The Corsair User Forums  

Go Back   The Corsair User Forums > Corsair Enthusiasts Section > Overclocking and Benchmarking

Notices

Reply
 
Thread Tools Search this Thread Rate Thread Display Modes
  #1  
Old 01-28-2020, 01:07 AM
Caleb_Dume Caleb_Dume is offline
Registered User
Caleb_Dume's PC Specs
 
Join Date: Jan 2020
Posts: 4
POST ID # = 1033089
Caleb_Dume Reputation: 10
Default Are my temperatures normal?

Hi Guys,

I have a h150i cooler. I am OCing an i9 9900k to all core 5Ghz.
I have AVX offset to -2.

When I run prime with AVX enabled, after about 7 minutes, my CPU package temperature goes above 90C and the CPU starts throttling at 100C.

Is this normal, or do I have an issue?

The temperature gauge in iCUE says 27C when idle. Assume this is pump liquid temperature?

I have the fans on balanced and pump on extreme. Under prime,
the iCUE temperature is 32C.

This is my first OC, so not sure if these temperatures are normal or not.

Thanks for any help!
Reply With Quote


  #2  
Old 01-29-2020, 11:34 AM
c-attack c-attack is offline
Registered User
c-attack's PC Specs
 
Join Date: Jun 2014
Posts: 12,294
POST ID # = 1033242
c-attack c-attack Reputation: 122
Default

Quote:
Originally Posted by Caleb_Dume View Post
The temperature gauge in iCUE says 27C when idle. Assume this is pump liquid temperature?
That is correct and the 60-70C difference between coolant and CPU temp suggests one of two things. You either have a physical contact problem between the CPU and cold plate or you need to set the BIOS properly to handle synthetic stress tests. Someone running 5GHz on all cores on a 9900K would typically have a differential of 40-50C -- not 60-70C and that is where the extra 20C come from.

The contact problem will present itself all the time. Idle temps will be 40-50C. Core temps will jump into the 60-70C range for trivial tasks like opening folders or launching browsers. Coolant temp never budges because heat is not being transferred across to the cooling system. When this is really bad, your stress test fails almost instantly. If you are just a little off, the initial temp is too high and very close to thermal shutdown.

The BIOS issue is far more common on the last several generations of Intel boards. They are designed to boot up and run fast with even the worst CPU of all time and the magic ingredient in that is voltage. It is usually not possible to test most recent motherboards with Prime 95 in pure default trim. It just keeps pouring on voltage. That is the missing data point in your information. Not just what you've set, but the actual level when under 100% load. Those are two different things.

Also, since Prime 95 is not really helping at the moment, get off it and use a milder version for testing. Try the CPU-Z stress test under the bench tab. It is linear. It should start off at XX CPU temp and basically stay there, only rising for each +1C of coolant temp increase. That should make it easy to monitor both voltage and coolant temp to see what's going on.
Reply With Quote


  #3  
Old 01-30-2020, 11:06 PM
Caleb_Dume Caleb_Dume is offline
Registered User
Caleb_Dume's PC Specs
 
Join Date: Jan 2020
Posts: 4
POST ID # = 1033446
Caleb_Dume Reputation: 10
Default

Hi, I took off my OC and went back to defaults (XMP enabled).
I wanted to see what would happen if I went back to default BIOS (no OC) and tried Prime w/AVX.

The temp's never got over 70C. Maximum Vcore was 1.2V. LCD Shut off after 10m and computer went to sleep after 30m. Strange W10 goes to sleep when Prime is running.
Anyways, let Prime w/AVX run for 50m total. No issues. Computer resumed from sleep just fine and continued running prime. All cores were running 4.7GHz. Makes me wonder why I am bothering for 300MHz more.
Here's the screenshots:



Reply With Quote


  #4  
Old 01-30-2020, 11:59 PM
Caleb_Dume Caleb_Dume is offline
Registered User
Caleb_Dume's PC Specs
 
Join Date: Jan 2020
Posts: 4
POST ID # = 1033450
Caleb_Dume Reputation: 10
Default

I think I am starting to understand the reason I am over 90C 7m into Prime w/AVX (AVX offset -3 now).

Look at HWINFO in previous post and this one. Core frequency is the same. Voltage is less! So, the only difference is Ring/LLC clock! It's 400MHz more and seems to blow up the temps!

