[SOLVED] i9-10850K Thermal Throttling. Inadequate Cooling? or Bad Silicon?

drxuess

Distinguished
Feb 1, 2015
43
1
18,565
Hi Friends,

I've recently upgraded my computer to a i9-10850K and would really like your thoughts and opinions around some thermal issues I've been having...
For reference I've drawn up my current build and airflow:

ComponentPart
CPUIntel Core i9-10850K
CPU CoolerCorsair Hydro H100i GTX (240mm Radiator) + 2 120mm Noctua F-12 Industrial 3000RPM fans in push configuration
CasePhanteks Enthoo Evolv ITX
Front Intake Fan200mm Phanteks PH-F200SP Fan
GPUEVGA Hybrid RTX 2080Ti (120mm Radiator) + 12 120mm Noctua F-12 Industrial 3000RPM in pull configuration
Exhaust FanNoctua NF-A14 140mm Fan

17p46C2jJA5yM5sOYQDPdrBohT-FmHLnB


Issue:
When I stress test this computer using CineBench or similar with an all core stress test, all the cores temp go to 100C straight away and stays there. What I've noticed:
StateTempFrequencyVoltagePowerQuestions
Idle (Windows 10) - 10% CPU utilisation max55C (+28C ambient)4.4GHz - 4.8GHz1.3V -1.4V29W - 50WWeird considering base clock is supposed to be 3.6GHz so no idea why its always clocked so high?
Intel XTU Benchmax 90C (+63C ambient)4.8GHz1.4V50W - 200WIs 3818 Marks considerably low for this bench?
Intel XTU Stressmax 100C (+73C ambient)4.8GHz (4.4 throttled)~1.36V125W - 200WStarts off at 4.8GHz 200W then after around 20 secs it thermal throttles to bring the temp down. Then soon after it starts power throttling keeping the power usage at around 125W under TDP. When power throttled, temps generally floated around 84C. Is this normal?
CineBench All Core Stress Test - 100% CPU utilisationmax 100C (+73C ambient)4.8GHz (4.4 throttled)~1.36V125W - 200WSame as above.
Gaming (Cyberpunk 2077) - 33% -66%max 100C (+73C ambient)4.8GHz
~1.36V
160WSeems to keep running at 100C which worries me. It doesnt power throttle either like the benchmarks

I've used this cooling rig for mutliple CPUs now:
  • i7-4960X (130W TDP)
  • i7-6700K (91W TDP)
  • i7-8700K (95W TDP)
and I've never had any issues. In most cases the idles were roughly around 15-20C above ambient and at 100% load, I've never exceeded 90C in temps before... The biggest concern for me is when its in gaming, it seems to sit consistently between 95C to 100C which is not ideal. I've read online with people similarly reporting high temps but they only say its in the high 80s....

What I've tried:
  • Replacing thermal paste in case it didn't spread well... Using the Noctua NH-1 paste.
  • Changing the CPU fans to pull config to pull in cold air instead
  • Running the CPU Radiator outside of case in push configuration.
  • Spinning the case fans to max rpm (not ideal)
  • Undervolting the CPU by 0.05V (didn't help) too unstable when more undervolting occurred
  • Turned off MCE (saw it on a post in some other forum)
  • Disabling short term boost to keep TDP under 125W. (didn't seem to work, kept going over 125 W anyway)
Question: Am I doing something wrong? Is my cooler inadequate? (Do I need a 280mm-360mm rad?) Did I just get some bad luck and get a really bad piece of silicon? (Should I return?)

If any of you have had similar experience or could shed some light that would be great. I just want to be certain about what the problem is before I either return it, or invest in some more heavy duty cooling...

Again really appreciate your help.

Cheers,

Morgan
 
Last edited:
Solution
Are there any other settings within BIOS I could tweak to bring it down?
It's a little difficult since I don't know what your motherboard is...
1)Turning off Multi Core Enhancement(Asus) if it's even active. The other vendors call it All Core Enhancement instead.
2)Make sure C-states are active.
3)For cpu Core Voltage, select Offset Mode and enter an offset of -0.01v.
4)If you are running programs such as Msi's Dragon Center or Asus' AI Suite, uninstall them. I have at least heard of the former automatically applying a 'performance boost' - not sure about AI Suite, but it isn't very good anyway.
5)Manually enter the PL1, PL2, and Tau values yourself, but it sounded like you already tried that and to no effect, which makes me...

