[SOLVED] Is there a solution to my system overheating that does not require new parts?

[sadakatsu]

I have been trying to resolve a series of system issues due to my old parts not working well anymore. Through a series of what now seem to be insufficiently researched purchases, I now have the following set of parts:

CPU: Intel Core i9 9900K
Motherboard: MSI z370 Gaming Plus (MS-7B61, UEFI version 1.A0; part was original to system)
Cooling: Corsair iCUE H150i ELITE CAPELLIX

Based what I have read now that I am trying to resolve problems, I guess it will not surprise any readers that I have overheating issues. It takes running my three radiator fans in the 70%-100% range to keep my idle system around 30 C. When I try to run the Prime95 stress test according to https://forums.tomshardware.com/faq/cpu-overclocking-guide-and-tutorial-for-beginners.3347428/ , individual cores spike at 80+C within the fifteen minutes as the whole package temperature slowly creeps toward that point. I have also noticed in HWMonitor that the CPU VCore value jumps eratically between 1.1V and 1.472V all while maintaining roughly 5GHz on each core despite the 100% load (though it does not seem to keep there constantly; it ranges between 4998 and 5006 MHz). I have also noticed that my motherboard's temperature sensors labeled CPUTIN and TMPIN5 can hit 100 C within those 15 minutes; TMPIN3 breaks 80 C.

I am fairly out of my depth at this point. I have been building systems since 2008 for myself and friends, and these problems are new to me. As best as I can tell, my only hope for a solution that doesn't require some combination of returning parts or buying new ones is to undervolt my processor. I tried following the steps from the above link and others to try to figure out which UEFI settings might lead to a lower voltage that supports hopefully close to the same speeds. My results have been confusing. The very first step of that guide after the memtest (setting the CPU VCore to the current setting - 1.066V at that point - and the CPU multiplier to the boost multiplier - 100 MHz x 50 for my CPU) resulted in my motherboard running contentedly at 3.6 GHz in defiance of my configurations. My second attempt to configure similar settings using information from https://www.msi.com/blog/intel-9th-cpu-overclocking-5ghz-with-z390-motherboards in addition to the previous link would not boot.

Does anyone know how I can find a solution to my CPU and motherboard trying to boil water that does not require me to replace parts yet again?

If I must return and replace parts, is there a way I can replace just the AIO or the motherboard to get good performance out of my CPU?

 

[Darkbreeze]

Where is the radiator mounted? Are the fans on the radiator configured to bring air in from outside or to blow air out from inside, ie, intake or exhaust? Are the fans mounted on the side of the radiator that is closest to the outside or are they mounted on the inside surface of the radiator?

HWmonitor sucks, stop using it. Use HWinfo. The readings MIGHT still be the same, but often, they are not. Especially for things like the various minor motherboard sensors. When using HWinfo, choose the "sensors only" option.

What case do you have? What OTHER fans are installed, where are they installed, and HOW are THOSE fans configured, as intake (Bringing air INTO the case) or exhaust (Blowing OUT of the case)?

 

[sadakatsu]

Case: https://www.corsair.com/us/en/Categ...40-High-Airflow-ATX-Cube-Case/p/CC-9011030-WW
Additional Fans: https://www.corsair.com/us/en/Categ...SP-Series-RGB-PRO-Performance/p/CO-9050094-WW

The radiator is mounted at the front of the case, tubes going up, with the fans set as intake along the inside.

The other fans are installed as exhaust along the top and at the back of the case.

From what I have learned from sources such as

View: https://www.youtube.com/watch?v=BbGomv195sk
, I know it is much better to install the radiator tubes down. I simply cannot until I find a vertical GPU mount kit either in a store or from an online seller that doesn't look like a scam. I would also prefer to have the intake fans installed against the case to get fresh air from as close to outside the case as possible; I originally did install them that way. Unfortunately, the tubes did not bend that way, so I could not fit the top fans inside the case. This led to the arrangement I hope you can see in the picture.

