Thanks or the update, mr. Lee.

If I can afford a naïve question: why, if Sandy Bridge is so efficient, the CPU temperatures are so worse than the previous ones (15-25°C more, with reference to the Athlon II X4 platform)?

Great article. However, I still find the tables confusing, perhaps a graph like CA_Steve posted for your fan roundup would be clearer.

I actually find the tables on the last page very simple and easy to understand.

+1

Thanks for the updated test platform - VRM monitoring! Woot!

Thank you Lawrence.
It would be interesting if you could add the difference between CPU and VRM temperatures with the stock heatsinks for comparison. Intel makes a big deal of the need for airflow on the VRM area but we have no reference for what that means in terms of the reported VRM temperature value.


It *is* more efficient, by a large margin.

We don't know the actual temperatures, only what the CPU reports.
Even the actual CPU temperature is not very interesting. The board is the part which is most at risk of overheating with these heatsinks, which is why the reported VRM temperatures are so useful. This gives us real data on which heatsinks are the most dangerous at low fan speeds.

Any difference between actual CPU temperatures is easily explained. Assuming the same heatsink and the same power consumption, the CPU temperature will be mostly determined by the heat transfer between the parts of the CPU which heat up and the contact with the heatsink. If that heat transfer is relatively inefficient, the CPU temperature will be higher.
It stands to reason that a smaller manufacturing process will cause the heat transfer to be less efficient.
Also, a manufacturer of efficient and heat-resilient CPUs would have less incentive to optimize the heat transfer because it's less of an issue with its products than with the products of its competitors. Overclockers might be frustrated by the inefficient heat transfer however...

Because temperature and power consumption are not the same thing.

If you say so (that it's so "simple")...

For the Noctua NH-L12, when only one fan underneath was mounted, was the fan blowing up or down? If the fan airflow direction was pointed up towards the fins, I'm wondering if it would be possible to test it once the other way to see the difference in temperatures (120 mm reference at full speed would be fine)?

In addition to other responses, I should point out that the test platform is both overclocked and overvolted.

Well, this reason looks like more sound than the other ones, as at stock settings the two platforms should have very similar thermals:



Thanks, I always forget that not-K SKUs can be overclocked (even when I read an article!).
However, if this is the reason, their oc'ing "attitudes" seem somewhat "inefficient" (and Ivy Bridge looks like even worse).

Well, when you overclock you normally try to undervolt as far as possible to reduce the heat generated.

SPCR did exactly the opposite for purposes of simply generating as much heat as possible so they could actually see the difference between the bigger heatsinks.

Conclusion: This test bed is about heatsinks and fans. Ignore the CPU