Question I've got to find a better CPU cooling solution.

BoomerD

No Lifer
Feb 26, 2006
62,678
11,019
136
2 years ago, I bought an IBuyPower desktop from Costco.
i7-9700k cpu, ASUS 2070 GPU, ASRock Phantom Gaming 4 mobo, 2x8 gb RAM. Decent specs.
CPU cooler was their own branded single radiator AIO liquid cooler.
Case is their own brand of case, designed for liquid cooling.(rear fan, room in the top for 2 x 120 or maybe 140mm fans, no side panel fan, front of the case is pretty tight, not sure there's any fan mounting room or air flow there.
Warranty was, 1 year factory, 1 year Costco.
After about 15 months, the water cooling unit crapped out. Since it was now into the Costco portion of the warranty, I had to ship it to some repair place in Wisconsin that Costco contracted.
They replaced it with a low-end CoolerMaster single radiator unit. (Their workmanship was fckng horrid, but that's another thread)
Worked relatively well for about 6 months...then the CPU started getting much hotter than before. (Not CRITICALLY hot...in the 80's during gaming) then taking forever to cool back down.
I grabbed a screwdriver and tightened the cooler mounts...all 4 screws were a bit loose, but not terribly so.
Didn't really make a difference.

So...looks like my options are...
1) grab some TIM, clean and reset the existing cooler pump. (I THINK I have some Noctua goop in the package left over from my last build...gotta be 10 years old...think it's still any good? Yeah, I have to make sure it hasn't dried out in the tube...)
2) replace the cooler with another liquid cooler...maybe dual radiator mounted in the top. Corsair? CoolerMaster? Other brand?
3) Try to fit a decent quality air cooler...Noctua? in there. (I haven't measured to see if a Noctua cooler @ 158 or 160 mm will fit.)
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Couple of possibilities off hand. The pump could be failing, these cheap units often burn out the pump motors fast. The coolant could be running low.
I would try your idea of re-seating the heat sink/pump block first. If that doesn't work. Replace the whole thing. You would have to measure from the socket to the edge of the case to see if a Noctua would fit, there are other options for good air-coolers.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
I'll add more to Ajay's comments.

It depends on the limitations of the IBuyPower's space-in-the-case and case-width, but I stand by my many earlier pronouncements that the ThermalRight Le Grand Macho RT is the best-performing heatpipe cooler still -- after more than four years on the market. It narrowly leaves the Noctua NH-D15 in second-place.

Second, have you even considered pulling the CPU and sending it to Silicon Lottery in Katy, TX for a $40 de-lid and re-lid with Grizzly Conductonaut? It will be returned to you with the appearance of a factory-fresh CPU. It would void your warranty on the processor -- theoretically -- but I really doubt that either eBuyPower or even Intel would be able to detect the TIM replacement unless they actually delidded the processor themselves. And I think you're no longer eligible for one of two warranties: I never heard of anyone needing the 3-year Intel warranty, anyway.

On some Intel processors with ~95W TDP specs, the TIM replacement can be worth 18C reduction in load temperatures. Silicon Lottery may have some statistics that include results with the i7-9700K. SL is even offering the relid service on more recent CPUs for which Intel returned to using their indium solder formula. Relidding for those processors was still worth another 5 to 6C improvement in temperatures.

Also, in regard to the external application of better TIM between IHS and cooler base, I think that IC Diamond paste will still outperform Noctua grease, and might be worth an additional 2 to 4C improvement. You could, of course, use Conductonaut with caution -- it's a conductive liquid metal -- provided you don't lap the IHS to bare metal -- next paragraph . . .

Again voiding your warranty in a way that the manufacturer would easily detect, you can lap off the nickel plate of the IHS and grab another 2C in thermal improvement.

I will qualify this and the centigrade values I've cited, in that I have this all from personal first-hand experience, but with an i7-6700K and an i7-7700K -- four cores against your eight. But you'd think the identical TDP spec -- 95W -- is the only significant factor in those temperature improvements. I can run the Kaby Lake processor under Intel-Burn-Test or LinX at 5GHz and the temperatures don't climb above the mid-80s C. OC'd to 4.8Ghz, they don't exceed 75C.

