Why do anandtech never OC the northbridge on AMD CPUs?

[Anarchist420]

I really never understood why AMD hasn't moved to a 32nm process for Phenom IIs. That way, they could have standard NB speeds of 3 GHz at the same voltage that they have the NB on 45nm.

What really sucks is that BD's NB is only going to be 400 MHz (20&#37 faster than PII's NB.

The other thing that sucks about AMD is that their Black Edition CPUs' Vcores vary wildly and it's just a matter of luck (they should have multiple Black Editions if the Vcore varies so much for the same SKU). That's why I'm not going to buy anything else from AMD. I understand that Intel does the same thing, but their vcores aren't as high to begin with, while they have much better performance, so it's not as bad that their vcores vary for the same SKU.

 

[Arkadrel]

I believe "HyperTransport 3.1" was mentioned with bulldozer.

Which according to Wiki is:
A Max. aggregate bandwidth (bi-directional) = 51.2 GB/s

Hypertransport 3.0 (2.6ghz) = 41.6 GB/s
Hypertransport 3.1 (3.2ghz) = 51.2 GB/s (23% or so faster)

http://en.wikipedia.org/wiki/HyperTransport

 

[formulav8]

Yeah buy PhII does't even hit HTT 3 limits. HTT 3 is 2.6ghz where-as PhII run's @ a 2ghz htt link. So you can't exactly go by that.

 

[Tuna-Fish]

I really never understood why AMD hasn't moved to a 32nm process for Phenom IIs. That way, they could have standard NB speeds of 3 GHz at the same voltage that they have the NB on 45nm.

Because GloFo has only just ramped up 32nm production? With Llano launching less than a month from now, in a sense they are moving to 32nm process for Phenom II. They just aren't Intel, so it took a while.

 

[Meaker10]

HTT clock != NB clock.

There is no real benefit raising the HTT clock.

Overclocking the NB increases memory bandwidth and reduces L3 cache latencies.