[SOLVED] i9-9900K + Cooler Master ML240R AIO - expected performance?

[b4rtpl]

Hi Everyone,

First of all, allow me to say hello, as this is my first post on this forum. I've been visiting Tom's Hardware for quite some time now as an anonymous user and I have always found the advice here very helpful. Hopefully, now that I'm "official", I would be able to share some of my experience as well.

Back to the topic at hand, I have recently built a PC featuring the i9-9900K and the Cooler Master ML240R AIO. When I first set the PC up, I ran some tests to see if everything was working properly. Unfortunately, I discovered that the cooler was offering quite poor performance in terms of thermals:

• Idle temperature was ~30-35C (which I suppose is ok) with occasional spikes to sometimes over 40C
• When running Prime95 (small FFTs, AVX off) the CPU temperature immediately jumped to ~80-83C for the duration of the Turbo. After the CPU got power throttled, the temperature would drop to ~60C and stabilize at ~68-69*C with CPU Package Power at ~95W
• When running them same Prime95 test with AVX on, behavior was similar, although the temperature on Turbo would immediately increase to over 90*C

Considering that the cooler seemed to work properly after power throttling, I decided to lift the CPU Power Management limitations in BIOS and see how it behaves:

• Idle temperature was similar
• When running Prime95 (small FFT, AVX off) the CPU temperature jumped to ~80-83C and slowly climbed up, as if the cooler didn't manage to distribute the heat created on the CPU. The temperatures would climb to over 90C where I would stop the test. CPU Package Power was ~140-150W
• Running Prime95 with AVX was impossible - the temperatures would start at ~90*C so I stopped the test after a short moment

As the cooler was also making weird noises (gurgling + metallic), I contacted Cooler Master and they told me to send it back. A week later I received a new unit, only to discover that the performance is ... exactly the same.
Since this is my first experience with AIO cooling, here's my question: is this normal behavior for this type of cooler? In their response, Cooler Master said that this model should be able to handle CPUs up to 250W - I can hardly see that happen, considering that it doesn't manage to cool the CPU at 140W already (assuming these are the same Watts we are talking about...).
Here are some more details based on questions I've seen in similar threads:

• I have a pull configuration on the radiator fans, which pull air into the case through the radiator from the front. There is a single 120mm fan at the back of the case that acts as an exhaust (I'm planning to add one more at the top)
• My motherboard is an ASUS MAXIMUS XI Hero. I'm using XMP I and answered "No" on the prompt to use AI OC, so settings are stock (MCE is off)
• When touching the hoses on the cooler, I don't really feel a difference in temperature (one of them should be hotter as far as I learned). The radiator itself also does not seem to be hot, but I admit I haven't touched the fins, only the edges
• I applied fresh thermal paste when replacing the cooler - pretty sure it's ok
• The radiator is mounted with tubes facing up (the only way I could fit it) - could that cause the pump to suck in air bubbles?
• Using HWiNFO64 for measurements

I would appreciate if you commented your thoughts on this case. I'm wondering if it's possible that I received a faulty cooler twice in a row, or is it that I'm expecting too much from a 240mm AIO and should only run strictly stock settings?
Thanks in advance!

 

[Phaaze88]

I have a few ideas, so I'll give it a shot.

There's no need to use AVX on P95 small FFT. The load that puts on cpu's is too high - though it makes for a great space heater in the winter!
Use P95 small FFT, AVX off for cooler thermal testing for about 1 hour.
Use Cinebench R20 'infinite loop' and Asus Realbench(8hr, 1/2 ram) for testing voltage stability of any overclocks.

When running Prime95 (small FFTs, AVX off) the CPU temperature immediately jumped to ~80-83C for the duration of the Turbo. After the CPU got power throttled, the temperature would drop to ~60C and stabilize at ~68-69*C with CPU Package Power at ~95W

There's a wild card with these results, and it's whatever the motherboard settings were. The auto settings use more voltage than necessary, by default, to guarantee stable operation.

The radiator is mounted with tubes facing up (the only way I could fit it) - could that cause the pump to suck in air bubbles?

