Linpack Xtreme Released

[Mussels]

VRM's could be overheating too

 

[hat]

VRM's could be overheating too

That's a very real possibility given the power draw of these things. You'd need a really robust cooling setup.

 

[Omegacain]

Something is wrong with your OC configuration or overheating.

It’s not temperature.... i know it has to do with the oc ... it passes the stress test with no errors but only at half the speed I checked hwinfo and all cores are at 100% and 5ghz ... so what in the bios would cause it to hold back .. I got power max out intel c state turn off turbo boost off with the 5ghz oc it’s 273 gflops with no oc just stock it’s like 364gflops

 

[Mussels]

VRMs could easily explain that

Go blast a fan onto them, and see if it helps. If so, you know the issue

 

[Regeneration]

It could be some kind of power limit or throttling. Are you using the same problem size (RAM usage)?

 

[EarthDog]

It’s not temperature.... i know it has to do with the oc ... it passes the stress test with no errors but only at half the speed I checked hwinfo and all cores are at 100% and 5ghz ... so what in the bios would cause it to hold back .. I got power max out intel c state turn off turbo boost off with the 5ghz oc it’s 273 gflops with no oc just stock it’s like 364gflops

its likely throttling. Have intel xtu up (just open) then run this program. At the bottom of xtu it has throttling reasons listed that will turn yellow when it is active. I'm betting it is vrm throttling ornpower limit of some sort.

My suggestion is to use something more in the realistic realm of loads... even p95. There's almost no reason to run something this stressful.

What motherboard are you using? Fill out your system specs bud...

 

[Omegacain]

It’s not my vrm either that is nice and cool I’ll post pictures later ... I think it might have something to do with my motherboard... I just order another one .. z390 master is that a good board

 

[Omegacain]

Here is a picture

 

[EarthDog]

Jesus man... really? Throw money at it?


Do what I said first and see what it is! You'll likely have to change the same thing on that board as you would here, lol!

 

[Omegacain]

I got power max out all settings ... I’m not a fan of the msi Meg ace their is something on the go with the board plus I got another 9900k on the way so

 

[EarthDog]

Again, have Intel XTU up (nut not running a stress test or anything) and run Linkpack. Look at XTU and at the bottom in yellow will be your throttling reason. You may have missed something.

Does this happen nearly instantly or after some minutes? 10 minutes? 20?

This application is overkill and there is little point of using it when the reality is the loads are unrealistic. Maybe see if the same happen with Prime 95 (small FFT or blend) as well (have XTU up also here to compare).


I don't know why you want to chuck more money at it when you don't even know what the problem is... more money than sense isn't a good thing.

EDIT: Are your radiator fans blowing IN the case? You may want to have them blow OUT of the case so you are not blowing CPU heated air across them.....front/sides = intake, top/rear - exhaust. Yes, the CPU will likely be a couple of C warmer, but if that isn't the issue here.... makes sense to setup the case for airFLOW.

Why is that bottom fan blowing out???? WTH? What are the fans in front doing???

 

[Omegacain]

There is no front fans it’s the lian li 011

Side and top. Intake and bottom is Exhaust

 

[EarthDog]

There is no front fans it’s the lian li 011

Side and top. Intake and bottom is Exhaust

That case doesn't have great airflow. That will be a problem no matter what board you have in there.

I am not sure what you said about how your fans are oriented, but your top fans are blowing in the case and the bottom fan is blowing out. You would want to switch that around.....



I'd still love to see a shot of this test with XTU running so you are able to see what is causing the throttling....................

 

[Omegacain]

I’m going to try to put the aio on the side for intake but the pump head wouldn’t reach the cpu with the fans showing so I’m going to switch it to fans first then the rad and see if that would reach that way

I’ll download intel Utility tonight and post a picture with it

 

[EarthDog]

If you actually think about it, that configuration is doing the same thing...blowing heated air into a case that has literally no exhaust.

If you put that rad on the side and the fans in pull, then you would want fans up top exhausting and the bottom flipped to intake.



