What is the better cooler nh-d15s w/ 1 fan vs EKWB XLC Predator 140 w/ 2 fans push/pull

[Laststop311]

My build is as follows.

Case: Fractal Design R5 , 2x 140mm front intake fans Fractal Venturi HF fans, 1X 140mm rear exhaust Fractal venturi fan

MoBo: Asus ROG IX Hero

CPU: i7-7700k with a hopeful overclock of 5Ghz+

CPU TIM: Coollaboratory Liquid Metal Ultra

Ram: Gskill Ripjaws V 2x8GB (16GB) DDR4 3200mhz Cas 14 w/ eventual upgrade to 4x8GB (32GB) (this may change as DDR4 is rapidly improving, currently 3200Mhz cas 14 is the fastest combo of specs for quickest reading of data)

PSU: EVGA 750 watt G1 80+ gold 10 yr warranty full modular w/ braided cables

GPU: EVGA GTX 1070 FTW Hybrid Gaming 8GB 1607Mhz Clock Boost clock 1797Mhz, with a hopeful overclock as close to 2000Mhz+ as I can, w/ 2x 100mm push/pull high static pressure SilenX effizio fans exhausting out of side panel

Storage: Crucial MX300 2TB SSD + HGST 6TB HE 7200 RPM HDD (eventually down the line I will add a 256-512GB Intel optane m2 SSD drive as my main system boot drive and most accessed apps and games as I plan on keeping this PC at least 10 years)

Front panel 5.25": Asus UPD 3.1 panel 2x USB 3.1 gen 2 type C w/ 100 watt power supply via 2x molex connectors and sata express data connection

For the CPU cooler I cannot find any information on how the EKWB XLC predator 140 AIO water cooler performs. I know it is much higher quality than the standard AIO uses a lot of copper throughout and has a pump that is suspended to not transfer vibration and uses an official EK waterblock with the pump built into the radiator, it will have 2x 140mm high static pressure EK fans in push/pull. The noctua nh-d15s performs within a couple degrees of the regular nh-d15 and is very quiet but I'm not sure which will provide lower temps the noctua nh-d15s will use 1x 150mm noctua fan in the center. Which cooler is better for overall noise and temp control? Another thing I'd like to note is with the EK predator it is much less weight on the mobo I'm sure the mobo can handle the nh-d15 but with the mobo vertical and all the weight it supports you have to be very careful when moving the tower, this slightly influences my decision also. If they are equal performers the weight difference would sway my decision.

Please vote in my poll and respond with why you think 1 is better than the other. Do not factor in price as why one is better than the other.

 

[BonzaiDuck]

My build is as follows.

Case: Fractal Design R5 , 2x 140mm front intake fans Fractal Venturi HF fans, 1X 140mm rear exhaust Fractal venturi fan

MoBo: Asus ROG IX Hero

CPU: i7-7700k with a hopeful overclock of 5Ghz+

CPU TIM: Coollaboratory Liquid Metal Ultra

Ram: Gskill Ripjaws V 2x8GB (16GB) DDR4 3200mhz Cas 14 w/ eventual upgrade to 4x8GB (32GB) (this may change as DDR4 is rapidly improving, currently 3200Mhz cas 14 is the fastest combo of specs for quickest reading of data)

PSU: EVGA 750 watt G1 80+ gold 10 yr warranty full modular w/ braided cables

GPU: EVGA GTX 1070 FTW Hybrid Gaming 8GB 1607Mhz Clock Boost clock 1797Mhz, with a hopeful overclock as close to 2000Mhz+ as I can, w/ 2x 100mm push/pull high static pressure SilenX effizio fans exhausting out of side panel

Storage: Crucial MX300 2TB SSD + HGST 6TB HE 7200 RPM HDD (eventually down the line I will add a 256-512GB Intel optane m2 SSD drive as my main system boot drive and most accessed apps and games as I plan on keeping this PC at least 10 years)

