MSI Katana GF66 core clock speed decreases when in a stress test by prime95 from 4.5GHz to 1.5GHz

[kopty]

Hello all,
I was thinking of buying a cooler since my laptop was getting heated up when playing games these days. I'm also planning to change the thermal paste, but before doing so, I was doing a stress test using prime95 "Small FFTs (tests L1/L2/L3 caches, maximum power/heat/CPU stress)" configuration on 3 scenarios -
1. flat on desk
2. using stand
3. using cooler
Both cooler and stand use a suitable elevation.
I noticed that the Core Clock speeds dropped from about 4.5GHz to 1.5GHz in the stress test, which is absurd. Thus, I also have recorded without doing a stress test. Each record is a 5-minute stress (+nostress) test record.

Below are some plots for the 4 scenarios. What do you infer from these?
The title of the plots mentions which stat is plotted, and the individual titles of the 4 subplots mention the scenarios. Max and min (of samples discarding some initial data points) are shown in red and green dashed lines with values.
Stats plot attached -
1. Average Effective Clock [MHz]
2. Core Clocks (avg) [MHz]
3. Core Temperatures (avg) [°C]
4. CPU Package Power [W]
5. Core Usage (avg) [%]
6. Core Thermal Throttling (avg) [Yes/No]

If any more plots or CSV log files are needed, I can attach them.

My analysis:
Looking at the "Core Temperatures (avg) [°C]", it seems the stand works best, whereas the "Core Clocks (avg) [MHz]" or the "Average Effective Clock [MHz]" says the flat is the best??
Thermal throttling occurred at some of the cores all the time (3 out of 8 cores), which should mean better cooling = better clock speed, but that is not the case??
Also, the "Core Thermal Throttling (avg) [Yes/No]" says that even without a stress test, there is sometimes throttling occurring, which means I have to change the thermal paste and clean the laptop (the laptop is 1-2yr old). I'll follow up with the same tests after that is done.

But the question persists, why has the clock speed decreased from 4.5GHz to 1.5GHz, as can be seen in "Core Clocks (avg) [MHz]"!!

 

[kopty]

Hi, so I have changed the thermal paste today. I have used Cooler Master Cryofuze 12.6 W/m-K .

Here are the improved results in the same format. (the extra 't' in the title (ex. coolert_df) means after changing thermal paste)
Looks like there is an improvement.

The difference now is that, in the nostress case, the "Core Clocks (avg) [MHz]" fluctuates from 0.8GHz to 4.5GHz, whereas before, it was staying at ~4.5GHz. In the stressed condition, rather than staying at 1.7GHz, it stays at 2.5GHz, which is definitely an improvement, but the question still lies as to why it can't go to 4.5GHz. Is it thermal throttling?

 

[kopty]

Here are the plots that compare before and after thermal paste change for the same quantities.
This is only shown for cooler usage, so we can easily see improvement.