[SOLVED] Cooling for an i-9 9900k?

JTidfk

Commendable
Mar 12, 2020
33
1
1,535
So i recently upgraded cpu's, and am now realizing i'm definitely going to need to upgrade cooling.
I only have room for a single radiator liquid cooler in my case, and i'm really between the Corsair Hydro Series H60 Liquid CPU cooler 120mm, and the NZXT Kraken M22 120mm.

Anyone have any opinions or thoughts on which one would be better? I'm leaning more towards the corsair from what i've heard about it.
 
  • Like
Reactions: Setab
Solution
Measure the height available for an air cooler.
If you have 160mm available, you can install a twin tower air cooler like the noctua NH-D15s.
165mm for the NH-D15 and some other good units.
There are other good twin tower coolers around.
In effect twin towers and 140mm fans are about the equivalent of two 140mm aio radiators.
To do better, you would need a liquid cooler with a 360 radiator.
Such a cooler may be difficult to mount, depending on the case.

How good do you need to be?
How good CAN you be?
That will depend on the quality of your chip primarily.
One can do stress tests than use heat inducing instructions, but those are not what a user workload will use.
The best test will be YOUR workload.
But, if you can't handle a...
So i recently upgraded cpu's, and am now realizing i'm definitely going to need to upgrade cooling.
I only have room for a single radiator liquid cooler in my case, and i'm really between the Corsair Hydro Series H60 Liquid CPU cooler 120mm, and the NZXT Kraken M22 120mm.

Anyone have any opinions or thoughts on which one would be better? I'm leaning more towards the corsair from what i've heard about it.
What is your case so we can see how much space you have for an air cooler?
 

JTidfk

Commendable
Mar 12, 2020
33
1
1,535
What is your case so we can see how much space you have for an air cooler?
I honestly have no idea what my case is, i bought this pc pre built from cyber power years ago. I can't find my case anywhere. i have one fan on the top of my case, 3 in the front, and the cpu fan is in the back.
 

Phaaze88

Titan
Ambassador
Yet reviews show the Noctua having performance of a 280mm cooler.
Depends on the fan speeds, as the Noctua will outperform the silence oriented models. And that's where the 'silence freaks' get bitten in the butt.

Looks like I should have worded my statement a tad differently:
NH-D15 isn't enough either in a number of cases, and neither is a 'silent' 280mm. Performance 280mm if not overclocking - but then there's the issue of noise... 360mm or custom loop is best overall.
 

Phaaze88

Titan
Ambassador
@jeremyj_83
Check this out:
View: https://www.youtube.com/watch?v=23vjWtUpItk


Overclocked 9700K was the cpu in question, so around 200w of waste heat. What do you notice about the testing results?
-all 5 coolers were allowed to run at their max rpm.
-with the exception of the struggling H60, the cpu was still within the other 5s' ability to handle.
-the 360mm H150i Pro did worse than the 240mm H100i V2. Why? The fans on the latter are more powerful than the former.
-although the fans on the H100i V2 are stronger than that of the NH-U12A, the front-to-back orientation of the latter's fans prove to be more efficient at moving air than the former, which are essentially moving air through the rad by themselves.
-the H150i Pro was the only silence-focused cooler among the 5.

If there were an actual OC'd 9900K in there, the results would've been completely different.

So with that out of the way: Why does the NH-D15/S outperform some 280mm AIOs? The fans.
Silence-focused and noise-normalized fans will NOT out perform performance oriented fans that are allowed to run as they please. Fans are like the 2nd most important part of a cooler:
Heatsink/rad capacity > Fans > Pump speed > Thermal paste application > Thermal paste brand

It's important to know how much noise other users are comfortable with.
IF the OP is a 'silence freak', then a 360mm AIO, or custom loop, is really the only option.
IF the OP is ok with some noise, then a high performance 280mm AIO can make for a cheaper alternative over the above.
 