MORE INFO: The TR Grand Macho RT has a height of 159mm, measuring from the processor cap or IHS to the pipe protrusions at the top of the cooler. I'm using CoolerMaster Stacker 832 cases, which have a normal mid-tower case-width and a 1-inch thick hinged plastic door (for mounting fans) inside the side panel which could potentially obstruct a cooler like the LGM RT, but it doesn't. I just can't install a fan in the door and in a position directly above the cooler.

You would just have to measure the distance between processor cap and the inner sidepanel surface of this iBuyPower case. There may also be additional air-cooling options, but not as good.

Of course, you might not want to spend $80 for a cooler. I personally don't have money to just throw away, but I don't consider an $80 cooler or $40 delid-relid to be an extravagance. It's less of an outlay than a custom water kit.
 

BoomerD

No Lifer
Feb 26, 2006
62,678
11,019
136
I'll add more to Ajay's comments.

It depends on the limitations of the IBuyPower's space-in-the-case and case-width, but I stand by my many earlier pronouncements that the ThermalRight Le Grand Macho RT is the best-performing heatpipe cooler still -- after more than four years on the market. It narrowly leaves the Noctua NH-D15 in second-place.

Second, have you even considered pulling the CPU and sending it to Silicon Lottery in Katy, TX for a $40 de-lid and re-lid with Grizzly Conductonaut? It will be returned to you with the appearance of a factory-fresh CPU. It would void your warranty on the processor -- theoretically -- but I really doubt that either eBuyPower or even Intel would be able to detect the TIM replacement unless they actually delidded the processor themselves. And I think you're no longer eligible for one of two warranties: I never heard of anyone needing the 3-year Intel warranty, anyway.

On some Intel processors with ~95W TDP specs, the TIM replacement can be worth 18C reduction in load temperatures. Silicon Lottery may have some statistics that include results with the i7-9700K. SL is even offering the relid service on more recent CPUs for which Intel returned to using their indium solder formula. Relidding for those processors was still worth another 5 to 6C improvement in temperatures.

Also, in regard to the external application of better TIM between IHS and cooler base, I think that IC Diamond paste will still outperform Noctua grease, and might be worth an additional 2 to 4C improvement. You could, of course, use Conductonaut with caution -- it's a conductive liquid metal -- provided you don't lap the IHS to bare metal -- next paragraph . . .

Again voiding your warranty in a way that the manufacturer would easily detect, you can lap off the nickel plate of the IHS and grab another 2C in thermal improvement.

I will qualify this and the centigrade values I've cited, in that I have this all from personal first-hand experience, but with an i7-6700K and an i7-7700K -- four cores against your eight. But you'd think the identical TDP spec -- 95W -- is the only significant factor in those temperature improvements. I can run the Kaby Lake processor under Intel-Burn-Test or LinX at 5GHz and the temperatures don't climb above the mid-80s C. OC'd to 4.8Ghz, they don't exceed 75C.

MORE INFO: The TR Grand Macho RT has a height of 159mm, measuring from the processor cap or IHS to the pipe protrusions at the top of the cooler. I'm using CoolerMaster Stacker 832 cases, which have a normal mid-tower case-width and a 1-inch thick hinged plastic door (for mounting fans) inside the side panel which could potentially obstruct a cooler like the LGM RT, but it doesn't. I just can't install a fan in the door and in a position directly above the cooler.

You would just have to measure the distance between processor cap and the inner sidepanel surface of this iBuyPower case. There may also be additional air-cooling options, but not as good.

Of course, you might not want to spend $80 for a cooler. I personally don't have money to just throw away, but I don't consider an $80 cooler or $40 delid-relid to be an extravagance. It's less of an outlay than a custom water kit.