I was running HWiNFO64 at the same time as I was running the Prime95 stress test. Most of the values you can see in the Maximum column come from that:

 

[USAFRet]

OK....ANY airspace in that rad/tube config WILL be right where the tubes are at the top.

"I know it is much better to install the radiator tubes down. "
This is absolutely 100% true.

 

[sadakatsu]

I think it's better up at the top of the reservoir than in the pump (though I take this on other's authority as opposed to knowing anything myself). I've heard no gurgling. The AIO has managed to cool the CPU as much as 20 C in a second after I stop the stress test. The AIO does seem to be doing SOMETHING... just not enough.

The AIO also can't do anything about the VRMs which I have seen described on various sites using expletives.

 

[USAFRet]

It is absolutely doing "something".

But, you're here due to an issue.
Fix the known things first.

 

[geofelt]

Water boils at 100c. You are not there yet.
Your monitor shows a maximum of 84c under a stress test. That looks ok to me.
And curiously enough, 100c. is also the temperature that the cpu will throttle or shut down to protect itself.

Out of curiosity, try leaving the case side open and direct a house fan at the innards. If you do significantly better, look for a better cooling case.
You could see if removing the front air filter helps.

 

[sadakatsu]

But, you're here due to an issue.
Fix the known things first.

I'm confused. Does this mean that mean I need to flip my radiator before I can expect any solution to work? The chipset heat sink, the SATA ports, the GPU's length, and the tube lengths mean I cannot route the radiator tubes under, over, or around the GPU. I am looking for a vertical GPU mount kit, but I do not know when I can get one. I would like to try what I can that is not blocked.

Out of curiosity, try leaving the case side open and direct a house fan at the innards. If you do significantly better, look for a better cooling case.
You could see if removing the front air filter helps.

Okay, I will set up as open a case as I can, point a box fan in the case, and try Prime95 again. I'll post a full report if I can figure out how.

 

[sadakatsu]

I had to stop the test in just under 11 minutes. My CPU package temperature hit 95 C. One of my motherboard temperature sensors (MOS?) broke 100 C and stayed up there until I stopped the test.

It seems that blowing additional air into the case did not help, even with all the grills and filters removed.

I could not find a way to generate a (non-exhaustive) log, so here are some HWiNFO64 screenshots:

 

[punkncat]

Sorry that I didn't dig through the various replies. Based on your OP alone, if you are topping mid 80's on an AIO of that caliber while running 5ghz on all cores I don't see what the issue is. 9900(K) are known to be hot. They use a lot of power and are 14nm. I personally feel like you are good.

 

[Darkbreeze]

I'm confused. Does this mean that mean I need to flip my radiator before I can expect any solution to work?

You've already watched the premier video on this subject, and you've seen what kind of problems they clearly showed will happen with a setup like yours and straight up says in the first minute of the video do not front mount your radiator with the hoses at the top or you WILL have problems with the cooler acting as though it has coolant loss/air lock, etc. , so yes, you pretty much need to do that and you need to hope that running it the way it is now hasn't caused any irreparable damage. Usually, flipping the radiator over immediately solves the problem, but not always.

Sorry that I didn't dig through the various replies.

Generally, if you can't be bothered to read through the whole thread, it's recommended (And preferred) that you don't bother injecting an uninformed reply in that case. That sounds worse than I mean for it to, but you have been here long enough that you know what I mean.

Also, in the OP they clearly state that INDIVIDUAL cores are hitting above 80°C and that the package is "creeping towards that" as well. I think we can assume that means they are not allowing it to reach that and are ending the test routines, which to me says that if they allowed it to continue running it WOULD exceed the recommended envelope. Honestly, for the 9900k I'd probably ultimately give it the allowance of 85°C while running the Small FFT test and you need to also be sure that you are not running ANY version of the AVX instructions while you are running small FFT.

