Another aspect that we absolutely can’t leave out is overclocking. AMD claims that the launch of the new processor coincided with certain progress on the way towards more advanced manufacturing process. Therefore, we can expect overclocking to be pretty successful. We decided to see if it was really so.
Overclocking experiments were performed in the same testbed as was used for performance tests. I would only like to add that the CPU was cooled with Scythe Mugen (Infinity) cooler with Noctua NF-P12.
Since the tested CPU belongs to the Black Edition series, we decided to use simple overclocking techniques: by increasing the multiplier. At the same time, I would like to remind you that the alternative overclocking method based on clock generator frequency increase is just as efficient, which we have seen many times during our previous tests.
Frankly speaking, the results of our overclocking experiments turned out a little disappointing. When we increased the processor Vcore by 0.175 V above the nominal, i.e. to 1.585 V, Phenom II X4 965 remained stable only at 3.8 GHz frequency.
On the other hand, we can’t really expect any serious overclocking improvements. Even specially selected Phenom II X4 TWKR 42 Black Edition overclocking processors can overclock with air cooling only to 4.0 GHz. So, even if we can claim a certain improvement of the overclocking potential, this improvement is extremely insignificant.
Unfortunately, we have to state that the overclocking attractiveness of the top Phenom II X4 processors starts to disappear little by little. At this point AMD has used up almost the entire frequency potential of their 45 nm Deneb cores. With air cooling alone we could only overclock the new Phenom II X4 965 by 10-15%, which is, in fact, another indication that we shouldn’t expect any faster quad-core processors based on Deneb core any time soon.
However, at the same time we have some very good news for overclocking fans. The thermal sensors integrated directly into the CPU cores of the new Phenom II X4 965 processors have finally been calibrated correctly. It means that during work in nominal mode and during overclocking you can finally use not only the readings off the thermal diode beneath the CPU socket, but also the readings off the CPU itself, which are more precise and have considerably lower lag effect.
The screenshot below, for instance, shows the temperature of the Phenom II X4 965 processor working at 3.8 GHz during the LinX run that we use to test the system stability.
I would like to remind you that processor sensors used to report absolutely unreal temperature that was about 20 °C lower than the actual one, which made it impossible to trust their readings. Unfortunately, it took AMD over 6 months to fix this issue, but we hope that we will now see correctly calibrated thermal diodes not only in the top Phenom II X4 processors, but also in other 45 nm CPUs.