Unable to cool 9900KS @5 Ghz on all cores without any limits

[Okys]

Hello guys.
Please help me understand what is wrong in my system or what I am doing wrong.
I recently bought some new hardware for my wifes PC:
ASUS Maximus XI Hero (just for Aura)
i9 9900KS
G.Skill F4-3600C16D-16GTZR
Thermaltake Water 3.0 ARGB Sync 360 AIO
A-Data XPG SX8200 Pro 1TB SSD
Old Gigabyte GTX970 (GV-N970G1 GAMING-4GD) videocard (saving money for something better later)
Everything is put inside a be quiet! Silent Base 801 (window) case and powered by Seasonic Prime Ultra Titanium 750W power supply.
Radiator is installed on top of the case with fans in push configuration, pushing air from case through the radiator. In front I installed 3x Silent wings 2 (3pin) 140mm fans (those came with the case) as intake for fresh air. On rear I put one 140mm Silent Wings 2 PWM fan as exhaust.
It's running Windows 10 Pro (1903) with all the latest drivers and 1302 BIOS.
So, at first, I set the XMP 2 profile in BIOS for RAM and left everything else untouched. Later, when I got a chance, I ran Cinebench 20 and noticed that PLL thing. It ran at 5 Ghz for some time, then dropped to around 4.7 Ghz at the end, to stay in the TDP of 127W.
I do not plan to overclock at all, but wanted all cores to run 5 Ghz all the time, when needed. I was reading 9900KS should be capable of that, so I went into bios, set it to disable any power limitations, set AVX offset to 0, kept voltage on auto and everything else on auto.
And here came the problem.. on a single cinebench 20 run, temperatures rocketed to 96 degrees C on some cores. If i remember right, voltage was around 1.33v. I didn't like that, so i thought, lets change the Thermaltakes pre-installed thermal paste to Thermal grizzly Kryonaut. It actually helped to reduce 2-3 degrees, but still i wasn't impressed... I was getting the score of around 5210-5225. Thats about the same, I found on the internet from another 9900KS (!!STOCK!!) reviews.
I also tried stress test from AIDA64, including GPU in it. CPU throttled after 8-10 seconds...
I then returned to BIOS, switched the voltage from Auto to Adaptive, with + sign.
Came back to windows, ran Cinebench, and wow. Temperatures dropped by around 5 degrees and also the score. Now, I was only getting like 5020-5040 score. With same TDP around 180 watts and same 5 Ghz clock (checked in HWMonitor).
At this point i decided I am an idiot, doing something wrong and came to google for help. It only made me feel even worse, after reading that some run 5.2ghz on air, some other easy does 5.2ghz, etc. How on earth is it possible, if i can't even run mine constantly on 5 Ghz?
I also tried these tests with open front, open cover... It almost doesnt matter at all... temperatures rocket to over 90 degrees anyway.
So my questions are:
How changing ONLY voltage from Auto to Adaptive+ was able to reduce temps by around 5 degrees and Cinebench score by 200?
What I am doing wrong and what could be wrong in my configuation? Or maybe nothing is wrong? I just want to run a stable, advertised 5Ghz on all cores, in any situation, while keeping the temps below 90 degrees. Is that even possible? I can consider installing radiator on front and fans in pull config, or even push-pull, but I tried with open case and open front without good results, so I could end up in the same situation...

Sorry for my weird european english and thanks in advance!

 

[Deleted member 185158]

The TIM under the IHS plate is probably the issue and has been for a long time with Intel chips. You'll find lots of reads with vastly improved temps replacing the TIM under the IHS plate.

The rest, pretty normal. It's throttling from 5ghz because it's just too hot.

You could gain some Cinebench points with Ram tweaking and overclocking too btw. It's not all CPU that gives good scores.

 

[FreedomEclipse]

The TIM under the IHS plate is probably the issue and has been for a long time with Intel chips. You'll find lots of reads with vastly improved temps replacing the TIM under the IHS plate.

its soldered on. Delidding a soldered CPU makes little to no difference. and when i say little i mean 1-2'c. plus the added risk you run of damaging your CPU when youre trying to scrape the solder off the CPU die with a razor blade or a craft knife is a lot lot higher.