 

[Karadjgne]

Liquid cooling requires a change in thought direction. The cooler does not 'cool' the cpu, the block transfers wattage to the coolant, which drops it off at the radiator. The cpu temp is what's left over from the work process.

The colder the coolant, the greater its ability to absorb wattage. Meaning when you start the test, the coolant will be very close to room ambient temp. With all the wattage absorbed into the coolant, it's going to transfer some of that heat to the block, the tubing, the rad itself which eventually will raise the temp of the coolant. It'll continue to do so at a relatively slow rate of climb until it reaches load equilibrium, which generally takes around 30 minutes or so.

So while under stress testing, you'll start at a lower coolant/cpu temp and with time, that'll slowly creep higher, until it reaches equilibrium and stays stable.

Stress test doesn't apply here, you should be running Prime95 small fft for temp with all 3 AVX technologies disabled, stress tests are different as they use different methodology to trip the cpu, meaning core usages change, voltages change etc. P95 small fft applies a constant 100% all core load, not a variable load mixing gpu, ram, cpu etc.

Z370 mid-range mobo's like the Gaming Plus were never designed to handle a 9900k all core 5GHz. Stock values only. It's why the VRM's are going nuts, nobody expected a 250w monster cpu to get stuck in a mid-range mobo. That's a ton of current at a constant rate with a 5GHz all core requirement. Changing airflow isn't going to do much of anything for the VRM's, only dropping the current will.

 

[sadakatsu]

It seems that the basic consensus is that I need to buy more new parts.

• I need to either install my radiator with the tubes down or along the top of my case. The only vertical mount kit I have found in-store is the https://www.microcenter.com/product/510585/Air_Vertical_Kit . It is apparently intended for two specific Lian Li cases, cannot be installed in my current case (I have HDDs where I am supposed to screw it in), and costs as much as some cases. So, I am going to try to find a case in-store that either comes with a vertical mount kit integrated or can fit and clear my 360mm radiator along the top of the case.
• I cannot expect my z370 motherboard to support my CPU without tripping the thermal alarms. I had hoped for advice on how I could undervolt my CPU while stably maintaining respectable clock speeds (maybe 4.6 GHz, like my old CPU). The closest I have gotten to such advice is the comment, "Changing airflow isn't going to do much of anything for the VRM's, only dropping the current will."

So, I am going to go to a Micro Center later today. According to their website, they have the following z390 chipset motherboards in stock:

• https://www.microcenter.com/product/512012/asus-z390-a-prime-intel-lga-1151-atx-motherboard
• https://www.microcenter.com/product...gaming-edge-ac-intel-lga-1151-atx-motherboard
• https://www.microcenter.com/product/512750/msi-z390-mpg-gaming-plus-intel-lga-1151-atx-motherboard
• https://www.microcenter.com/product/512631/gigabyte-z390-ud-intel-lga-1151-atx-motherboard

MicroCenter reports Core i9 support in the CPU SUPPORT section only for the Gigabyte motherboard. I cannot tell if that means it is the only board I should consider.

Since my ultimate goal is to use this system for both gaming and running CPU and GPU-intensive AIs, can any of these motherboards withstand the 9900K running at high or max boosts for hours at a time with fairly high loads? I don't know what sort of thermal response I should expect during a fifteen-minute test, but I would prefer that my motherboard not ever break 60 C.

 

[Karadjgne]

You don't need a vertical mount kit, that's intended for the gpu to be mounted vertically. The radiator with mount anywhere where a 360mm fans will.

The aio is a decent performer, but that whole thing with tubing lengths can be easily solved by use of a different aio such as the Fractal Design S36 which uses standard custom loop G1/4 fittings, you can make your own tubing lengths as needed.

You could also use a custom loop as is, using a copper radiator, 3.2 ddc pump and small reservoir, which increases cooling capacity without increasing size, although you could opt for a thicker rad like a 45mm version.

You do have options.

