Huge CPU temp differences between different stress tests

[Mentis]

Hi,

 

So I was wondering why the CPU temps are so different between the Prime95 small FFT test (~85c) and Intel Xtreme tuning utility stress test (~66c)? Both are using all the 6700k cores at 100% @4.4GHz 1.33Vcore (NH-D15 cooler)

 

Thank you

[airdeano]

just like it is presented, it develops maximum heat, might try using this as a custom blend test (min/max FFT 1344):

[zMeul]

different tests utilize different instruction sub sets (extensions)

• AVX
• SSE4

Prim95, I know for a fact, uses AVX instructions; while Intel's test might not

[SpaceGhostC2C]

3 minutes ago, Mentis said:

Hi,

 

So I was wondering why the CPU temps are so different between the Prime95 small FFT test (~85c) and Intel Xtreme tuning utility stress test (~66c)? Both are using all the 6700k cores at 100% @4.4GHz 1.33Vcore (NH-D15 cooler)

 

Thank you

The % of CPU usage is only a coarse measure of the utilization of the CPU. Each of those tests uses the CPU in a different ways and is able to saturate particular units within the CPU to a different extent.

[Imakuni]

3 minutes ago, Mentis said:

Hi,

 

So I was wondering why the CPU temps are so different between the Prime95 small FFT test (~85c) and Intel Xtreme tuning utility stress test (~66c)? Both are using all the 6700k cores at 100% @4.4GHz 1.33Vcore (NH-D15 cooler)

 

Thank you

Because they are different types of workloads. P95 Small FTT is extremely well optimized to squeeze as much performance as possible out of your CPU, making use of the latest instruction sets due to the kinds of calculations done. Intel XTU, however, doesn't do that.

 

Think of it as playing a match of CS GO or swimming. Both require 100% of your attention, but one generates a lot of heat on your body and the other, not so much.

[Mentis]

3 minutes ago, zMeul said:

different tests utilize different instruction sub sets (extensions)

• AVX
• SSE4

Prim95, I know for a fact, uses AVX instructions; while Intel's test might not

It would be interesting to know how different extensions have different power draw!!! They're all using the same chip after all

[Mentis]

4 minutes ago, SpaceGhostC2C said:

The % of CPU usage is only a coarse measure of the utilization of the CPU. Each of those tests uses the CPU in a different ways and is able to saturate particular units within the CPU to a different extent.

 

2 minutes ago, Imakuni said:

Because they are different types of workloads. P95 Small FTT is extremely well optimized to squeeze as much performance as possible out of your CPU, making use of the latest instruction sets due to the kinds of calculations done. Intel XTU, however, doesn't do that.

 

Think of it as playing a match of CS GO or swimming. Both require 100% of your attention, but one generates a lot of heat on your body and the other, not so much.

So can one test the stability better than the other? Or is it only good to test maximum temps?

[Imakuni]

5 minutes ago, Mentis said:

So can one test the stability better than the other? Or is it only good to test maximum temps?

Back to the CS GO and swimming analogy: one stresses mainly the arms and legs while the other goes more for fingers and the brain.

 

If your OC is stable, it'll pass both tests. But if it's unstable, it might fail only one or the other. Or it might not even fail either and instead just crash upon opening a game. Personally, I prefer P95, but that doesn't necessarily mean that it's a be all, end all stress test.

 

For heat (aka cooler stress testing), however, P95 straight up wins, no questions asked.

[SpaceGhostC2C]

4 minutes ago, Mentis said:

 

So can one test the stability better than the other? Or is it only good to test maximum temps?

That's beyond my knowledge, but my experience, for the same OC, is:

 

- Intel Burn Test gives me the lowest temperatures, and tend to be more forgiving (passes some OCs that other tests don't)

- Prime95 generates the highest temps, and is not as forgiving, but may need to run for a long time before detecting a problem, making heat the main concern.

- OCCT is somewhere in between in terms of temperature, and is the one that fails the quickest of these 3.

 

Your mileage may vary.