in 42 seconds, offload 300 pounds of potatoes, race back taking 42 seconds, and then the next 2 go, offloading 300 pounds of tomatoes. It took them a little less than 3 minutes to move the 600 pounds of potatoes. This is kinda like nehalem. Lower clock speed or MPH than the Thubans.
Last example. Now we have our SB trucks. They can go 110 MPH, and carry 110 pounds in their cargo bays, 50 in their cabs. Ill save you the time and just tell you it takes them almost exactly 2 minutes to do all 600 pounds.
If anybody was smart enough to see the parallelism, good for you
😀 If not, MPH number represents clock speed. SB at 3.4 GHz, Thuban at 3.2 GHz, Nehalem at 2.66/2.8 GHz. The 2 lanes represent a dual threaded app. Most apps actually are dual threaded, some single threaded. The amount the cab and cargo bay could carry is the all important IPC. I guess its IPL in this case, instructions per lap. :lol: But this is where AMD struggled with Phenom and Thuban. They had a lot of cores, at a high speed, but each core couldnt process nearly as much as their Intel counterparts. Of course it isnt 3x advantage for intel, like 50 vs 150 pounds suggests, but i do believe SB put the advantage around 50% over K10 (Phenom and Thuban). Now, give the 6 AMD trucks 6 lanes, and its obviously a different story, which is why the Thuban CPU's kept pace with the Nehalem ones in highly threaded apps. Less so with SB.
Now, the lesson to the story. If AMD cant make a DRAMATIC jump in IPC, BD can NOT be competitive clock for clock with the fastest Intel has to offer. I cant remeber the last time if any an architecture was made that gained 50% IPC over the previous generation. SB was something around 10% over Nehalem. A Q6600 to a SB i7 2600K might even be less than 50%.
So, will BD overcome its IPC deficit? Probably not. The design just doesnt look like its really single core IPC power house orientated. It looks a lot more based on multi-threaded apps and apps highly dependent on integer work (servers fit into both of those). This leads me to believe as far as gaming goes, it wont be as fast as Sandy Bridge in general. Same goes for other single thread and dual threaded apps. If BD can make a large gain in IPC, combined with a generous clock speed and core count, it can still be very very competitive. In the server market it looks like an absolute monster. Its design will make it use less power than the traditional core approach, as well as cost less. Those 2 things are extremely important to servers, along with the facts its server based performance should be fantastic. Im slightly less confident in the real world though. I think it will give stout competition to SB, especially on a performance/price basis, but im fairly skeptical it will manage to dethrone SB as far as flat out performance goes. Especially not the 50% like some rumors say. Atleast not in low threaded apps. Just my opinion though. I certainly hope it blows SB out of the water!
He is being mean to me!" 
Facebook
Twitter
Instagram