[SOLVED] I9 10900k NZXT Prebuild overheating above 92c when running cinebench on 1.3V @ 4.7ghz all core with NZXT Kraken Z63 - 280mm

Jul 17, 2021
6
0
10
2 days ago I purchased a prebuild with these specifications.


  • NZXT Kraken Z63 with LCD display
  • Standard Service (US)
  • Microsoft Windows 10_Home
  • NZXT H510 Elite (Black)
  • MSI Z590 ACE
  • Intel Core i9-10900K 10-Core 3.7 GHz
  • Team T-FORCE XTREEM ARGB 4000MHz 4000 MHz (max. Geschwindigkeit) 32GB (2X16GB) More Info
  • NVIDIA GeForce RTX 3080 Ti - EVGA FTW3 ULTRA GAMING
  • EVGA SuperNOVA 1000W G5 Gold
  • Seagate FireCuda 520 Cyberpunk 2077 Limited Edition 1 TB
  • NZXT Aer RGB 2 140mm (RGB)
  • RGB & Fan Controller
  • Interner USB-Hub (Gen 3)

I have enabled my XMP profile and then ran cinebench with enhanced turbo enabled and the cpu bios settings all on default/stock. After about 5 minutes under load the temps were hitting above 92c and my distance to tjmax being under 8. The cpu core voltage looking at HWinfo was somewhere around 1.35V average under full load as it was on auto in the bios.

I then tested it again but first went into the bios and set the frequency per core manually to 4700mhz per core as well as changed the Voltage to 1.3 with loadline calibration set to auto and enhanced turbo disabled to see if I would get a different result. From my understanding with my cooler and my chip, I should be able to exceed 4.7 with ease and near the 5.0 range without too much issues, despite chips varying from chip to chip.

When I booted, I went back into cinebench and tested again with the same results. Temps hitting as high as 92c and distance to TJmax being around 8 at the worst moments. This was only a 10 minute test.

My CPU is idling around 29-33c regardless of the voltage I set as I have changed the frequency and voltage to 3.7 and 1.225 to see if there was a difference in the idling temps as well as the cinebench tests. When I did the test on this low frequency and low voltage, clearly the temps were lower. My highest core temp during the test was 74c and my distance to TJmax was 26. Idling temps are about the same around 29-33c.

At this point, I am considering multiple options and would like some feedback on them as well as some additional ideas for OCing or just in general about this matter.

1. replace the thermal paste and hope for a different outcome

2. move the AIO to the top of the case(which likely wont fit as I bought it prebuild from NZXT and the case is tiny,)(Read something about how the way they install them can cause performance issues with the air)

3. buy a whole new case that can fit a better cooler

4. Reconsider the cooling strategy in the build which I have no idea about

I would ideally like to at least hit 5 ghz on my 10900k on all cores as I feel that is a very reasonable overclock for this chip. Most OC on this chip I have seen people started their OC testing at 5.0 but I am having issues at even 4.7 due to the system overheating at full load at only 1.3V
 
Solution
You would have to (manually) bring the voltage down. Start with dynamic voltage default value with -0.05 offset and make your way down as much as it would go staying stable. Use small increments and run stability tests. I would leave 4.7 until the temps are under control.
I would use Intel Extreme Tuning Utility for that.
If 1.3v is the lowest stable voltage for 4.7, you've lost the silicon lottery…
You would have to (manually) bring the voltage down. Start with dynamic voltage default value with -0.05 offset and make your way down as much as it would go staying stable. Use small increments and run stability tests. I would leave 4.7 until the temps are under control.
I would use Intel Extreme Tuning Utility for that.
If 1.3v is the lowest stable voltage for 4.7, you've lost the silicon lottery…
 
Last edited:
Solution
Jul 17, 2021
6
0
10
Totally understand that, I suppose my end goal is to be able to be in the ranges of 1.3-1.36V regardless. So either something is wrong with my cooling system or my cooling system just sucks. Would like an answer more leaned towards fixing the problem rather than accepting it and reducing the voltage. Hopefully everyone who sees this understands my main goal.
 
Totally understand that, I suppose my end goal is to be able to be in the ranges of 1.3-1.36V regardless.
  1. Your goal is to be 1.1-ish or lower most of the time (dynamic load/frequency)
  2. It may spike higher for short periods and your cooling should be able to accommodate those spikes
  3. 1.3v will produce large amounts of heat with such a fat chip. I have two systems with 280mm aio, and both would probably boil with this voltage…
  4. Forgot to mention AVX offset, at least -2 for starters
Here is how it looks on my ITX i9 build (disregard the numbers, it is detuned for the ITX case, find your own).
5KoOQjw.png
 
Last edited:
Here is my settings Multi Core Enhancements- Enabled Remove all limits All Core Ratio- Set at 50 Cpu Core/Cache Voltage - Manual Mode Cpu Voltage Override- set at 1.365 volts I game alot. ( You will have to experiment with your own voltage setting for your usage) I have a 360 cooler and temps never go above 70c in heavy gaming with own fan curve set.
 

Phaaze88

Titan
Ambassador
Those are some high power parts(cpu + gpu) you've got in there, along with a chassis that does not provide for such a config very well.
PCs are space heaters. Air needs to enter and exit effectively. NZXT's H500 has restricted intake and exhaust - it does work ok with less power hungry configs though.
Negative offsets and undervolting will help some, but ultimately, the cpu and gpu cooling efficiency is only as good as what the chassis allows; you still could run into scenarios where one or the other gets too hot for your liking.


Weakest link is the chassis, IMO. Popular for its looks, but it's design doesn't play well with certain configs, namely those with high energy use:
-a cpu, when OC'ed and run under stress tests, can pull up to* 300w+ of power if the limit is removed. [These devices dynamically adjust themselves if allowed to, so it's not like they'll sit at peak draw all the time.]
Sure, you won't see that kind of energy use everyday, but you'll still need to test OC stability if you are going to go that route...

-a gpu, which can draw up to* 450w, according to the vbios. By far the bigger power user of the 2 here.
I'm not going to get into the whole transient power spike thing that these cards are notorious for, because it complicates things.

Put both of those potential numbers together with the H500 effectively having 2 small areas of exhaust at the rear and top...
There's a traffic jam with exhausting heat.