I'm not "enthusiast crazy" enough to bother with the de-lidding process...nor with trying to lap the processor. Way more trouble than I'm willing to go through. I'll have to remove the pump head to try to replace the TIM, that would be the best time to measure the distance to the side panel. $80 isn't really that bad for a quality air cooler...actually cheaper than Noctua's NH-U12A...which is the shorter of their "better" coolers. That one is over $100 everywhere. I'll look at the ThermalRight...
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
I'm not "enthusiast crazy" enough to bother with the de-lidding process...nor with trying to lap the processor. Way more trouble than I'm willing to go through. I'll have to remove the pump head to try to replace the TIM, that would be the best time to measure the distance to the side panel. $80 isn't really that bad for a quality air cooler...actually cheaper than Noctua's NH-U12A...which is the shorter of their "better" coolers. That one is over $100 everywhere. I'll look at the ThermalRight...
Well, you don't have to "bother with" it. You just pull your CPU, put it in an anti-stat envelope and then a small mailing box, go to Silicon Lottery, place your order for relidding with Conductonaut, and follow the instructions. If you mail it out on some day and date, it can probably come back to you in a week and less than 2 weeks. You pay $40, and there's a warranty.

It was worth 18C temperature reduction at OC-load to me.

But I can't see that there's any trouble with having it done, especially by someone who's done it 10,000 times. You don't have to fiddle with a razor-blade, and he doesn't use a razor-blade. He probably uses the Der8auer* De-Lid Die-Mate tool (also $40), and he applies liquid tape insulation on the PCB around the die. He's apparently damn good at applying the Conductonaut. I've had a Skylake 6700K that I bought from Silicon Lottery for the last four years, and there is no change in its thermal stats.

If you're going to change out a water-block for a heatpipe cooler, You only have to flip a switch to remove the CPU and put it in a bag. Put it in a box. Take it to the post-office.

And wait. Then put it back in the socket, install your water-block and heatsink with a good TIM -- I recommended IC DIamond -- and . . . . DONE.

No sandpaper, no lapping, no fuss, no bother -- no mess.
---------------------------------------
* Per the young German engineer who designed the De-Lid Die-Mate II, it shouldn't have taken me this long to figure out the name with the "8". "8". Das ist acht. Der-acht-auer.
 
Last edited:

BoomerD

No Lifer
Feb 26, 2006
62,678
11,019
136
From the Silicone Lottery wepsite:

Peak core temperatures under a heavy overclocked load typically decrease anywhere from 15°C to 25°C when delidded for Intel 3rd through 8th generation, 3°C to 7°C for Intel 9th generation, and 5°C to 12°C for Intel 10th generation.

$49.99 plus shipping for 3-7 degrees doesn't seem worth it to me.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
From the Silicone Lottery wepsite:



$49.99 plus shipping for 3-7 degrees doesn't seem worth it to me.
Then Gen 9 was prepared with the older, more effective Intel solder. I can see that you might not think it's worth it for only 3 to 7 C improvement.

Different strokes for different folks. I probably fall into the category of "enthusiast" because I've been "doing computers" much of my life.

I stuck to air-cooling, even after investigating exotic Rube Goldberg solutions like evaporative chilled water. I've flirted with the idea of custom water cooling, and AiO coolers. But in the end, I didn't want the maintenance, the possibility of pump failure,, leaks or any of it.

So I followed the "pursue the last grain of rice" strategy, since all cooling improvements like lapping, superior TIMs -- and ultimately, delidding -- are additive.

There IS another option -- much better for an AiO water-cooler than a 900-gram heatpipe cooler with some amount of torque on the motherboard. You could apply the waterblock on a bare-die processor -- removing the IHS. You would definitely NOT use a TIM like IC Diamond. You could use Kryonaut, possibly Conductonaut -- the latter with some caution because of its conductive properties.

Someone else could advise you further. Otherwise, your existing choices should get you far enough to keep the temperatures within spec for that processor.

Also, I cannot criticize anyone for buying a desktop or gaming rig at COSTCO. It's probably a very good machine. I personally haven't bought an OEM machine since 1994. It costs me more money, but I prefer it that way. So I guess I'm an "enthusiast" for sure.

One more thing I just thought of.