 

[geofelt]

1. I do think you need a more robust Z390 based motherboard to handle a i9-9900K.
Try to find reviews of each of the 4 candidate motherboards and pay attention to the vrm cooling and how they do with a 9900K overclock.. As usual, you are likely to find that the best units are more expensive. My thought is that all 4 are reasonably priced.

2. I really like the case. It is an excellent one for air cooling.
You need to do something about your 360 aio or it's mounting.
It is not now effective.
Consider one of the noctua NH-D15 coolers.
The twin 135mm radiator towers have about the same amount of radiator size as a 360 aio with 3 120mm radiators. Cooling potential should be similar.
The main difference is in where the heat exchange takes place.
As mounted now, the heated cpu air is flowing over the motherboard reducing any vrm cooling.
With air, that no longer becomes an issue.
The noctua web site rates it suitable for maximum turbo and overclocking .
https://ncc.noctua.at/cpus/model/Intel-Core-i9-9900K-307
Two front 140mm fans should provide all of the cooling air you might want.
If needed, such fans can be changed out to higher airflow versions.

 

[Karadjgne]

Noctua NH-D15 is rated at @250w±, a 360mm AIO is closer to 300-350w. It's more than just area, it's fin density, fin composition, heatpipes etc. Aircoolers rely totally on the small area where the heat pipe passes through the fin for thermal transfer, aios have much larger channels, more of them, so each individual fin gets better saturation and spread of wattage. In a word, a liquid cooling rad is more efficient than a heatsink, allowing for greater capacity per Sq.inch.

Its why the 120mm rads equal most 120mm budget coolers, even though they are only 27-30mm thick vs the 40+mm of an aircooling heatsink.

The D15 can handle a 9900k at stock values easily, especially with most games not using the full core capability of the cpu. Pushing a 5GHz OC is pushing it a little, you'll see some games knocking in the 90's range. An all core 5GHz temp test like P95 is seriously stretching your luck, depending on not only the vcore, but also the current needed to get the wattage, silicon lottery.

 

[geofelt]

Just sat through the 26 minute video on aio mounting.
I learned a lot. A sobering discussion.
If anything, it reinforced my personal bias against using a aio cooler if not necessary.
It bothers me that a cooler should deteriorate over time.

One possibility:
On the replacement motherboard, there will usually be a couple of slots available to take a graphics card.
Normally, the top slot nearest the cpu will be a X16 slot and be the preferred place to mount a graphics card.

But, a lower slot can run at X8 which still gives good graphics performance.
Tests I have seen show a minimal difference in graphics performance except for the very highest powered cards.
It would be worth the experiment to put your graphics card in a lower slot if that would let you mount your current aio in the best orientation with the tubes on the bottom.

If you absolutely can not mount the 360 aio properly, then you should replace it.
If a 280 aio is the best you can do, then the noctua solution should at least be a contender.

Or, look into a new case.

 

[sadakatsu]

All of the motherboards I listed have reviews that note they cannot handle the 9900K. Fortunately, I found that another MicroCenter that is not too much farther has more options when you include open boxes. I have reserved an ASUS Z390 ROG Maximus XI Hero WiFi for pick-up. It appears to be a very solid choice, though one review noted they had to drop the boost speed to 4.9 GHz to get stable performance. But, hey, 16 threads at 4.9 GHz is still a lot better than my current meltdown or my previous 12 threads at 4.4 GHZ.*

There is a good chance I will return the Corsair H150i Elite Capellix. It depends upon what I find while shopping their cases.

* I got the Intel Core i9 9900K to replace my Intel Core i7 8700K after it suddenly started overheating.

 

[Darkbreeze]

CPUs don't generally overheat because something is wrong with the CPU. They overheat because something is wrong with the cooling configuration. I can probably count on one hand over the course of 30 years how many CPUs I've encountered that ACTUALLY had some kind of internal problem causing them to overheat. Three of them, out of hundreds and hundreds that I've either owned or were in systems that I worked on. Three. Ever.