@jeremyj_83
Check this out:
View: https://www.youtube.com/watch?v=23vjWtUpItk


Overclocked 9700K was the cpu in question, so around 200w of waste heat. What do you notice about the testing results?
-all 5 coolers were allowed to run at their max rpm.
-with the exception of the struggling H60, the cpu was still within the other 5s' ability to handle.
-the 360mm H150i Pro did worse than the 240mm H100i V2. Why? The fans on the latter are more powerful than the former.
-although the fans on the H100i V2 are stronger than that of the NH-U12A, the front-to-back orientation of the latter's fans prove to be more efficient at moving air than the former, which are essentially moving air through the rad by themselves.
-the H150i Pro was the only silence-focused cooler among the 5.

If there were an actual OC'd 9900K in there, the results would've been completely different.

So with that out of the way: Why does the NH-D15/S outperform some 280mm AIOs? The fans.
Silence-focused and noise-normalized fans will NOT out perform performance oriented fans that are allowed to run as they please. Fans are like the 2nd most important part of a cooler:
Heatsink/rad capacity > Fans > Pump speed > Thermal paste application > Thermal paste brand

It's important to know how much noise other users are comfortable with.
IF the OP is a 'silence freak', then a 360mm AIO, or custom loop, is really the only option.
IF the OP is ok with some noise, then a high performance 280mm AIO can make for a cheaper alternative over the above.
That video just disproved EVERYTHING you were trying to state earlier. They were using the NH-U12A and it out performed the 5 AIOs. The only difference will be bursting performance where the AIO can come out on top due to thermal mass and transportation.

All that being said I have an issue with their testing methodology. From a laboratory perspective it has too many variables. What you really want is basically a hotplate that you can put programmable thermal loads onto and then measure the temperature above ambient. Put into these constraints then yes the Noctua NH-U15 doesn't perform as well as massive AIOs. However, the results are usually within a few degrees of each other, especially when acoustic performance is measured with it. https://www.anandtech.com/show/9415/top-tier-cpu-air-coolers-9way-roundup-review/11
With a 340W continuous load the NH-U15 achieved a 28.8 degree temp above ambient using full fans that produced 38.6 decibels at a distance of 1 meter. Noctua is supposed to be releasing an updated version of the 15S with 10% more heatsink surface area which will increase the cooling performance. They state that it will be able to dissipate 400W of heat. We already see that the current one can do just fine with 340W of heat.

Seeing that the NH-U15 can easily remove 340W of heat, we can say that this statement is false.
To play it safe when recommending coolers to others, it's 280mm if not overclocking, 360mm or custom loop for overclocking.
Another way of looking at the statement is I have a little Athlon 200GE, 35W TDP, according to your logic I need a $100+ AIO to cool it. I want to overclock said CPU you are stating I need $150+ AIO. See where your logic starts to go a bit nuts. That is a CPU that you can OC to the max with a 212 EVO twice over.

Your statement regarding temperature does make sense though. However, lets put some math behind this. With a static 340W load the NH-U15 was able to do 28.8 above ambient. You have a pretty average case and other components that aren't as thermally conscious. The ambient temp in your case can be 46C/114F and the NH-U15 can still keep the CPU at or below 75C/167F with a 340W continuous load and the noise it produces is in the quite/very quiet range. Say you want it almost inaudible with that 340W load. Exact same case conditions your CPU temp is now 83C/181F. A little toasty for my taste so I drop my continuous load to 250W, which is only seen on a 9900k @ 5GHz running Prime 95, and my CPU temp is now 73C/163F. https://www.tomshardware.com/reviews/intel-core-i9-9900k-9th-gen-cpu,5847-11.html

Can you see what I am getting at here. Sure the NH-U15 can be out performed by very large AIOs. But you don't need an AIO for running a 9900k, even massively overclocked.
 

Karadjgne

Titan
Ambassador
It's simple. The 95w TDP rating on the i9 9900k is bogus. TDP is calculated by averaging cpu voltage use over a series of middling apps. At base speeds only. No turbo, no hyperthreading.

Which isn't why you bought a i9 9900k. At stock values, running all core/all threads the i9 9900k is well capable of 200w thermal output. At a measly, but very popular, OC from base speeds to a solid all core 5.0GHz (as opposed to standard turbo of just 1 core at 5.0GHz), the i9 9900k can reach 250w thermal output.