After Intel stopped publishing a "safe voltage spec" after Nehalem, they were meeting with their "motherboard partners" to offer guidelines on processor lifespan, electro-migration, heat degradation and voltage. For instance, not too long ago -- when Skylake and Kaby Lake were "new" -- they told the mobo makers that their "auto" voltage setting could take VCORE up to 1.4 V for turbo speed.

This meant that, in many cases, processors would be overvolted unnecessarily if the motherboard setting was left at "Auto" for VCORE. If you can enter the BIOS of your iBuyPower system, you might be able to reduce voltage enough to run with total stability -- maybe even with overclock settings -- and cooler temperatures.

The first thing I'd do is to enter the BIOS to see if the VCORE has an "Auto" setting. Then, go from there.

My Kaby Lake VCORE on default settings ("Auto" voltages, etc.) would show a loaded VCORE of 1.3V for 4.5Ghz speed. Right now, with my cooling solution, it only peaks to 1.36V for 4.8 Ghz, and the stock speed required only less voltage to run at stable and optimal performance. And these settings have a margin of surplus. They allow me to run AVX2 stress tests at that 4.8 speed without missing a lick.

Someone else might offer more information about this, or they might correct me, or disagree. But it is definitely something to look into.
 
Last edited:

BoomerD

No Lifer
Feb 26, 2006
62,678
11,019
136
Then Gen 9 was prepared with the older, more effective Intel solder. I can see that you might not think it's worth it for only 3 to 7 C improvement.

Different strokes for different folks. I probably fall into the category of "enthusiast" because I've been "doing computers" much of my life.

I stuck to air-cooling, even after investigating exotic Rube Goldberg solutions like evaporative chilled water. I've flirted with the idea of custom water cooling, and AiO coolers. But in the end, I didn't want the maintenance, the possibility of pump failure,, leaks or any of it.

So I followed the "pursue the last grain of rice" strategy, since all cooling improvements like lapping, superior TIMs -- and ultimately, delidding -- are additive.

There IS another option -- much better for an AiO water-cooler than a 900-gram heatpipe cooler with some amount of torque on the motherboard. You could apply the waterblock on a bare-die processor -- removing the IHS. You would definitely NOT use a TIM like IC Diamond. You could use Kryonaut, possibly Conductonaut -- the latter with some caution because of its conductive properties.

Someone else could advise you further. Otherwise, your existing choices should get you far enough to keep the temperatures within spec for that processor.

Also, I cannot criticize anyone for buying a desktop or gaming rig at COSTCO. It's probably a very good machine. I personally haven't bought an OEM machine since 1994. It costs me more money, but I prefer it that way. So I guess I'm an "enthusiast" for sure.

One more thing I just thought of.

After Intel stopped publishing a "safe voltage spec" after Nehalem, they were meeting with their "motherboard partners" to offer guidelines on processor lifespan, electro-migration, heat degradation and voltage. For instance, not too long ago -- when Skylake and Kaby Lake were "new" -- they told the mobo makers that their "auto" voltage setting could take VCORE up to 1.4 V for turbo speed.

This meant that, in many cases, processors would be overvolted unnecessarily if the motherboard setting was left at "Auto" for VCORE. If you can enter the BIOS of your iBuyPower system, you might be able to reduce voltage enough to run with total stability -- maybe even with overclock settings -- and cooler temperatures.

The first thing I'd do is to enter the BIOS to see if the VCORE has an "Auto" setting. Then, go from there.

My Kaby Lake VCORE on default settings ("Auto" voltages, etc.) would show a loaded VCORE of 1.3V for 4.5Ghz speed. Right now, with my cooling solution, it only peaks to 1.36V for 4.8 Ghz, and the stock speed required only less voltage to run at stable and optimal performance. And these settings have a margin of surplus. They allow me to run AVX2 stress tests at that 4.8 speed without missing a lick.

Someone else might offer more information about this, or they might correct me, or disagree. But it is definitely something to look into.