Two of those were physically abused because the owners had dropped them onto the floor while doing something with them like changing paste etc. One of them simply developed a short or something inside. The rest, in every case where there was an overheating problem, were due to problems with bad CPU cooler mount, dried out paste (This generally doesn't happen overnight. It takes a couple of years to dry out unless it was not mounted correctly in the first place and air has been easily getting in between the heat spreader and bottom of the heatsink.), failed fans, misconfigured fans, fans facing the wrong way or in some cases a faulty motherboard that is reporting the wrong temperatures. Also, using the wrong monitoring software is sometimes to blame. I don't recommend using anything for thermal monitoring except for HWinfo, Core Temp and on AMD processors, for older models that use the thermal margin style of monitoring (FX processors for example. AM3, AM3+), AMD Overdrive, or for newer AMD processors, Ryzen master.

So if you were having thermal issues with your old CPU and are still having thermal issues with a newer CPU, there's a good chance the two problems could be related, and that your older CPU was fine.

 

[Karadjgne]

Growing up, mom drilled into my poor, tired brain 'homework first, then play'. That's most definitely true regarding building or maintaining a successful pc.

I've never seen a cpu overheat when the cooling is functioning correctly. Ever. I've seen cpus overheat because of faulty cooling (both air and liquid), I've seen plenty overheat because the cooling was insufficient for the job. Never seen a cpu 'suddenly' develop thermal issues non-cooling related.

 

[Darkbreeze]

I should clarify an exception to that. I HAVE seen four or five CPUs "suddenly develop thermal issues" when they are in systems that have been improperly or overly overclocked beyond or "right at" the safe level for that configuration (Meaning the CPU sample itself, the cooling configuration, allowance for "did the system get packed with dust and dirt or did something block off the airflow") if it was that way for a while and had degraded the CPU.

But we're talking CPUs that were MANY years old and had been overclocked or had other cooling issues that went unnoticed for an extended period of time, and again, you'll notice that part of that IS related to cooling issues. It doesn't have to be though. A CPU configured with TOO high of voltage, can definitely experience electromigration and VT-shift, and develop overheating problems or a need to further increase voltage in order to keep it stable which itself then causes heating problems.

But none of that should apply to THIS thread.

 

[sadakatsu]

I bought a new case (Fractal Design Meshify S2) and a new motherboard (ASUS ROG Maximus XI Hero WiFi ). I now have my Corsair H150i Elite Capellix installed at the top with my fans pushing air up and out through them. The case had extra space, so I have one additional 140mm fan for intake and exhaust each. The new motherboard's UEFI has a page where it estimates the clock speed it can support for my processor; it notes that I can sustain a clock speed just over 4.8 GHz with non-AVX instructions.

When I ran Prime95, my computer ran stably with a clock speed just over 4.7 GHz with no core ever exceeding 69 C.

I really wanted a free solution if possible, but at this point, I am glad to have a working computer again.

 

[Karadjgne]

Anybody would be glad to have a working pc 😁, as opposed to a temperamental pc. You probably wouldn't have an issue with 4.9GHz as long as you didn't do full core burns like the p95. But 4.7GHz is good enough.

 

[Darkbreeze]

Sometimes, unfortunately, especially with high end configurations, is that there IS no good answer without having to spend some additional money. I think you did the right thing, for the most part. I think if you could get the radiator front mounted in an intake configuration AND have the radiator configured correctly with the hoses at the bottom, you'd probably see an additional 5-8°C Delta difference overall, which could afford you some additional headroom for your clocks, but considering where you are right now thermally that is certainly not seeming to be a necessity anyhow.

One thing to note however is, WHICH Prime95 test are you running? Small FFT, or something else? And, if you ARE running the Small FFT test, are you also disabling all of the AVX options so that you are running a steady state workload, or is that thermal value/result being obtained WITH AVX instructions of any kind, enabled? If it is without AVX and you are running Small FFT, then that is a very acceptable result and I think you should be pretty happy with that in contrast to what you were getting before.