The Corsair H60, or any other 120mm AIO, is rated at @ 140w. Inadequate is a seriously excessive understatement. To maintain decent temps, somewhere around 70°C ish, under full stress loads requires a cooler with at least 1.5x but more likely 2x the thermal output of the cpu. For the i9 9900k that means full custom loop. You can be somewhat successful with a 280/360mm AIO as they run @ 300w-350w, you'll see higher temps with a @250w 240mm AIO or the largest aircoolers like the NH-D15S or deepcool assassin III, but that'll still be temps of @ 85°C ish or more.

The rule of thumb on that i9 is 'go biggest', not 'go big', and even then that may not be enough.

Outside ambient temps have little to do with load temps, they affect mainly idle temps where the cpu output is far lower than the heatsink ability. Load temps are different in that it's more about the capacity of the cooler than efficiency.
 
At stock values, running all core/all threads the i9 9900k is well capable of 200w thermal output. At a measly, but very popular, OC from base speeds to a solid all core 5.0GHz (as opposed to standard turbo of just 1 core at 5.0GHz), the i9 9900k can reach 250w thermal output.
The only time you see a stock 9900k hit 200W power draw is doing Prime 95 with AXV, this is also the often quoted 250W @ 5.0GHz. https://www.tomshardware.com/reviews/intel-core-i9-9900k-9th-gen-cpu,5847-11.html Move forward a couple months and reviewers are changing how the test power. "We're moving away from using AVX-based stress tests for our CPU power testing, though we will continue to use them for their intended purpose of validating overclocks. AVX-based stress testing utilities essentially act as a power virus that fully saturates the processor in a way that it will rarely, if ever, be used by a real application. " https://www.tomshardware.com/reviews/ryzen-9-3900x-7-3700x-review,6214-3.html Now a stock 9900K pulls make 138W and the overclocked @ 5.0GHz does hit 213W on Handbrake. Handbrake makes heavy use of AVX which is why the power draw is so much higher, but still a good ways away from 250W. Note that Tomshardware used the MSI Z390 Godlike so there are no worries about VRMs holding back the CPU.

In the video that @Phaaze88 posted, to keep an all core 4.7GHz boost, required 150W running blender. That is quite easy to do with any top shelf cooler. In that video it shows all 3 cooling options posting very high temps, higher than anyone would ever expect. The only thing I can think is that they measured something wrong. Remember a lot of places are using the internal sensor for temperature readings. "Although the testing of a cooler appears to be a simple task, that could not be much further from the truth. Proper thermal testing cannot be performed with a cooler mounted on a single chip, for multiple reasons. Some of these reasons include the instability of the thermal load and the inability to fully control and or monitor it, as well as the inaccuracy of the chip-integrated sensors. " I know the 9900K requires a lot of cooling power, hence why I wouldn't say to someone to use a stock Intel cooler, but to think that one needs the industrial chiller from the 28 cores @ 5.0GHz demo is also crazy.
 

Karadjgne

Titan
Ambassador
I'll make it easy then. If a cpu pulls 150w on a 150w capacity cooler, expect temps of over 100°C. As coolers reach thermal limits, efficiency tanks, hard. If you want to keep a 9900k in happy temps under most normal conditions, it will take a minimum of a 250w+ cooler.

All core isn't a boost at 4.7GHz, it's only an overclock in which all cores are set to 4.7GHz instead of being allowed normal turbo numbers which would be 5.0GHz and down. Overall, 4.7GHz all core uses less wattage than normal turbo all core as you don't see 4.8GHz+ for core speeds.
 
I'll make it easy then. If a cpu pulls 150w on a 150w capacity cooler, expect temps of over 100°C. As coolers reach thermal limits, efficiency tanks, hard. If you want to keep a 9900k in happy temps under most normal conditions, it will take a minimum of a 250w+ cooler.