This is the first pre-built PC I've bought since the early 90's. I prefer to build my own...but just didn't want to mess with finding "just the right parts," assembling everything and HOPING it works...and I've never quite gotten over that "crunch" when seating a CPU in the socket...
I've never really been into overclocking...I'd rather just buy one that will do the job right out of the box. (yes, I understand the appeal, and have played with it a bit over the years...but I almost always revert to stock.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
This is the first pre-built PC I've bought since the early 90's. I prefer to build my own...but just didn't want to mess with finding "just the right parts," assembling everything and HOPING it works...and I've never quite gotten over that "crunch" when seating a CPU in the socket...
I've never really been into overclocking...I'd rather just buy one that will do the job right out of the box. (yes, I understand the appeal, and have played with it a bit over the years...but I almost always revert to stock.
Oh -- You don't need to overclock it. You might find to your pleasant surprise that you can under-volt it.

It shouldn't take very long, and if you have a utility like OCCT for the stress-testing, it should trap errors at the low margin of voltage before your system throws a blue screen. The free version allows for maximum 1-hour tests; the one-year license allows longer test definitions. The one-year license is $25. That's why my build-it project for this year needs to be finished in about 10 months. I don't want to pay for it again this year.

You need to find the loaded VCORE value that would be shown in the OCCT monitoring screens at the "Auto" setting, provided that the BIOS shows you it's set to "Auto". Then, you'd find an "Adaptive mode" setting -- the number of millivolts above idle speed to which the processor "turbos" to give you the stock maximum VCORE, that allows you to boot to Windows. You'd then adjust this Adaptive setting -- using OCCT tests -- to find the lowest stable voltage. You might then add maybe 10 mV for a comfortable margin. But it is quite possible that this final, stable setting is lower than the motherboard is currently volting the processor for turbo.

Also, for that processor, you should be able to find some reviews -- possibly here at Anandtech -- where someone has plotted stable VCOREs against speed, so you have an idea of what the most efficient and coolest voltage actually is -- more or less. You would tweak the Adaptive voltage to give you totally stable tests with AVX2 settings in OCCT until you're satisfied.

But I'm pretty sure what I said about the "motherboard partners" and their default, Auto settings is true. I can only be sure of it for my generation of processor, motherboard and chipset, but it's something worth exploring in your case.

The first step, of course, is to poke around in the BIOS to find out what is actually going on there, then proceed accordingly.

IN FACT, there's another thing. If the RAM used by iBuyPower is not low-end "basic" models, it might be "OC" RAM. All of us who are building systems with DDR4 RAM are buying the kits spec'd at above DDR4-3000. I built my rigs with DDR4-3200 TridentZ RAM.

What I discovered, however, is that you cannot just set the XMP RAM profile in BIOS for the spec speed and expect the "Auto" VCCIO and VCCSA voltages for the motherboard to be "reasonable". The board may push the IMC's VCCIO to close to the "safe limit". When I did this with my G.SKILL TridentZ RAM -- forgetting to initially set the memory speed to the Intel spec, the board was volting the IMC to 1.33V, and 1.35V is considered a threshold between "OK" and "excessive". Then, you might find that the VCCIO monitored value at a "spec" Intel speed like 2666 is a lower limit of what will suffice to run the "OC" modules at their full spec speed.

The memory controller voltage has a lot of impact on processor temperature, just as VCORE does.

I"m wondering if iBuyPower didn't just throw in some RAM spec'd at a certain speed, and left the motherboard "Auto" setting to determine the memory controller voltage and/or the system agent voltage.

I'm not familiar with motherboards after the Z170 chip, but the processor you're using deploys the same Skylake cores that I have, with some minor improvements and a greater number of cores.

I think you'll find other threads and posts here or in more appropriate forums suggesting that the future of "overclocking" may simply be "undervolting". And that includes undervolting at stock speed settings.

There may even be a recent article here at Anandtech exploring that theme.

I think if I bought an OEM system today, I"d be attempting to find out how much they gave a damn about pushing up processor temperatures with sloppy BIOS settings.
 
