Re: That 8W is at the CPU, not at the VRM input. You have to multiply by 1.3-1.5 depending on how good the VRM efficiency is.
That's completely incorrect. VRM efficiency is quite good, and your methodology for calculating it below is wrong. Instead of guestimating Intel's power efficiency, how about using a review that tests the new G-0 step CPUs? There have been about a half dozen other sites which have, and they all find them to be lower power.
Re: On page 30, look at the power for the 65nm 65W TDP 2.6GHz X2 5000+. It has the same frequency area as the E6700 but at lower power in the minimum C1E power state of 3.04A
The C1E state is a new state for the AMD 65nm G-step chips. The C1 state already operates at Min P-state of 1.1V, at which voltage the processor has a leakage value of 4.4A. AMD gets to a lower leakage value in C1E state by disconnecting the HTT link, putting the memory in self-refresh mode, and tri-stating the DDR SDRAM pins (except CKE pins). See note 13 for all of these points.
Intel's C1E state achieves lower leakage through lowering the voltage. However, their datasheet does not state exactly how low the voltage goes in C1E state (the datasheet merely says in note #3 on table 5 that the VID differs during power management events). I suspect Intel reduced their C1E state from B-step to G-step primarily through further voltage reduction techniques.
Re: Since it measured 8.8W at the VRM input
Hold on, Pete! Your review did not use a Brisbane 5000+ (2.6GHz). It had a Brisbane 2.5GHz measuring 8.8W. It's unclear if their samples are implementing all the power manage techniques, but since we know that motherboard vendors often leave some power management on the table for AMD parts, it's not hard to imagine that this is the reason for the slight power increase over spec.
Re: that puts the VRM efficiency at 43%
Bull. VRMs are very efficient, never that low at any point. There is another reason why they are measuring high for the Brisbane core, but it has nothing to do with VRM efficiency, since the Intel core measures nearly spot-on to its 22W C1E spec (see below about your wrong assumption for 12W C1E power).
Re: The E6700 still has a 12W TDP in C1E same as the E4300
Actually, the E6700 is 22W or 12W, depending on the sku. You obviously wish it to be the 12W sku in order to prove your point, but it's clear that they are using the same measurement from their first review of the Core 2 Duo last July.
lostcircuits.com
Since Intel only had 22W skus available at that time, it could only be that version of the product. They have actually not tested a newer Intel sku in about a year. |