Front panel 5.25": Asus UPD 3.1 panel 2x USB 3.1 gen 2 type C w/ 100 watt power supply via 2x molex connectors and sata express data connection

For the CPU cooler I cannot find any information on how the EKWB XLC predator 140 AIO water cooler performs. I know it is much higher quality than the standard AIO uses a lot of copper throughout and has a pump that is suspended to not transfer vibration and uses an official EK waterblock with the pump built into the radiator, it will have 2x 140mm high static pressure EK fans in push/pull. The noctua nh-d15s performs within a couple degrees of the regular nh-d15 and is very quiet but I'm not sure which will provide lower temps the noctua nh-d15s will use 1x 150mm noctua fan in the center. Which cooler is better for overall noise and temp control? Another thing I'd like to note is with the EK predator it is much less weight on the mobo I'm sure the mobo can handle the nh-d15 but with the mobo vertical and all the weight it supports you have to be very careful when moving the tower, this slightly influences my decision also. If they are equal performers the weight difference would sway my decision.

Please vote in my poll and respond with why you think 1 is better than the other. Do not factor in price as why one is better than the other.

I just happened to have addressed that question myself -- indirectly. I was planning to use a Predator 240 for my Skylake rig. Planned to do it for four months, then crawfished and chose a ThermalRight Le Grand Macho.

I think you could safely infer that the Predator 140 will be less effective in CPU cooling than the Predator 240. I found a review comparison that included the NH-D15, the Predator 240 and the LG Macho. The Macho just edges out over the D15 by 1C degree on the same test-bed. By comparison, the PRedator 240 beats the Macho by 5C.

So you would guess -- infer--- that the Predator 140 would either be close, slightly ahead or behind the D15 by so little that other factors might affect your choice instead. I would guess that for the single-fan and radiator size, the D15 would outperform.

 

[Laststop311]

I just happened to have addressed that question myself -- indirectly. I was planning to use a Predator 240 for my Skylake rig. Planned to do it for four months, then crawfished and chose a ThermalRight Le Grand Macho.

I think you could safely infer that the Predator 140 will be less effective in CPU cooling than the Predator 240. I found a review comparison that included the NH-D15, the Predator 240 and the LG Macho. The Macho just edges out over the D15 by 1C degree on the same test-bed. By comparison, the PRedator 240 beats the Macho by 5C.

So you would guess -- infer--- that the Predator 140 would either be close, slightly ahead or behind the D15 by so little that other factors might affect your choice instead. I would guess that for the single-fan and radiator size, the D15 would outperform.

OK, Well that makes things a lot easier. I've been installing big air coolers for years. Was kinda leaning that way anyways due to story after story of the predator leaking.

 

[BonzaiDuck]

OK, Well that makes things a lot easier. I've been installing big air coolers for years. Was kinda leaning that way anyways due to story after story of the predator leaking.

I'm not going to push you one way or the other. Last time I looked, the PRedator 240 was about $200. I think I spent close to $80 on the Macho, which is now $10 less:

Le Grand Macho at the Egg

It's actually some 90 grams lighter than the NH-D15, is more compact, and doesn't hang over the RAM modules. If by chance you take a fancy to the Macho, buy the $8 accessory duct for it, and port it to your rear exhaust fan.

I apparently was the only one so far who left a review at the Egg about it. You might want to look at that.

I spent another $50 to have my Skylake re-lidded with CLU for a 12C improvement. So the Macho with the CLU-relid beats the EKWB by 7C.

And I don't have all the failure points, integrated pump that can fail, etc. You have to embrace the complexity of water cooling even with an AiO. I figured I didn't need to do any "embracing" for now.

 

[XavierMace]

Why are you limiting yourself to the 140? You can fit 280mm radiators in your case.

