Stress tests and adaptive mode voltage

[alexnt]

Which stress test should I run in order to test system stability on adaptive mode. I assume that prime is prohibitive, right?

 

cpu: 4690k

[CarterDavison]

XTU is the hardest stress test but AIDA64 could be another, possibly more realistic one if it's just a gaming oc.

[airdeano]

do not stress test in adaptive mode. the extra voltages associated with adaptive

mode will create higher than expected temperatures, excessive over-voltage

and just not good practices. use of p95 is fine, but using the right variables:

another free utility is intel XTU. ability to change the modes in OS to verify tests.

 

[PrimeSonic]

If you're going to stress test with Intel-XTU, have something like CPU-Z open as well to keep an eye on voltages.

 

In my personal experience, Intel-XTU doesn't correcly report voltages after they go over a given threshold.

 

If you see your voltages spiking absurdly high, stop the test and adjust them as needed.

[MageTank]

do not stress test in adaptive mode. the extra voltages associated with adaptive

mode will create higher than expected temperatures, excessive over-voltage

and just not good practices. use of p95 is fine, but using the right variables:

 

another free utility is intel XTU. ability to change the modes in OS to verify tests.

Aw, you are skipping the hottest tests in that screenshot. 

 

 

XTU is the hardest stress test but AIDA64 could be another, possibly more realistic one if it's just a gaming oc.

XTU is not the hardest stress test. Not even in the top 10. It uses a very light AVX 2 Linpack, so it gets credit for adding AVX2 heat to the equation, but I can think of several more intense stress tests above XTU. XTU is just one of the more safe AVX2 workloads to test, as it is less stressful than Prime95, and custom Linpack.

[CarterDavison]

Aw, you are skipping the hottest tests in that screenshot. 

 

 

XTU is not the hardest stress test. Not even in the top 10. It uses a very light AVX 2 Linpack, so it gets credit for adding AVX2 heat to the equation, but I can think of several more intense stress tests above XTU. XTU is just one of the more safe AVX2 workloads to test, as it is less stressful than Prime95, and custom Linpack.

I know this but he clearly doesn't know what he's doing so i'm putting it in to more basic terms.

 

I also recommended XTU as it's very user friendly and you can even set how long you should stress for.

[MageTank]

I know this but he clearly doesn't know what he's doing so i'm putting it in to more basic terms.

 

I also recommended XTU as it's very user friendly and you can even set how long you should stress for.

I agree. I think for most people, XTU is probably a better test. Prime95 is mostly used to test specific aspects of my overclocks. 1344k or 2688k for testing vcore, 512k-4096k with 75% memory capacity for testing CPU IMC, 8k - 48k for testing my cooling solution, etc. I use it because I can customize FFT lengths and test the very specific aspects of my overclocking, and see if my thermal solution is enough to handle worst case scenario. Most overclockers won't need to do this. If you are not using AVX 2 that often (PSCX2, GROMACS, GRID 2, etc) then you don't really need to subject yourself to that kind of torture. Running a less stressful test for a longer duration would be ideal in that scenario. It will simulate a more realistic workload. XTU with its ability to run it as long as needed, and still be a decent amount of stress makes it perfect for that scenario.

 

It also won't get as hot as Prime95 or Custom Linpack, so the people with absurdly high OC's will still be happy.

[alexnt]

My question was only if it is good to stress in adaptive mode. I have already tested in manual mode...

[airdeano]

My question was only if it is good to stress in adaptive mode. I have already tested in manual mode...

 

i think this clear above:

do not stress test in adaptive mode.

 

[alexnt]

I have set:
Core ratio 46 @ 1.15v
Ram 2000mhz(from 1866) auto voltage and timings

In digi+ power control:
Cpu LLC: level 4
Cpu phase control to optimized

Eist is on and epu is on

Temps: 66-69

I m running aida extreme stability test for about 30' and everything is stable. How can it be stable at so low voltage?

 

OCCT and HWMonitor show VID: 1.15v(as I have set) and Vcore: 1.68v.

Cpuz: 1.149v

[alexnt]

do not stress test in adaptive mode. the extra voltages associated with adaptive

mode will create higher than expected temperatures, excessive over-voltage

and just not good practices. use of p95 is fine, but using the right variables:

 

another free utility is intel XTU. ability to change the modes in OS to verify tests.

Thank you for the reply! I have one more question. Can you give me your opinion on my post above?

[airdeano]

we call it the silicon lottery when CPUs have larger than normal processing

speed to low voltage/temperature showings. not everyone has these results.

 

what mobo are you using? LLC hasn't been a real issue for CPUs unless you

are having erratic CPU voltage drops/inadequacies.

[alexnt]

we call it the silicon lottery when CPUs have larger than normal processing

speed to low voltage/temperature showings. not everyone has these results.

 

what mobo are you using? LLC hasn't been a real issue for CPUs unless you

are having erratic CPU voltage drops/inadequacies.

I m going to run aida for a couple of hours just to be sure and then I will reduce the voltage to 1.14 because I have already tried and I can log in to windows.

 

I have sabertooth z97 mark 2 and the reason I ask about LLC is that if it is set to "auto" I get high temps(63+ C) at VRM. So I googled and found out that it has to do with digi+ power control LLC level. When overclocking it sets LLC at level 8. In bios it says auto but in asus ai suite it shows it at level 8. So I set it at level 4 and that temp was reduced.

 

What about the voltage readings?

[airdeano]

volts are fine, run the LLC as low as necessary.

[alexnt]

volts are fine, run the LLC as low as necessary.

OK thank you!