Last edited:

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
2 years ago, I bought an IBuyPower desktop from Costco.
i7-9700k cpu, ASUS 2070 GPU, ASRock Phantom Gaming 4 mobo, 2x8 gb RAM. Decent specs.
CPU cooler was their own branded single radiator AIO liquid cooler.
Case is their own brand of case, designed for liquid cooling.(rear fan, room in the top for 2 x 120 or maybe 140mm fans, no side panel fan, front of the case is pretty tight, not sure there's any fan mounting room or air flow there.
Warranty was, 1 year factory, 1 year Costco.
After about 15 months, the water cooling unit crapped out. Since it was now into the Costco portion of the warranty, I had to ship it to some repair place in Wisconsin that Costco contracted.
They replaced it with a low-end CoolerMaster single radiator unit. (Their workmanship was fckng horrid, but that's another thread)
Worked relatively well for about 6 months...then the CPU started getting much hotter than before. (Not CRITICALLY hot...in the 80's during gaming) then taking forever to cool back down.
I grabbed a screwdriver and tightened the cooler mounts...all 4 screws were a bit loose, but not terribly so.
Didn't really make a difference.

So...looks like my options are...
1) grab some TIM, clean and reset the existing cooler pump. (I THINK I have some Noctua goop in the package left over from my last build...gotta be 10 years old...think it's still any good? Yeah, I have to make sure it hasn't dried out in the tube...)
2) replace the cooler with another liquid cooler...maybe dual radiator mounted in the top. Corsair? CoolerMaster? Other brand?
3) Try to fit a decent quality air cooler...Noctua? in there. (I haven't measured to see if a Noctua cooler @ 158 or 160 mm will fit.)
Just looking at your initial post again. This has gone a bit far afield, discussing the under-volting possibilities. But if you're collecting sufficient observations that show processor temperatures increasing, and you've been through a couple AiO or low-end liquid coolers, the hardware is the first thing I'd look at. I'd recommend something like the TR "Le Grand Macho RT". Just get the best, heatpipe cooler, because that will put you within 4 or 5C of EKWB"s "customizable" AiO and some other good ones. Then, look at the BIOS and proceed from there.

The MTBF spec for a heatpipe cooler as offered many years ago by NASA's calculation was 1 million years. And conventional computer cooling usually has two, three or more fans for intake and exhaust, so even the heatpipe-mounted CPU fan has some backup, and good fans last.

Then, we have also a current thread on the ICE GIANT ProSiphon, the video showing only its comparison with a Noctua heatpipe unit and a 64-core ThreadRipper. The Ice Giant could probably be described as a heatpipe cooler that works differently than either the TR or other models, or it depends on gravity. So it has to be mounted a certain way. But it all proves that water is unnecessary, or low-end water coolers may be outperformed by high-end heatpipe coolers.

But I can't see any damage yet to your electronic hardware, like the CPU. The temperatures of which you complain are still within Intel specs for "warranty survival". Ajay could be right about the pump not operating properly. The processor will throttle if temperatures get too high, and yours doesn't seem to be heading into the 90's C range. There certainly can't be anything wrong with the ASRock board, if it's been running this long, and it's not threatened by the problem. It's probably as Ajay suspects.

After that, well, you can look at my long digressions.
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
This is the first pre-built PC I've bought since the early 90's. I prefer to build my own...but just didn't want to mess with finding "just the right parts," assembling everything and HOPING it works...and I've never quite gotten over that "crunch" when seating a CPU in the socket...
I've never really been into overclocking...I'd rather just buy one that will do the job right out of the box. (yes, I understand the appeal, and have played with it a bit over the years...but I almost always revert to stock.
Well, when you come to your senses and do your own build again, post a thread and we'll help you get the right parts. If we don't notice it, there are plenty of guys you can PM, including me.
 

BoomerD

No Lifer
Feb 26, 2006
62,678
11,019
136
Well, when you come to your senses and do your own build again, post a thread and we'll help you get the right parts. If we don't notice it, there are plenty of guys you can PM, including me.

Thanks for the offer. ..often, though, what others think are the right parts are different than what I think they should be...:p

This IBP rig has been OK, the only problem (so far) has been the low end AIO water coolers. I need to see if there's room enough for a good air cooler. I have a Noctua cooler in my last build. It worked flawlessly...and still does, 10 years later. Unfortunately, Noctua doesn't recommend it for my current processor.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,839
3,174
126
be quiet dark pro4 gets a vote from me, if your looking for an air cooler similar to the Noctua D15.