Phaaze88

Titan
Ambassador
What are gpu thermals like under load?

Also, what does cpu thermals look like in bios?


Looking up reviews - the Newegg ones tend to be more reliable than Amazon's - this chassis has poor overall airflow, with lots of air blowback with how they designed that top panel.
Oh, and that 200mm fan. They just don't do well with obstructions, period - they don't have the static pressure for it. 200mm fans should be used against open mesh panels.

With what I'm seeing so far, the chassis is just choking your hardware.
 
not all radiators/pumps are adequate enough, and, frankly, possibly too many many think all liquid solutions are more than sufficient when that is not the case. (I've seen folks confused that a 120 AIO was not adequate for a 9900K, etc.!) You'd have to sort out the comparisons of the assorted AIO 240 vs. 280 vs. 360 mm configurations...;but, clearly, your current configuration is not cutting it...

You can fine-tune all core turbo clock freqs a tad lower within Intel's XTU easily enough. (It it hits 88-90C under load at all-core 4.9 GHz, try a 'mere' 4.7 GHz, etc... (Setting the boost goals within XTU is child's play) That should hold you over while researching a better pump /radiator solution.

Good luck
 
  • Like
Reactions: drxuess

drxuess

Distinguished
Feb 1, 2015
43
1
18,565
What are gpu thermals like under load?

Also, what does cpu thermals look like in bios?


Looking up reviews - the Newegg ones tend to be more reliable than Amazon's - this chassis has poor overall airflow, with lots of air blowback with how they designed that top panel.
Oh, and that 200mm fan. They just don't do well with obstructions, period - they don't have the static pressure for it. 200mm fans should be used against open mesh panels.

With what I'm seeing so far, the chassis is just choking your hardware.

Thanks for responding! The GPU thermals are actually doing fine (they peak at around 60C during gaming and 75C under stress on kombustor) hence why I thought it was strange since I didn't have this issue earlier with any other CPUs. I am aware the 200mm might not be running suboptimally but thats why i took the radiator out of the case just to test, and the temps are pretty much the same... Is it just that the 10-series runs extremely hot?
 

drxuess

Distinguished
Feb 1, 2015
43
1
18,565
not all radiators/pumps are adequate enough, and, frankly, possibly too many many think all liquid solutions are more than sufficient when that is not the case. (I've seen folks confused that a 120 AIO was not adequate for a 9900K, etc.!) You'd have to sort out the comparisons of the assorted AIO 240 vs. 280 vs. 360 mm configurations...;but, clearly, your current configuration is not cutting it...

You can fine-tune all core turbo clock freqs a tad lower within Intel's XTU easily enough. (It it hits 88-90C under load at all-core 4.9 GHz, try a 'mere' 4.7 GHz, etc... (Setting the boost goals within XTU is child's play) That should hold you over while researching a better pump /radiator solution.

Good luck

Thanks! Yeh most definitely the first time I've encountered thermal problems with this cooler but I thought considering it had handled the 4960X like a champ another 125W TDP CPU should be no problem... I'll give the underclocking a test and see if it drops the temps :)
 

Phaaze88

Titan
Ambassador
Is it just that the 10-series runs extremely hot?
When they're not running the Intel defined power limits, 10th gen gets toasty.

I mean, if you're still hitting 100C with Turbo Boost off and holding the radiator outside the chassis... there must be something going on beneath the IHS.
Another thing that needs examining: the difference between the hottest and coolest cores. 10C and below would be considered normal, but this is going to be hard to determine if multiple cores are hitting 100C.
 

drxuess

Distinguished
Feb 1, 2015
43
1
18,565
When they're not running the Intel defined power limits, 10th gen gets toasty.

I mean, if you're still hitting 100C with Turbo Boost off and holding the radiator outside the chassis... there must be something going on beneath the IHS.
Another thing that needs examining: the difference between the hottest and coolest cores. 10C and below would be considered normal, but this is going to be hard to determine if multiple cores are hitting 100C.

Thanks! Yeh all cores seem to hit 100C which is why its worrying. Also is it normal in windows for it to always run at 4.8GHz even on idle? like the base clock for this thing is at 3.6 and it has never gone down.... What do you mean by running Intel defined power limits? I thought by turning off MCE the chip manages its own power states?
 