 

[R-T-B]

The TIM under the IHS plate is probably the issue and has been for a long time with Intel chips.

This is a soldered chip.

There are always dud chips that still technically meet spec. Mine is one. My present 9900K (basically your chip) won't go an eyelash over 4.8 Ghz allcore. It's still an awesome chip... don't worry so much about the "even 5" number.

 

[Deleted member 185158]

Thanks fellas. Was not aware it was soldered.

 

[sneekypeet]

Any chance that the pump is running stupidly slow or not at all?

 

[Deleted member 185158]

its soldered on. Delidding a soldered CPU makes little to no difference. and when i say little i mean 1-2'c. plus the added risk you run of damaging your CPU when youre trying to scrape the solder off the CPU die with a razor blade or a craft knife is a lot lot higher.

Was on mobile on the last reply, so couldn't portray more on de-lidding....

Well aware of the ups and downs of de-lidding. Just wasn't aware the 9900KS is soldered.
For sure de-lidding soldered chips is for enthusiasts. I've done quite a few myself, but mostly AMD processors.

Not sure about the 1-2c claim. This really depends on a few factors such as re-using the IHS plate with LM (which should have worse temps) or using a lidless plate and direct cooler contact.
I don't typically recommend de-lids on soldered processors, I'm sorry to the OP for that suggestion. I'll research more on which Intel chips are soldered and which are not so I don't make this mistake once again. Thank you for the correction.

 

[Vario]

there aren't any real world gains to be made at 5 GHz vs 4.5 GHz, don't drive yourself crazy trying to achieve it.

 

[lsevald]

The solder they use seems pretty bad, maybe the solder layer is too thick or something. Anyway, I'm running my 9900k naked, no IHS at all, and use liquid metal directly between the die and a nickel plated water block. Gained about 18°C better temps (full P95 AVX load, quick test before and after at the same OC'ed settings), posted about here. My process was a bit too risky, if I were do do it again, I probably would try this kit.

 

[Deleted member 185158]

The solder they use seems pretty bad, maybe the solder layer is too thick or something. Anyway, I'm running my 9900k naked, no IHS at all, and use liquid metal directly between the die and a nickel plated water block. Gained about 18°C better temps (full P95 AVX load, quick test before and after at the same OC'ed settings), posted about here. My process was a bit too risky, if I were do do it again, I probably would try this kit.

Thats very nice results.
Been thinking to delid my 8700k, but it already benches at 5.4ghz on stock TIM. Still contemplating the worthyness while Ill be closer to 1.6v at 5.5ghz vs the 1.520v at 5.4ghz. Starts to become a hog.
Obviously Im not using this daily... Well I bench on it almost daily, but with sub ambient liquid cooling instead.

 

[Agentbb007]

I just ran Cinebench R20 on my 9900K running all core 4.9GHz. Highest temperature was 87C on core 2, Vcore reported 1.225V during benchmark. Obviously with these temps my X62 fans and pump all run at 100%. My CPU voltage is set to Adaptive Mode + Add Turbo Voltage=1.2 Offset Voltage=0.05.
I've attached some screenshots of my sensors during benchmark and UEFI voltage so you can compare with your setup. I like what sneekypeet suggested, make sure your pump is working and everything is blasting to 100%.

 

[bubbleawsome]

Everyone saying "don't focus on the 5Ghz" is out of their mind. The entire point of the chip is the 5Ghz all-core turbo. If his can't make it without throttling it's not getting it's rated speed, and should be warrantied.

 

[Okys]

Thanks for many answers.

I just ran Cinebench R20 on my 9900K running all core 4.9GHz. Highest temperature was 87C on core 2, Vcore reported 1.225V during benchmark. Obviously with these temps my X62 fans and pump all run at 100%. My CPU voltage is set to Adaptive Mode + Add Turbo Voltage=1.2 Offset Voltage=0.05.
I've attached some screenshots of my sensors during benchmark and UEFI voltage so you can compare with your setup. I like what sneekypeet suggested, make sure your pump is working and everything is blasting to 100%.