Yeah. Those units are not topped off with fluid, so there is air present. It will 'hang around' the highest point in the loop.
The best front mounted orientation is with the tubes entering from the bottom of the rad; the pump will push the air in the loop into the radiator, where it will then rise to the other end and get trapped there, as opposed to the opposite, where it can go back and forth between rad and pump.
Is mounting the rad at the top an option?


Also does your chassis allow for adequate airflow through the radiator fins? Liquid coolers are basically hybrid air coolers.
If chassis air flow is poor, then cooling will be more dependent on the flow rate of the liquid - without raising rad fan speeds to try and force more air through, at least.
Is the header that you have the pump plugged into set to PWM mode instead of auto in the bios?

In their response, Cooler Master said that this model should be able to handle CPUs up to 250W...

Up to, but not exactly 250w. The closer to the cooler's capacity you get, the closer it is to thermal throttling. Pulling 250w on a 250w cooler leads to a cpu pretty much running over 90C.
The 9900K easily runs over 200w on P95 small FFT, AVX off, if you remove the power limits. With that said, a 240mm unit isn't adequate for stress testing this cpu without power limits.

 

[b4rtpl]

Thanks a lot for replying and sharing the tips. It's good to know that AVX is not necessary with small FFT - I don't worry about this one in that case and will follow your recommendation for testing.
Answering the points above:

There's a wild card with these results, and it's whatever the motherboard settings were. The auto settings use more voltage than necessary, by default, to guarantee stable operation.

Yes, this is something I noticed when looking at the voltage applied across different cores. Sometimes it's very high, reaching even more than 1.4V. Would you recommend setting a fixed, stable voltage to the CPU rather than allowing it to take as much as it needs?
Other than the automatic voltage and Multi-Core Enhancement being off, the settings are pretty much stock. I have actually seen a pretty interesting video from Gamers Nexus lately, showing that the MAXIMUS XI Hero is one of the few motherboards that allows not to violate Intel's TDP specification:

View: https://www.youtube.com/watch?v=IBrumDWpl-c

Is mounting the rad at the top an option?


Also does your chassis allow for adequate airflow through the radiator fins? Liquid coolers are basically hybrid air coolers.
If chassis air flow is poor, then cooling will be more dependent on the flow rate of the liquid - without raising rad fan speeds to try and force more air through, at least.
Is the header that you have the pump plugged into set to PWM mode instead of auto in the bios?

Please find below a picture of how it looks like right now. Unfortunately, mounting the radiator at the top is not an option at the moment, the case (Corsair Spec-05) is too small and the cooler won't fit. In general, you can probably tell from the picture that things are pretty packed inside. The GPU can also reach temperatures of up to 80C. I was thinking of upgrading the case to something bigger to allow more space for the components - do you think that would be a good solution? I've only done basic research until now, but are there any particular brands/models you would recommend based on your experience?
The AIO Pump is plugged into a dedicated header on the motherboard and it's operating at ~2400RPM all the time. The fans are plugged into the CPU Fan header and they have a standard curve set up in BIOS. I cannot set a screenshot right now, but I can try getting one if you'd like. Based on the current noise level I think I should be able to move the curve a bit up, to have more RPM at temperatures below 70C.

https://imgur.com/a/K3fYPR6

View: https://imgur.com/a/K3fYPR6

 

[Phaaze88]

1)Yeah, 1.4vcore is definitely way more than needed. Start at 1.32 manual vcore and make your way down by 0.01v until it's no longer stable - crashes, then go back up to the last stable vcore.

2)Yep. The mobo models that don't adhere to Intel's TDP guidelines are just the vendors competing against one another.
Those adjustments, either base clock OC, or simply removing the TDP power limits, can screw the unawares end user over though.
Base clock overclocking, OCs core frequency, memory frequency, cache/uncore frequency.
The problem with this is that some cpus don't react too well to base clock OCs, and the end user could spend days or months trying to troubleshoot crashes caused by this and never figure it out...
Removing the power limits is self-explanatory.

3)Thanks for the image, that's helps.
Corsair Spec 05, too... yeah, I see a potential case airflow issue with that. The rad fans are definitely going to have to work harder to get air through the rad and the tight spacing in the front panel.
You could max out the rad fans, but you may not - well, not everyone is - be comfortable with the extra noise created by that.