Phaaze88

Titan
Ambassador
10850K has a single core turbo up to 5.20ghz, and 4.8ghz all core, with Thermal Velocity Boost active up until 70C.
Some motherboards do not follow the stock Intel or AMD specifications out of the box, with the vendor opting for a performance tweaked profile.
It's a more common thing with the fancier boards. It's to make their products look better than the competition, but it tends to screw over the end user, because the cpu runs hotter.
If you're seeing different values from stock, it's likely been tweaked.

Also is it normal in windows for it to always run at 4.8GHz even on idle? like the base clock for this thing is at 3.6 and it has never gone down....
No, that doesn't sound normal. Even just idling on the desktop, Windows background tasks alone should have this cpu boosting to 5.2ghz on a couple of cores just to get them finished with faster, and dropping as low as 2.Xghz when not in use to save power.
Are you running the balanced power plan?

What do you mean by running Intel defined power limits?
Take the first paragraph in this post, and combine it with this:
View: https://imgur.com/YwqWtBb

^10th gen power table.
The 10850K shares the same power limits with the 10900K.
 

drxuess

Distinguished
Feb 1, 2015
43
1
18,565
10850K has a single core turbo up to 5.20ghz, and 4.8ghz all core, with Thermal Velocity Boost active up until 70C.
Some motherboards do not follow the stock Intel or AMD specifications out of the box, with the vendor opting for a performance tweaked profile.
It's a more common thing with the fancier boards. It's to make their products look better than the competition, but it tends to screw over the end user, because the cpu runs hotter.
If you're seeing different values from stock, it's likely been tweaked.

Thanks man that explains a lot. Yep looking at it I think you are right, single core boost to 5.2 + 4.8 all core. However I think because the CPU is consistently running at 4.8 and even with the slightest loads goes above 70c, I think that's why thermal boost is never activated. Are there any other settings within BIOS I could tweak to bring it down? Its pretty much stock right now.

No, that doesn't sound normal. Even just idling on the desktop, Windows background tasks alone should have this cpu boosting to 5.2ghz on a couple of cores just to get them finished with faster, and dropping as low as 2.Xghz when not in use to save power.
Are you running the balanced power plan?

Yep, I read about this many times but for some odd reason changing the power plan has no effect whatsoever on the clock speeds that it runs. I am currently enforcing a clock limit on XTU to keep the temp down. But whatever speed I set, all cores are basically running that clock speed nothing higher nothing lower. Sometimes a few cores might drop like 100MHz or so but nothing significant or even down to the base clock of 3.6... I'm thinking it might be the board itself tweaking something?


Take the first paragraph in this post, and combine it with this:
View: https://imgur.com/YwqWtBb

^10th gen power table.
The 10850K shares the same power limits with the 10900K.
Yeh I didn't know about the second power limit. But the numbers seem consistent with what I see around the power limit. Is the tau number how long the processor is allowed to run at that pl2 for?
 

Phaaze88

Titan
Ambassador
Are there any other settings within BIOS I could tweak to bring it down?
It's a little difficult since I don't know what your motherboard is...
1)Turning off Multi Core Enhancement(Asus) if it's even active. The other vendors call it All Core Enhancement instead.
2)Make sure C-states are active.
3)For cpu Core Voltage, select Offset Mode and enter an offset of -0.01v.
4)If you are running programs such as Msi's Dragon Center or Asus' AI Suite, uninstall them. I have at least heard of the former automatically applying a 'performance boost' - not sure about AI Suite, but it isn't very good anyway.
5)Manually enter the PL1, PL2, and Tau values yourself, but it sounded like you already tried that and to no effect, which makes me think some software is overriding that.
6)Windows Search > Edit Power Plan > Change advanced power settings > Processor power management > Maximum processor state. Set it to 98%, click Apply, and exit.
7)Enable Intel Speed Shift or Speed Step, but not both at the same time - there's no need to. I think Shift is the better one.

I'm thinking it might be the board itself tweaking something?
Maybe it's Thermal Velocity Boost kicking in once the cpu cools down even a little bit?
I'm sorry, this feature on the i9s is new to me. I don't know too much about it beyond that 70C limit.

