Temp problems with XC7

I just recently built a system based around hydro x cooling.

 

Problem is the xc7 has stupid poor cooling performance for me. I use a 9900ks (stock settings so far, cant oc with these temps). The temperatures spike a ton when idling/doing simple tasks. Opening google chrome will make it spike to 60c.

And it's much worse when I try to benchmark. Running prime95 I'll be lucky if I stay at 80c, but in most cases I'll reach 100c instantly and the computer then freezes if I dont stop it.

 

I've tried using different thermal compunds: Arctic Silver 5, Thermal Grizzly Cundoctonaut and Artic MX-4. All with the same result.

 

I've also looked at how the Silver 5 and MX-4 spread across the IHS, and it seemed fine and even.

 

I also use the XG7 on a 2080ti, and the temps there are much better, as it stays under 45c when running timespy.

 

After disassembling the cpu part of the loop 5 times I'm not sure what to do. Does anyone else?

What motherboard do you use?

The asus rog strix z390-e, which is on the compatibily list.

There is a very, very, very important value that's missing here - your vCore (vCPU in iCUE).

 

What are your vCore values? I know that you are saying "stock" but Asus is well-known for auto-overclocking with stock settings.

If by vCore you mean the voltage the CPU is running then it's 1.306V currently when looking in HWMonitor.

That looks to be well within the norms for voltage for that cpu. I take it before this upgrade you didn't have problems with it overheating? It sounds like the heatsink it's making good contact with the cpu. That's the only thing I can think as to why it's acting like this. When you tighten it down, do you gradually go around the waterblock tightening a little at a time? I use the criss cross method, like tightening lug nuts on a car.

I use the criss cross method too so that I can be sure that the thermal compound is spread evenly. And when I've removed the block I've checked if it's spread evenly, and it has been every time. It also covered the whole cpu.

This is why I have a hard time figuring out what's causing the high temps.

CPU temp spikes are either voltage based on or a contact issue. There isn't really another option. So if you are sure the physical contact is good, then you need to dive into the BIOS and see what's going on.

 

1.30v seems a touch heavy for 5.0GHz, but not out of bounds and probably pretty standard for the KS series that is preset to 5GHz out of the box. However, I wonder if it is really staying at that level.

 

If you are using Auto (adaptive) voltage, go into the Adv. BIOS -> Extreme Tweaker column. Scroll down to Internal Power Management. Enter that sub menu and set IA AC/DC Load Line to 0.01. That should calm down voltage shifting. You may also need to set a manual Load Line Calibration Level instead of Auto, but it would be better to only change one thing at a time. Most modern motherboards cannot handle synthetic stress testing right out of the box and need to be tweaked for it. However, the general temp spikes when performing mundane tasks suggests something else is amiss.

I personally think it's a contact issue. With what he is describing it most certainly sounds contact related. Shooting to 60c just opening a browser. Is this a new build, or did you have a different cooling solution prior?

It's a completely new build

And unfortunately I didn't check thermals when I tested the hardware with air coolers prior to installing HydroX.

Since you're saying your thermal imprint is fine I assume this is not the case then.

 

Must be the motherboard and it's auto-overclocking function, ASUS has a function called MCE, which effectively overclocks all cores to the single-core limit. By doing so, it automatically overshoots Vcore by a significant margin as it uses single-core load Vcore using all-core load. Yikes!

 

Just to be 100% sure - can you post a picture of your TIM imprint on the block and CPU?

 

Normally, this block performs perfectly with 9900K/990KS, see TechPowerUp's review: https://www.techpowerup.com/review/ek-momentum-msi-z390-meg-ace-monoblock/6.html

 

I hope you solve your issue very soon.

Edited January 27, 2020 by tiborrr

...

Must be the motherboard and it's auto-overclocking function, ASUS has a function called MCE, which effectively overclocks all cores to the single-core limit. By doing so, it automatically overshoots Vcore by a significant margin as it uses single-core load Vcore using all-core load. Yikes!

...

 

Yikes indeed. I have a brand new ASUS motherboard in my new build. Note to self - Make sure the first thing I do is head into the BIOS and disable all this auto overclock nonsense until I have it bedded in !!

Sorry for the slow response, was kinda busy.

I did however take a wild gamble after remembering a post I saw a while back mentioning that if unlucky, the block might not be properly flat.

So I went and bought a new cpu block and quickly installed it today (the xc7 again of course, need the aesthetic of using the same brand)

And now the max temp it reaches is 70-75c when running the torture test in Prime95 (mind you the room temp is around 27c). I do still see a little bit of spiking, but not in the same way.

I also checked the bios, and it is running with MCE on.

I can also see that the package is reaching a max of 150w in CPUID, which also suggests it has been overclocked.

 

All in all I think it's pretty acceptable. Might try reducing the voltage and see if that will make the temps better whilst still running stable.

 

And thanks a ton for the help guys, I appreciate it.

70-75°C is excellent in P95 torture test. Do note the usual temperature differential (dT) is around 40-45°C for non-overclocked 9900K. 9900K is a furnace, but you still have good 25°C of thermal headroom left, plenty for more overclocking.

Yeah, and considering it was actually running all core 5Ghz I think I'm gonna try and reach 5.2. But this time manually instead of the machine doing it by itself.

That is great news man. Great news. If you want to use something that I think is a bit better than prime95. I use intel cpu burn. Will do all the work of prime95, but in a shorter time period. But in the end, to each his own.

Thanks man

Being able to test stability quicker is pretty valuable, I'll definitely try it out.

Just set it to anything other than standard. High, very high preferably. If it can make it though 10 tests, nothing is going to crash that system. Except bad code. But it won't crash because of anything temp related.

 

It might also be worth downloading the intel extreme tuning utility. It gives you the abliity to overclock specific cores. So if using 1 core it will turbo to a selected speed. if using 2 cores, etc. Very configurable. I have my 10900x at 4.5ghz on all 10 cores. Once I get the new corsair liquid cooling system installed, I am betting I can go higher.

Edited January 29, 2020 by acemaninwa871