Reply With Quote


  #5  
Old 01-31-2020, 06:56 AM
c-attack c-attack is offline
Registered User
c-attack's PC Specs
 
Join Date: Jun 2014
Posts: 12,294
POST ID # = 1033462
c-attack c-attack Reputation: 122
Default

Core VID and “Vcore” are two different things. The actual applied voltage (Vcore) is hiding in the motherboard section - bottom center column. It jumped to 1.36v and there is the problem. 1.35v is the absolute sane limit for a 9900 series. It has a different scale than the last several 8700/7700 CPUs and the whole curve is offset -0.10. Regardless, this is good news and means you didn’t do anything wrong on the install. However, you absolutely cannot “auto overclock” modern CPUs by raising frequency and letting the bios pick the voltage. You’ll need to make some manual changes.

There probably are detailed well planned overclocking guides for the 9900 series CPUs. My experience is on Asus board for the last several and the terminology is different. This is very BIOS specific.

1) You are going to want to set a Vcore in the 1.25-1.30v range. Currently you are at 1.36 on Auto. You can still use adaptive voltage, but it need to have a specified limit. The way this is done is specific to the manufacturer.

2) Most MBs have some type of "auto multi core overclock" system. This sets all cores to the turbo max frequency. That is fine and how most people want to overclock, but do this by setting all cores to 50 (multiplier) x 100 rather than using the multi-core enhancement or whatever MSI calls it. There are usually hidden parameters.

3) Several of the 390 series boards now have a VID or expected voltage modifying element. On asus this is described as "Best case scenario, normal, worst, etc". This effectively pads (or no pad) the voltage curve when under stress. In your current set up, the VID is 1.22 and yet the BIOS is delivering 1.36. That is classic worst case scenario and obviously way to much.

4) There may be other specific MSI settings. This is why an MSI specific guide is best, if you can find it.

Last edited by c-attack; 01-31-2020 at 10:55 AM.
Reply With Quote


  #6  
Old 02-01-2020, 01:08 PM
Caleb_Dume Caleb_Dume is offline
Registered User
Caleb_Dume's PC Specs
 
Join Date: Jan 2020
Posts: 4
POST ID # = 1033569
Caleb_Dume Reputation: 10
Default

Thanks for the help! I set the Vcore to 1.32V, but as you see in the screenshot I posted, it is going over the 1.32V. CPU Core Voltage Mode was in Override.

I'll post my bios screenshots:







Reply With Quote


  #7  
Old 02-01-2020, 06:50 PM
c-attack c-attack is offline
Registered User
c-attack's PC Specs
 
Join Date: Jun 2014
Posts: 12,294
POST ID # = 1033584
c-attack c-attack Reputation: 122
Default

It will be pretty tedious for me to request sub menu pictures for all those so I can learn your BIOS. You've selected 1.32v as the target and since that is the voltage recommended in the MSI blog guide here, I suspect you found a decent source. I would follow their recommendations to start, check temps, and then you can begin to chip little bits off the 1.32 value to see if you can bring it down further. CPU voltage levels vary from chip to chip, so they need to be pretty conservative in a broad spectrum guide. I know people who run their 9900K@1.25v for 5GHz.

AVX offsets - I am of the opinion if you run programs with AVX, then your overclock should be able to handle that or don't clock that high. For that reason, along with unexpected programs using AVX instructions, I set mine to 0 and only clock as high as I am AVX stable. However, even if you don't fall in line with that philosophy, you want an AVX and non-AVX data point for temperature. You can certainly run Prime at 1.32 to see how much better things are, but I would also run the CPU-Z stress test (bench tab). Non-AVX, linear, relatively mild. This will test you at 5.0 vs 4.8 and with voltage to match. It is also a more realistic indicator of the temps you will see. The temp you see 1 second in is the max temp, except for additional rise to the coolant temp (+1 coolant temp = +1C CPU temp). I often use this test as a cooler check/assessment. I don't think there is need for that here, but it still helps people understand how it works. Either way, you don't need to run this for hours. 30 seconds will tell you if the voltage is too high or not. 10 min will be a decent cooler evaluation and show you your limits. And coolant temp creep should peak before the 10 minutes are up.

Overclock.net is also usually a good source for motherboard specific advice and there is generally a thread for each board so you can help directly from other owners. Here is on for MSI Z390.

https://www.overclock.net/forum/6-in...rd-thread.html
Reply With Quote


Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -4. The time now is 05:41 AM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2020, vBulletin Solutions, Inc.