Is the tau number how long the processor is allowed to run at that pl2 for?
Yep, and depending on the kind of workload, the power limit either:
-remains at PL1 until the work is finished(think Blender or Cinebench)
-or it literally gets to sit at PL2; variable loads can have the Tau timer reset before it finishes(games are not constant loads)
 
Last edited:
  • Like
Reactions: drxuess
Solution

Viorala

Reputable
May 20, 2019
150
19
4,615
Commenting to bookmark this.

FWIW, mine does the same thing.

 

drxuess

Distinguished
Feb 1, 2015
43
1
18,565
Thanks so much for you help so far @Phaaze88

It's a little difficult since I don't know what your motherboard is...
1)Turning off Multi Core Enhancement(Asus) if it's even active. The other vendors call it All Core Enhancement instead.
2)Make sure C-states are active.
3)For cpu Core Voltage, select Offset Mode and enter an offset of -0.01v.
4)If you are running programs such as Msi's Dragon Center or Asus' AI Suite, uninstall them. I have at least heard of the former automatically applying a 'performance boost' - not sure about AI Suite, but it isn't very good anyway.
5)Manually enter the PL1, PL2, and Tau values yourself, but it sounded like you already tried that and to no effect, which makes me think some software is overriding that.
6)Windows Search > Edit Power Plan > Change advanced power settings > Processor power management > Maximum processor state. Set it to 98%, click Apply, and exit.
7)Enable Intel Speed Shift or Speed Step, but not both at the same time - there's no need to. I think Shift is the better one.
Yeh sorry about that its an ASUS ROG STRIX Z490I but I think uninstalling AI suite did the trick! The clocks are dropping down to 3.6 Ghz on idle :)

Maybe it's Thermal Velocity Boost kicking in once the cpu cools down even a little bit?
I'm sorry, this feature on the i9s is new to me. I don't know too much about it beyond that 70C limit.
Yeh its really weird moving from an older intel chip. I actually have Thermal Boost Disabled earlier because I thought it might be the issue.

Yep to be honest in hindsight my i7-8700K (3.7Ghz - 4.7Ghz) was a 6/12 chip and that ran around 38C idle to about 75C load. I may have been too ambitious to think that the cooler could maintain similar cooling for this chip which has 4 extra cores + runs 4.8 boost up to 5.2...
I'll be on the lookout for a better cooling solution. Maybe its time to dip into custom loops again or the new ML360 Sub-zero haha.

@Viorala for me it seemed that the motherboard software was keeping the cores on load all the time and as a result it was never allowed drop to its base clock. Not sure if you have the same problem... I also switched voltage mode to offset with -0.01 if that helps.
I also underclocked for the CPU to 4.5GHz instead for the boost limit. This has kept my CPU floating around 75C-82C for the meantime until I get better cooling solution. Please let me know if you find any other alternatives :)
 

Phaaze88

Titan
Ambassador
Yeh sorry about that its an ASUS ROG STRIX Z490I but I think uninstalling AI suite did the trick! The clocks are dropping down to 3.6 Ghz on idle :)
Nice! (y)(y)

I actually have Thermal Boost Disabled earlier because I thought it might be the issue.
I think I would leave it enabled. It's like free performance - when it's active, anyway.

Yep to be honest in hindsight my i7-8700K (3.7Ghz - 4.7Ghz) was a 6/12 chip and that ran around 38C idle to about 75C load. I may have been too ambitious to think that the cooler could maintain similar cooling for this chip which has 4 extra cores + runs 4.8 boost up to 5.2...
I'll be on the lookout for a better cooling solution. Maybe its time to dip into custom loops again or the new ML360 Sub-zero haha.
That probably applies to most people who stayed in the Intel camp.
They stayed on the same 14nm process since Core i 6xxx and just kept squeezing more speed out of them with each 'new' release - there were some refinements here and there, but... ehh.
Their latest chips are even faster and have more cores/threads... simple physics says they're gonna be hotter - especially if you throw the power limits out the window, a-la overclocking.


Once I'm off the X299 platform, I'm giving up OC'ing - well, I've already given it up on my gpu since hybrid cooling it. So, let's change that to giving up cpu OC'ing.
With there being more cpus and gpus that can push for higher speeds on their own as long as you give them the 'cooling budget' to do so, the endeavor of manually overclocking has lost value in my eyes.
I'm not a min-maxer; I did it to sate my curiousity. I did get some enjoyment out of it and learned some things throughout.
I shouldn't go into a 3 year life story over this, so I'll cut it here.
 
  • Like
Reactions: drxuess