Understanding TPD and clock

So, I've got Acer Triton 500 - i7 8750H / RTX 2080 MAX-Q. And I'm trying to understand clock logic here. Take a look:

I've stressed CPU with AIDA, TDP is 45W and clock stuck on 3600mhz. Okay.
https://sun9-53.userapi.com/c858324/v858324109/1f8e23/hs4bLP1J2R4.jpg

But then I've used powerMAX stress test and now i've got 3200mhz on 45TDP.
https://sun9-22.userapi.com/c858324/v858324109/1f8e19/pUIrnI7kCSo.jpg

Both tests loaded my CPU for 100%. I don't understand. One test loaded CPU more than another and it doesn't get enough power to support a higher clock?

 

yes, TDP (or rather, the has priority. Some tests, like Prime95, load the CPU more, so it hits the TDP/power limits at lower clocks.

There's a whole lot of instruction sets in modern CPUs, many with their own separate part on the chip. The more of them a program uses, the higher its power usage and temperature, and the lower the TurboBoost clocks.

AVX, for example (used by Prime95 and others), brings Intel CPUs to their knees, temperature and power usage so high they can only manage base max clocks or barely above that.

You can increase the TDP limit I believe (with Intel XTU or Throttlestop)... if your temperatures are fine (90ish under load)

 

Clock rate is an indicator of work done but not the only one.

X86 cpus have many different logic units which cater for a wide range of different capabilities.
Some units are complex and require more power to run (floating point, and the AVX, FMA type stuff in heavy stress tests). At the other extreme a CPU can appear 100% busy but be just sitting there doing nothing for millions of clocks in a row waiting for data to be fetched from ram, burning almost no power.

People test with crazy heavy loads because any problems with stability will show up when the CPU is under the heaviest hottest load it ever encounter by doing these power hungry instructions 100% of the time. If its stable there, you can be pretty confident it's stable during anything lighter you might do during your everyday usage.

In the olden days this used to hurt performance since who cares if your cpu can't handle an hour straight encoding or calculating Pi to a billion do if you never ever do either of those things. These days turboboost fills that gap and allows the CPU to increase its clockspeed if it hasn't reached its power cap. The only speed Intel guarantees the CPU to maintain is base clock. Everything above that is opportunistic.

TDP is also not power limit, its a cooler design spec that usually coincides with the long term power limit (PL1) of the CPU.

In the short term the turbo power limit is higher, usually 25% above PL1 for 28 seconds. But both are set by the manufacturer, and commonly adjusted up or down with preset max performance/silent fans type system behaviour modes. They can also be used to overclock if left user editable.