Some of the information below will not be applicable to older AMD systems that used "thermal margin" temperature monitoring methods rather than actual "temperature" readings. Therefore, this should only be considered applicable primarily to 1st through 10th Gen Intel platforms and AMD Ryzen or TR platforms.
Also, it is fully understood that "there are more ways than one through the woods", so in this case we should clarify that this is not the ONLY way out there, but is the MOST recommended way based on what is generally accepted as a "standard practice" for most systems and with the further clarification that "steady state" is an all important consideration to anybody trying to accurately assess whether or not their system can conform to the appropriate thermal specification when under an all core full load.
The first thing you should do is to download HWinfo and install it. When running HWinfo it is important to select "Sensors only" and uncheck the "Summary" option. You may also use "Core Temp" as a recommended thermal monitoring utility, if you prefer.
Click here to download HWinfo.
Click here to download Core Temp.
Next, click here to download Prime95.
It's important to note that when you run Prime95 there are a few specific settings that must be configured in order for the test to meet the desired "steady state" criteria, which is very important for thermal testing. While stability testing, which is an entirely different thing than thermal testing, usually requires involves a fluctuating workload, thermal testing is best performed using a steady state workload that will not allow the CPU package or individual cores to "rest", which would skew the results and offer false reassurances.
The first screen you'll see is the Welcome screen. Click on "Just stress testing".
To configure the necessary and desired data set choose the "Small FFT" test option on the pop up options window you are greeted with when you open Prime95 OR when you select "Torture test" from the drop down Options menu. Do not choose "Smallest FFT", "Large FFT" or "Blend". Use ONLY "Small FFT".
Also, it is important to disable ALL AVX instruction options. Depending on what model of CPU you have you may see options for AVX, AVX2 and AVX512. Disabling one option by placing a checkmark in the box next to it may result in another option becoming available. Be sure to disable ALL available
AVX instruction options before proceeding.
Open HWinfo and choose the "Sensors only" option. Uncheck the "Summary" option.
(Or Core Temp, if you prefer to use Core Temp)
Open Prime95, configure as outlined above and click "OK" to begin the test.
Keep an eye on HWinfo or Core Temp periodically to ensure that all listed cores or threads are running at 100% for the entire time the test is running. ANY core or thread shown to not be running at 100% means an error has occurred and that "worker" has dropped out of testing. At that point, you should
end the test by clicking on the "Test" column on the Prime95 menu bar and selecting Stop or Exit. Do not click the "X" in the top right corner of the program to close the program like you would with most applications. That will only reduce it to the task bar and the test will continue to run in the background.
You MUST use the "Test>Stop or Test>Exit" options to end the Small FFT test.
Run the test for 15-20 minutes. If no cores exceed 80°C during that time and no workers error out, then the current configuration is acceptably thermally compliant.
If you exceed 80°C but are below 85°C, you are technically probably still ok since most applications and environments are never going to see this kind of real world workload, but be aware that you are pushing the boundaries of what is acceptable and what isn't.
For most users though, anything over 80°C on 1st through 10th Gen Intel or any Ryzen CPU up to the current Zen 3 (Ryzen 5000 series) models, should be considered excessive and testing should be stopped at that point
and corrections to your overclock configuration or changes to your cooling system should be addressed.
It will be obvious to more advanced users that there may technically be additional factors involved, such as adjustments for "normalizing" based on degrees above or below ambient, but for the average user if you simply ignore
any such considerations and stick with a simple "below 80°C is good and above 80°C is, not so good", you will seriously reduce the chances of causing any harm or degrading your CPU prematurely.
Since configurations using AVX instructions in it's various forms is still somewhat of a niche use case, and most users won't be concerned with those configurations, we will not cover that here.
If you need AVX compliant thermal testing procedures you can find a good beginning primer on that here:
Testing with Prime95 AVX enabled