The Scythe Fuma and Mugan would also be another recommendation if you really want to save that little bit of extra cash, although the dark rock pro4 is a really good heat sink and i feel worth the extra.

There is also Thermalalright Macho.

Personally i am not a big fan of AIO's unless i am space constraint near the cpu socket.
I prefer large air sinks, unless its too large with good mounts.
 
Last edited:

BoomerD

No Lifer
Feb 26, 2006
62,678
11,019
136
be quiet dark pro4 gets a vote from me, if your looking for an air cooler similar to the Noctua D15.

The Scythe Fuma and Mugan would also be another recommendation if you really want to save that little bit of extra cash, although the dark rock pro4 is a really good heat sink and i feel worth the extra.

There is also Thermalalright Macho.

Personally i am not a big fan of AIO's unless i am space constraint near the cpu socket.
I prefer large air sinks, unless its too large with good mounts.

I've looked at the Dark Rock coolers before. They look great...and are a bit cheaper than the Noctua NH-U12A , but only a 3 year warranty compared to 6 with Noctua...and they're within a couple of mm the same height. Rough estimate is about 160-165 mm clearance for a cooler...that leaves out the Noctua NH-D15...their flagship model. Just -> <- much too tall.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,839
3,174
126
but only a 3 year warranty compared to 6 with Noctua

but you know with air sinks the only real thing that goes bad is the fans.

Also how big of a overclock are we talking about?
If its not too big, i would push a Noctua C14s and not D15, as it pushes air down to the board and keeps other things cool on the board as a secondary.
 

BoomerD

No Lifer
Feb 26, 2006
62,678
11,019
136
but you know with air sinks the only real thing that goes bad is the fans.

Also how big of a overclock are we talking about?
If its not too big, i would push a Noctua C14s and not D15, as it pushes air down to the board and keeps other things cool on the board as a secondary.

I don't really overclock. I DO let Intel's "Extreme Tuning Utility" work SOME of its magic...but nothing too heavy.

I've looked at that noctua cooler...their website's compatibility page for the i7-9700K sez:
If power limits are disabled in BIOS, CPU might not be able to keep maximum turbo clock under prolonged AVX loads, suggested maximum power limit: 205W
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,839
3,174
126
well considering the rock pro4 and d15 are simular and raited at about 250W, well, 205W isn't bad for a sink which also throws air at your board and not out the rear.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
well considering the rock pro4 and d15 are simular and raited at about 250W, well, 205W isn't bad for a sink which also throws air at your board and not out the rear.
The manufacturers -- or whomever -- may rate the coolers according to a maximum thermal wattage, but I've always stuck to comparing two or more coolers tested on the same overclocked platform. Usually, the tests include a "stock" platform test result, and an "OC'd" result. The OC'd results are usually based on some processor within the last three generations -- sometimes the 95W TDP processors, other times those with 125W TDP's.

The measured results are shown in the temperature changes between room-ambient (idle) and a load value, and they are rank-ordered. So if you have coolers A, B and C where A and B are in one lab-test review of 20 different coolers, and B and C are rank-ordered among another 20 coolers, and you're interested in the performance of A versus C, the two reviews together will give you the answer.

Other measurements are the thermal resistance, and its inverse of thermal conductivity. AT least, that's the way I understand those two concepts. But generally, the temperature spreads between idle and load will rank-order consistent with thermal resistance.

That's how I stand behind my pronouncement about the TR LGM "Macho" RT cooler. You can say that you like the shape, size or fan-orientation of this or that cooler, but ultimately its performance can be measured against others. Over the last several years, they've been including AiO water-coolers among air-cooling results.

The only loose joint in all this is the cooler fans bundled with each make and model. Certainly, they compare noise ratings, but other than that, not all fans are equal, although they're only unequal to a limit. You can push so many CFM between cooler fins before it no longer makes any difference.