I will definately check the pump tomorrow. It has only 3 pins, btw. This is first time I'm ever using an AIO, because my wife wanted one so badly. Im happy with my Noctua NH-D15S, but she wanted the looks..
And seeing youre running the same bench at 87C, 188Watts on 9900K using Kraken X62, makes me think something is wrong for me...
I will be back tomorrow with more info on this.

 

[FreedomEclipse]

Was on mobile on the last reply, so couldn't portray more on de-lidding....

Well aware of the ups and downs of de-lidding. Just wasn't aware the 9900KS is soldered.
For sure de-lidding soldered chips is for enthusiasts. I've done quite a few myself, but mostly AMD processors.

Not sure about the 1-2c claim. This really depends on a few factors such as re-using the IHS plate with LM (which should have worse temps) or using a lidless plate and direct cooler contact.
I don't typically recommend de-lids on soldered processors, I'm sorry to the OP for that suggestion. I'll research more on which Intel chips are soldered and which are not so I don't make this mistake once again. Thank you for the correction.

Maybe the 1-2'c might of been a little exaggerated. but even a 4'c difference is a waste of time unless youre going for extreme OCs or bench runs with LN2. Still not worth the risk of deliding. I would check other things like if the cooler is making good contact or if the AIO might even be faulty. We all know 9900Ks run hot at 5Ghz anyway.

:EDIT:

also if you want to read the article...

 

[Okys]

Everyone saying "don't focus on the 5Ghz" is out of their mind. The entire point of the chip is the 5Ghz all-core turbo. If his can't make it without throttling it's not getting it's rated speed, and should be warrantied.

I can finish Cinebench R20, several times in a row and it doesn't throttle. I just dont think temps should be that high using a 360mm AIO. Only thing that throttled almost immediately, was AIDA64 stress test with GPU included in the test.

 

[TheMadDutchDude]

Running at 5G all the time is an overclock. It is meant to run at 4.8 after 27 seconds by default. Anything above that for an infinite amount of time IS and always WAS an overclock.

 

[Okys]

I would check other things like if the cooler is making good contact or if the AIO might even be faulty. We all know 9900Ks run hot at 5Ghz anyway.

On first try I installed the block using Thermaltakes thermal paste. After taking it off and examining, everything was kinda ok with contact. I then cleaned both surfaces, applied Thermal Grizzly Kryonaut and temps improved by 2-3 degrees.
How do I inspect a faulty AIO? I will check what I can about the pump, but not sure about everything else..

Running at 5G all the time is an overclock. It is meant to run at 4.8 after 27 seconds by default. Anything above that for an infinite amount of time IS and always WAS an overclock.

I may agree on that, but seeing 9900K running at lower temps, more TDP and using 280mm AIO, makes me feel even worse for my "premium" purchase

 

[FreedomEclipse]

On first try I installed the block using Thermaltakes thermal paste. After taking it off and examining, everything was kinda ok with contact. I then cleaned both surfaces, applied Thermal Grizzly Kryonaut and temps improved by 2-3 degrees.
How do I inspect a faulty AIO? I will check what I can about the pump, but not sure about everything else..


I may agree on that, but seeing 9900K running at lower temps, more TDP and using 280mm AIO, makes me feel even worse for my "premium" purchase

Normally its pump noise but if thats working correctly. Its clearly working as youre not throttling or going into thermal shutdown.

On the other hand, your PC case....

Its a bit of a hotbox. The words 'silence' and 'performance' dont really go together in some cases unless you throw a lot of money at it. If you want a silent machine then it will come at the cost of performance. If you want Performance then that will come at the cost of silence - Both these can kind of be worked around if you throw money at custom water cooling but that is an expensive route.

Best thing you can really do is either get a new case or get fans that are static pressure orientated like Corsair MLs or SPs because they will pull more air into the case then normal airflow fans - even then this is more of a bandaid over a cut then addressing the actual problem.

Its either one or the other, a middleground is extremely rare but it falls down to how much noise you are willing to tolerate.

