Overclocking
Overclocking
We now move to a part that was not initially planned. After the change of thermal paste and the gain of temperatures on the CPU, we are interested in solutions to gain in performance. Changing the thermal paste did not allow us to gain in performance because of the power limit that we were already reaching before. The solution is simple, we had to change the power limit values.
There are two ways to approach this, the first is through the BIOS, the second is via the Intel XTU software. We chose the first one and took the opportunity to set the ventilation to 100% without the cover.
So the PL1 power limit was increased from 28W to 64W and PL2 from 64W to 80W with a PL2 hold time of 228 seconds instead of 28 seconds. We then tested this on Cinebench R20, geekbench 3, 3DMark Time spy and 3DMark Firestrike.
Cinebench R20
On Cinebench R20 we observe a gain of 13% with a frequency maintained between 4100 MHz and 4300 MHz against 3200 to 3900 MHz before. The power consumption stagnates at 70 W against 64 W then 40 W before.
Geekbench 3
On geekbench 3 the gain is less with only 2.5%, here we still can’t reach the i7-12700h of the Dell Inspiron 14 Plus.
3DMark Time spy
On 3DMark Time spy the gain is negligible, here the overall score is more influenced by the GPU than the CPU. The gain on the CPU is only 3.8% and only overclocked RAM or more cores could grant a significant gain.
3DMark Firestrike
On 3DMark Firestrike same observation as on Time spy, the gain on the CPU is only 1,9 %.
This upward manipulation of the power limit therefore provides a performance gain only in certain cases. It is therefore not very useful for a machine dedicated to office work. However, it should be pointed out that the fact of being able to touch the PL1 and PL2 can be applied in the other direction. A reduction of the power limit would be judicious to better contain the heating and the noise of this NUC in certain cases of use.
Let’s find out right away with some performance tests on Cinebench R20 at different power limit levels.
As we can see, by opting for a power limit of 40 W we manage to lower the temperature by 16 °C while remaining ahead of the MSI Prestige 13 Evo A13M.
For my part, I will choose a PL1 of 28 W and a PL 2 of 40 W. This would have a definite impact on multi-thread performance but not for office use. With this setting the average temperature drops to 67°C on Cinebench R20 for 4068 points. But this has very little impact on the office suite for example with 14 105 points against 14 117 points before on the PCMark 10 Application benchmark.