Oh, and the radiator tubing, that looks really tight, especially the 'bottom' tube. It doesn't seem to have any slack at all compared to the 'top' one.
The tube could come loose from the barb if it's too tight, and you'll have a leak on your hands.

4)The gpu temps are also from the lack of airflow. It would be getting most of it's air from the rear of the chassis in the current situation.

5)I could recommend a new chassis if you don't care for the current one.
Is there a particular style you prefer?

 

[b4rtpl]

1)Yeah, 1.4vcore is definitely way more than needed. Start at 1.32 manual vcore and make your way down by 0.01v until it's no longer stable - crashes, then go back up to the last stable vcore.

Ok, thanks - I'll check it out. I made some attempts at setting a stable vcore a couple of weeks ago, but for some reason the motherboard ignored the setting ... will investigate how to overcome that.

Oh, and the radiator tubing, that looks really tight, especially the 'bottom' tube. It doesn't seem to have any slack at all compared to the 'top' one.
The tube could come loose from the barb if it's too tight, and you'll have a leak on your hands.

Yes, it was pretty stiff when I first installed it. I could probably rotate the pump to the right, so that the tubes are not that tense.

4)The gpu temps are also from the lack of airflow. It would be getting most of it's air from the rear of the chassis in the current situation.

5)I could recommend a new chassis if you don't care for the current one.
Is there a particular style you prefer?

Every time I set up a PC in the past, I just picked a reasonably priced case that was acceptable in terms of aesthetics. I suppose it worked, since I was always using stock coolers and settings. It seems the world has gone one step further
I don't mind changing the case at all if that means better comfort and airfrlow for all the other components. I did some research today and found the be quiet! Dark Base 700 to be quite interesting. It's design is elegant (not too futuristic or filled with RGB) and it has the side window. I also like the dedicated radiator shelf at the top, although I know this might be a gimmick. The case is a little pricey though...
My main concern with other cases that I looked at is that they will not fit the radiator above the mainboard and RAM. I measured that I would need +6cm in height vs the Corsair Spec-05 to fit the combined width of the radiator and the fans.

 

[Phaaze88]

I don't mind changing the case at all if that means better comfort and airfrlow for all the other components. I did some research today and found the be quiet! Dark Base 700 to be quite interesting. It's design is elegant (not too futuristic or filled with RGB) and it has the side window. I also like the dedicated radiator shelf at the top, although I know this might be a gimmick. The case is a little pricey though...

This case actually isn't that bad, save for the price, and the poor thermals on Silent Mode Level 1, but, you can always adjust fan speeds yourself.
https://www.gamersnexus.net/hwreviews/3122-be-quiet-dark-base-700-case-review
It's got competition with the likes of NZXT's H710(i) Phanteks Eclipse P600S, and even their own Pure Base 500 DX that recently came out. Perhaps one of those will catch your interest without pushing the bank?

My main concern with other cases that I looked at is that they will not fit the radiator above the mainboard and RAM. I measured that I would need +6cm in height vs the Corsair Spec-05 to fit the combined width of the radiator and the fans.

It shouldn't be too much of an issue with rads at 30mm or below.

 

[b4rtpl]

Perfect, thanks a lot for your suggestions - I will then pick a new case and hopefully move all the components later during the week. Once done, I'll share the new thermals with you in this thread.
Thank you and keeping my fingers crossed!

 

[Phaaze88]

🤞 Good luck!

 

[b4rtpl]

Thanks for your patience! Update from my side:

I ordered the new case and re-fitted all the components during the weekend. After long consideration I went with the NZXT H710 as it seemed to offer slightly better airflow than the Dark Base 700. Observations so far:

• The case is bigger than the previous one I had and it was much more convenient to build in thanks to all the cable management options available.
• The overall material quality is quite good. The only 2 things that could be better are:
  • The front audio connection - there is only one jack at the front and microphone sound is quite bad in my case when using the attached splitter
  • The back exhaust fan is 3-pin (so no PWM) while the three front fans are 4-pin and can be fully controlled by the motherboard. Under load the front fans spin faster than the back one - I assume this is not optimal, since air gets in faster that it can be released.
• With additional space and the 4 fans that came inside the case, I feel the airflow is much better and the PC has more "breathing room" inside
• During gaming sessions the temperature of the GPU dropped from max ~80C to max <70C (I have an RTX 2080S). The CPU reaches max <80C during short bursts of Turbo, but keeps ~60C under normal load

Overall, I feel changing the case was a good decision and will improve the longevity of the PC.