 

[oxrufiioxo]

On first try I installed the block using Thermaltakes thermal paste. After taking it off and examining, everything was kinda ok with contact. I then cleaned both surfaces, applied Thermal Grizzly Kryonaut and temps improved by 2-3 degrees.
How do I inspect a faulty AIO? I will check what I can about the pump, but not sure about everything else..


I may agree on that, but seeing 9900K running at lower temps, more TDP and using 280mm AIO, makes me feel even worse for my "premium" purchase

A lot of 280 AIO beat 360 AIO just look at the h115i platinum Vs the h150i pro.
I can't find any reviews on your specific model so it hard to say if its adequate or not. Most reviewers test on an open bench in a 20-22c room.

My guess is either your ambient is high or the pump in that specific aio is weak.

My vanilla 9900k is cooler at 5ghz than your chip ( h150i pro ) you should be able to run around 1.25V and achieve 5ghz on the KS model according to reviews.

5ghz all core score Although comparing this system to system without identical specs is pointless.


Temps doing 3 back to back runs. 22C ambient Front mounted radiator fans fixed at 1500RPM. I keep my ambient under 23C year round though.

 

[Okys]

Guys, I just checked the pump... So, its a 3-pin, i have connected it to the AIO Pump connector. In BIOS the "AIO pump speed" is reported as 1760-1778 RPM and the AIO pump control disabled at Q-Fan Control, so, in my understanding, if nothing controls the pump, it must be running at it's full speed, which is advertised as 3000 RPM.
I also tried DC Mode (for 3-pin connectors) and selected the low-mid-high (30-40-70) temp limits and added 100% percentage to every limit, just to check if something would change.
After entering BIOS again, the speed is still reported around 1778 RPM.
I am not sure where to check AIO pump speed in windows.

An update:
The same speed is reported on HWMonitor as Chassis #6 speed. And it doesn't increase while doing a benchmark., just staying at those 1778 RPM's...

Normally its pump noise but if thats working correctly. Its clearly working as youre not throttling or going into thermal shutdown.

On the other hand, your PC case....

Its a bit of a hotbox.

I agree and was aware of that, but wife liked the case, especially the orange one...
That's why i took the front completely off, so, the 3x140mm fans could grab every possible cool air. I opened the side cover also, to check. There was a little to no change on the temps..

A lot of 280 AIO beat 360 AIO just look at the h115i platinum Vs the h150i pro.
I can't find any reviews on your specific model so it hard to say if its adequate or not. Most reviewers test on an open bench in a 20-22c room.

My guess is either your ambient is high or the pump in that specific aio is weak.

My vanilla 9900k is cooler at 5ghz than your chip ( h150i pro ) you should be able to run around 1.25V and achieve 5ghz on the KS model according to reviews.

The ambient temperature is around 25-26 degrees C. This Thermaltake Water 3.0 ARGB was quite good on reviews I was able to find before buying. I could have bought Corsair or Kraken, i think they are all similar Asetek. Wife choose this one.

 

[Deleted member 185158]

Maybe the 1-2'c might of been a little exaggerated. but even a 4'c difference is a waste of time unless youre going for extreme OCs or bench runs with LN2. Still not worth the risk of deliding. I would check other things like if the cooler is making good contact or if the AIO might even be faulty. We all know 9900Ks run hot at 5Ghz anyway.

:EDIT:

also if you want to read the article...

Interesting read. Found benching points increases, lower temps and lower fan RPM.
Very well noted at the very bottom of the article that this would be for enthusiasts for sure, looking to squuek that extra mhz.

Pretty cool. That's what I always look to do!!! Not until recently have I gotten an Intel Rig for benching, but I wanted to try my hand in 3D benches so needed an Intel rig to do so. I'm still on a very steep learning curve. But having fun is what I like to do most.

through the years, been mostly an AMD enthusiast. I've de-lidded quite a few soldered processors. It's rather easy as I've done quite a few and the risk is not as high as you'd think. There's always that chance of a mistake. I've de-lidded 20 and killed 2 in the process in the last 10 years or so.

Thank you for sharing that link. Very interesting to me!