All core isn't a boost at 4.7GHz, it's only an overclock in which all cores are set to 4.7GHz instead of being allowed normal turbo numbers which would be 5.0GHz and down. Overall, 4.7GHz all core uses less wattage than normal turbo all core as you don't see 4.8GHz+ for core speeds.
I haven't disagreed that you need a massive cooler for the 9900k. I also haven't said that it only needs a 150W cooler to run well. I said "to keep an all core 4.7GHz boost, required 150W running blender. That is quite easy to do with any top shelf cooler." That means I wasn't talking about using a 212 EVO for cooling, that would be moronic to use on the 9900k, that means 280mm, 360mm, 420mm AIOs or the biggest air coolers you can buy. The NH-U15 is a 340W+ cooler, so that falls into the biggest air coolers you can buy category. Under almost all circumstances the NH-U15 should provide more than enough cooling for the 9900k. If I were living in the Sahara or Death Valley without AC, well that wouldn't suffice then.
 

Ancient_1

Distinguished
Oct 18, 2006
319
3
18,815
I will say the if the D15 will fit and he can't go bigger than a single 120 AIO the 15 is his best bet as long as he has enough ventilation in his case. My cpu I would guess is putting out as much or more heat than a stock 9900k and the D15 keeps my cpu cool enough for me. I do have a case the is about the best for keeping the cpu as cool as possible (SS Fortres FT01) feeding fresh outside air to the cpu/HS and doesn't do very well for the video card but does a decent job on the 7 mechanical hard drives and cpu.

Here is a shot of the time since I rebooted last nite before I ran two instances of HandBrake converting 10hr of edited HDTV mgp2 to x265 HEVC mp4 for archive on my plex server. It shows the fan speed of an AIO but it actually is powering my main intake fan. Also My ambient temp was on the cool side at about 65f.

i-HsGMSHn-X3.png
 
Measure the height available for an air cooler.
If you have 160mm available, you can install a twin tower air cooler like the noctua NH-D15s.
165mm for the NH-D15 and some other good units.
There are other good twin tower coolers around.
In effect twin towers and 140mm fans are about the equivalent of two 140mm aio radiators.
To do better, you would need a liquid cooler with a 360 radiator.
Such a cooler may be difficult to mount, depending on the case.

How good do you need to be?
How good CAN you be?
That will depend on the quality of your chip primarily.
One can do stress tests than use heat inducing instructions, but those are not what a user workload will use.
The best test will be YOUR workload.
But, if you can't handle a abnormal shutdown, you need to stress test.
I might suggest OCCT, as a realistic workload, but there may be others that work better for you.

Another alternative would be to run the intel performance maximizer app.
https://www.tomshardware.com/reviews/intel-cpu-auto-overclock-performance-maximizer,6179.html
It measures your chip quality and your cooling capacity and tries to do the best it can with your actual setup.
 
Solution

Phaaze88

Titan
Ambassador
I mean, who gets a 9900K and doesn't want to, or isn't tempted to do so at some point in time? That alone eliminates air cooling.

That means I wasn't talking about using a 212 EVO for cooling, that would be moronic to use on the 9900k
Well, do I have NEWS for you! Check out this person's signature: https://forums.tomshardware.com/members/sydneyblue203.1560910/
They claim to be doing just fine with it... I'd bet they didn't bother to thermally test it for the worst case scenario, of which Blender IS a realistic workload, mind you.

@Ancient_1
Believe it or not, your 3950X is at times more power efficient, and runs cooler than the 9900K - perhaps not by much, but still... and that gap only gets worse when the latter is overclocked.
 
I mean, who gets a 9900K and doesn't want to, or isn't tempted to do so at some point in time? That alone eliminates air cooling.
True. I tried overclocking my 4770K but after a while I removed the OC. I personally would still try to OC a 9900k with a NH-U15. If stability or temps were bad I would either take the OC off or go with a 360mm AIO.
Well, do I have NEWS for you! Check out this person's signature: https://forums.tomshardware.com/members/sydneyblue203.1560910/
They claim to be doing just fine with it... I'd bet they didn't bother to thermally test it for the worst case scenario, of which Blender IS a realistic workload, mind you.
Just when you think you have seen everything....nice catch on finding that. Totally agree that Blender, much like Handbrake, are realistic work loads and both will stress the CPU as much as a realistic workload can.