However...

...when testing thermals in Prime95 Small FFT (AVX off), the CPU still jumped to >80C and climbed steadily until power throttling kicked in. I made an attempt at setting a stable Vcore, applying the following settings:

1. Manual Voltage = 1.32V
2. SVID = Disabled
3. BCLK Adaptive Voltage = Disabled
4. LLC = Level 5

Under 100% load HWiNFO showed a Package Power of 2W (I read that's because of SVID being disabled) and the thermal results were actually worse than on stock settings. I went back to BIOS and enabled settings 2 and 3 + set LLC to Auto. Under load with P95 the CPU would not surpass a Vcore of 1.296V, but the thermals would still be higher than on stock settings, which made me set them back to default.
Did I do the whole thing wrong (e.g. missed some settings or should have tried with lower Vcore) or is it just that a 240mm AIO is not enough for the i9-9900K?
Thanks in advance for your input

 

[Phaaze88]

That's a little too much manual voltage.

	SSE Frequency	AVX2 Frequency	Vcore	% Capable
9900K	4.80GHz	4.60GHz	1.275V	100%
9900K	4.90GHz	4.70GHz	1.287V	Top 91%
9900K	5.00GHz	4.80GHz	1.300V	Top 30%
9900K	5.10GHz	4.90GHz	1.312V	Top 5%

That's Silicon Lottery's binning statistics for the 9900K.
You should be able to run with less manual voltage if you also use an AVX offset.

Leave SVID disabled if you want to overclock. If you exceed a specific amount of current, it will trip the Voltage Regulator's safeguard and shut the PC off.
BCLK: same deal as SVID. Leave it disabled if overclocking.

Already mentioned before:
Use P95 small FFT, AVX off for cooler thermal testing for about 1 hour.
Use Cinebench R20 'infinite loop' and Asus Realbench(8hr, 1/2 ram) for testing voltage stability of any overclocks.

I also already mentioned this in the last part of my first post:
The ML240R covers up to, but not exactly 250w. The closer to the cooler's capacity you get, the closer it is to thermal throttling. Pulling 250w on a 250w cooler leads to a cpu pretty much running over 90C.
The 9900K easily runs over 200w on P95 small FFT, AVX off, if you remove the power limits. With that said, a 240mm unit isn't adequate for stress testing this cpu without power limits.

The current cooler is enough for the 9900K at stock operation. There's no headroom for overclocking.

 

[b4rtpl]

Ok, thanks, all clear then - I will play with the settings to find the optimum, let's see how low I can get with the voltage. Probably will not overclock at this point, as I'm satisfied with current performance but might re-visit in the future.
Thanks a lot for your help!

 

[Karadjgne]

A 240mm AIO is roughly 250w capable. The i9 9900k generally runs @ 200w at stock boosts and standard Intel turbo core settings. Applying an OC or locking cores can change that, just like enabling MCE does. That cpu is capable of 250w or higher.

With coolers the way the temp works is a slight rise in temps as thermal load increases, upto a point. At that point, the curve starts a sharper curve upwards on a logarithmic scale, not linear until temps become thermal wattage gain squared.

Literally the closer to saturation you get, the faster the temps go up.

Under normal usage, even heavy gaming, that AIO is plenty, but p95 is not anything remotely close to gaming, it's a maximum 100% load on all cores, all threads.

LLC is a pre-emptive voltage addition. It's purpose is to add voltage to vdroop to maintain voltage demand levels. At stock settings that's pretty minimal, even on auto. But that added voltage also applies to the top of the droop as well as the bottom, so using extremes of LLC is counter productive and can create instability. Asus has 1-8 as a setting? 5 would be medium, which is perfect for a decent OC, but more towards extreme without the cpu voltage demands of closer to stock speeds.


Honestly, with as strong as a 9900k is, and with as little gains as OC gets on that cpu, leaving it stock is plenty good. My i7 went from 3.9GHz to 4.9GHz, a 1GHz OC showed plenty of results, but moving from 4.9GHz to 5.0GHz on a 9900k is not really worth the effort.