Guys, I just checked the pump... So, its a 3-pin, i have connected it to the AIO Pump connector. In BIOS the "AIO pump speed" is reported as 1760-1778 RPM and the AIO pump control disabled at Q-Fan Control, so, in my understanding, if nothing controls the pump, it must be running at it's full speed, which is advertised as 3000 RPM.
I also tried DC Mode (for 3-pin connectors) and selected the low-mid-high (30-40-70) temp limits and added 100% percentage to every limit, just to check if something would change.
After entering BIOS again, the speed is still reported around 1778 RPM.
I am not sure where to check AIO pump speed in windows.



I agree and was aware of that, but wife liked the case, especially the orange one...
That's why i took the front completely off, so, the 3x140mm fans could grab every possible cool air. I opened the side cover also, to check. There was a little to no change on the temps..



The ambient temperature is around 25-26 degrees C. This Thermaltake Water 3.0 ARGB was quite good on reviews I was able to find before buying. I could have bought Corsair or Kraken, i think they are all similar Asetek. Wife choose this one.

hit F6 in bios and see if you can manually set the pump speed.

 

[FreedomEclipse]

You can use HWmonitor to monitor a lot of things.

Since the pump is 3-pin anyway it should always run at max rpm - my corsair H105 did exactly the same thing. What fans do you have on the front of the case??

through the years, been mostly an AMD enthusiast. I've de-lidded quite a few soldered processors. It's rather easy as I've done quite a few and the risk is not as high as you'd think. There's always that chance of a mistake. I've de-lidded 20 and killed 2 in the process in the last 10 years or so.

Not everyone is a dab hand like you friend. I have never delidded a CPU though i'll get around to it some day.

 

[Deleted member 185158]

You can use HWmonitor to monitor a lot of things.

Since the pump is 3-pin anyway it should always run at max rpm - my corsair H105 did exactly the same thing. What fans do you have on the front of the case??



Not everyone is a dab hand like you friend. I have never delidded a CPU though i'll get around to it some day.

Yes I often forget that. My experience also differs from other enthusiasts. Amd temp dropping is harder to come by than Intel for sure. Id kill to knock of 14c with a delid.... My average is only about 10c through a variety of processors. This is because some processors where low wattage and you could only tell the difference with high volt overclocking. Ive done some of that 2v FX on LN2. Its a real gas.

If I could turn the average users into overclocking beasts, PC gaming and overclocking would thrive again.

Old tech is cheap. Practice and mod and have fun. Sheat, Phenom II and FX so cheap, Dry Ice pretty cheap...
Ya. Youll need to overclock. Us older gents wont be around forever......

 

[newtekie1]

Guys, I just checked the pump... So, its a 3-pin, i have connected it to the AIO Pump connector. In BIOS the "AIO pump speed" is reported as 1760-1778 RPM and the AIO pump control disabled at Q-Fan Control, so, in my understanding, if nothing controls the pump, it must be running at it's full speed, which is advertised as 3000 RPM.
I also tried DC Mode (for 3-pin connectors) and selected the low-mid-high (30-40-70) temp limits and added 100% percentage to every limit, just to check if something would change.
After entering BIOS again, the speed is still reported around 1778 RPM.
I am not sure where to check AIO pump speed in windows.

An update:
The same speed is reported on HWMonitor as Chassis #6 speed. And it doesn't increase while doing a benchmark., just staying at those 1778 RPM's...

It is probably reading the pump at half speed. The pump on the Thermaltake Water 3.0 ARGB is specced to run at 3600RPM, 1778 is pretty close to half of that. Pumps read weird, sometimes they read as double speed, sometime at half speed.

 

[Okys]

Anyways, I found why is my pump speed reported only half the speed. It's something related to dual pulse and motherboard can read only one, so you should double the reading.
Tomorrow i will start lowering the voltage manually. At the moment, I am on adaptive+, HWMonitor reports voltage up to 1.3-1.32v, i have lost 200 pts of Cinebench R20 benchmark, because of adaptive voltage. As soon as i set it to Auto, temps increase by 5 degrees and bench is back to 5210 pts.
I have never had worst experience with things that should be working straight out of the box...

What fans do you have on the front of the case??

Those that came with the case - Silent wings 2 (3pin) 140mm fans (non-pwm).