Still.....the cart is before the horse man. You don't even know what the problem is and you are wanting to swap out parts and configs? For Pete's sake, find out what the issue is before you waste more of your time (and ours trying to help).

 

[Omegacain]

Lian li 011 Dynamic has good airflow it’s like top five on gamers nexus list of airflow... the 011 air is the one that is bad

 

[EarthDog]

Whatevs....... I don't see much intake on that thing (zero fans moving cool air in - 3 moving heated air in)......that really isn't the point though... this is getting dizzying.......

 

[Omegacain]

I got five fans pulling cool air in and 2 Exhaust

Two on the side intake three on top intake two on bottom Exhaust

 

[EarthDog]

I see 3 fans up top blowing warm air in the case. That isn't cool air intake (but is intake).

You have fans at the BOTTOM exhausting air... that doesn't make sense. Heat rises and all so you are fighting that natural thing...

If I was you, I would have that rad up top blowing out, your two fans on the side we cant see blowing in, and the bottom fans blowing up as intake. This will give you a good balance of airflow in that case and keep cooler air moving across the motherboard.


All this over a stress test that is 10x overkill, LOL...

 

[Omegacain]

Ok I will change around the config later tonight

It could be some kind of power limit or throttling. Are you using the same problem size (RAM usage)?

I’m using setting 5 for ram i think it’s the max

 

[Regeneration]

Anyone using v1.0.0 with a modern Intel CPU should update to v1.1.0.

v1.0.0 throws false positive hardware faliures.

 

[Obihoernchen]

Hey @Regeneration,

nice tool!
I have a questions regarding the "default" settings:

Why do you use logical (HT) threads by "default" (I think a lot of people will choose use all threads because they think it's better )?
Linpack should be used without HT. Running Linpack with HT threads just lowers the performance, power consumption and system strain.
Intel also mentions it in their developers guide: https://software.intel.com/en-us/mk...ions-of-the-intel-optimized-linpack-benchmark
For instance my 9900K has ~20% less performance ~10 less power consumption with 16 instead of 8 threads.

Furthermore, I get ~410 GFlops using the official Intel MKL binary with the same Size, LDA, Align, etc.
Using LinpackXtreme extended benchmark I only get ~180 GFlops.
Using LinpackXtreme stability test with 10GB with 16 threads I get ~320 GFlops.
Using LinpackXtreme stability test with 10GB and 8 threads I get ~410 GFlops, too.

The power consumption is also way lower.

Which KMP_AFFINITY settings do you use?

 

[Regeneration]

Hey @Regeneration,

nice tool!
I have a questions regarding the "default" settings:

Why do you use logical (HT) threads by "default" (I think a lot of people will choose use all threads because they think it's better )?
Linpack should be used without HT. Running Linpack with HT threads just lowers the performance, power consumption and system strain.
Intel also mentions it in their developers guide: https://software.intel.com/en-us/mk...ions-of-the-intel-optimized-linpack-benchmark
For instance my 9900K has ~20% less performance ~10 less power consumption with 16 instead of 8 threads.

Furthermore, I get ~410 GFlops using the official Intel MKL binary with the same Size, LDA, Align, etc.
Using LinpackXtreme extended benchmark I only get ~180 GFlops.
Using LinpackXtreme stability test with 10GB with 16 threads I get ~320 GFlops.
Using LinpackXtreme stability test with 10GB and 8 threads I get ~410 GFlops, too.

The power consumption is also way lower.

Which KMP_AFFINITY settings do you use?

The benchmark runs only on true cores, but it uses a special binary to provide accurate comparable results vs. other CPUs, brands and OSes.

Most CPUs tend to crash quicker with HT, and therefore, it is enabled by default for stress testing. Especially, on Linux, Linpack with HT is considered brutal and tend to crash unstable OCs in no time.

I don't use KMP_AFFINITY at all, it doesn't work well on legacy CPUs and AMD. The OMP parameters work better.

 

[EarthDog]

Sounds like it is throttling your system....

 

[Omegacain]

Same with me ... it ain’t throttling either