 

[Laststop311]

It wont fit with the asus 5.25" 2x usb 3.1 type c port add on. I require a new pc build to have 10gbps 100 watt front type c ports for future proofing. Also I'm not really liking the dozens of stories saying it leaks and pumps are known to go bad after a few years. And the nh d15s with 1 fan won't interfere with my ram. I think the d15 will be good enough to keep the processor 75c or lower at 5ghz and it's more reliable plus its like half the price as a bonus.

 

[StefanR5R]

The DH15S has a lot more fin area than the Predator 140.

Noctua dual tower air coolers:

I am using the DH14 on an overclocked 6-core Ivybridge-E. Under some circumstances, this could run into the temperature limit and throttle mildly.

I am using DH15S on several mildly overclocked Sandybridge i5. Under worst CPU load, the fan merely reaches 800 RPM which is barely audible (inaudible in an typical office environment) while temperatures remain moderate.

I am using DH15S also on a few 14-core 140 W Broadwell-EP Xeons. Under full CPU load, the fans spin not much faster than 700 RPM and keep the Xeons at about 62 °C, one even at merely 55 °C (room temperature is 23 °C).

EKWB all-in-one water coolers:

I am using Predator EK-XLC 360 with a moderately overclocked 10-core Broadwell-E. I replaced the Predator's 3 stock fans by 3 Noctua NF-P12 which are slower than the stock fans, have a psychoacoustically optimized noise profile, and lack the bearing noises that the original fans exhibit. However, the P12s at full speed (circa 1200 RPM) are barely able to keep the CPU temperature in check under full load.

Recently I added a GTX1080 into this loop, and surprisingly, matters worsened only a little even when both CPU and GPU are under load. I.e. the EK-XLC 360 is able to keep up with the combined heat output of CPU and GPU even with the slower fans. But since 1200 RPM are very much audible, I am going to add at least one more large radiator into this loop when I find the time.

Also note that while the fans spin down to inaudible RPMs when the PC is idle, you cannot spin down the pump as far as the fans can be spun down. On my Asus motherboard, the lowest pump speed which I can achieve is 1200 RPM. (Actually I keep the pump at 1200 RPM even when CPU and GPU are under load; this is totally sufficient. Only fan speed matters; and radiator surface of course.) At already little more than that minimum speed, the pump would become audible for me.

I currently have this setup in a Fractal Design R5 PCGH edition (basically an R5 with a closed and dampened top side), whose closed design is beneficial to reduce the remaining pump noise, but is somewhat detrimental to air flow and, in the PCGH edition, lacks space for a second large radiator.

In my opinion, water coolers only make sense if provided with very large radiator surface, or/ and if it is not possible to place a sufficiently large air cooler directly on the CPU or GPU.

CPU: i7-7700k with a hopeful overclock of 5Ghz+

With this CPU and this aim for overclocking, core temperatures will probably depend more on whether or not you de-lid the CPU, and less so on how large a cooler you choose (air or water alike). Size of the air cooler or radiator would still have an effect on how fast = how loud the fans need to be, but less on how soon you are going to run into a temperature limit. (This is just my guess from what I have read about i7-7700k, not from own experience.)

 

[StefanR5R]

PS,

CPU: i7-7700k with a hopeful overclock of 5Ghz+
[...]
I plan on keeping this PC at least 10 years

One goal looks contrary to the other.

 

[BonzaiDuck]

The DH15S has a lot more fin area than the Predator 140.

Noctua dual tower air coolers:

I am using the DH14 on an overclocked 6-core Ivybridge-E. Under some circumstances, this could run into the temperature limit and throttle mildly.

I am using DH15S on several mildly overclocked Sandybridge i5. Under worst CPU load, the fan merely reaches 800 RPM which is barely audible (inaudible in an typical office environment) while temperatures remain moderate.

I am using DH15S also on a few 14-core 140 W Broadwell-EP Xeons. Under full CPU load, the fans spin not much faster than 700 RPM and keep the Xeons at about 62 °C, one even at merely 55 °C (room temperature is 23 °C).

EKWB all-in-one water coolers:

I am using Predator EK-XLC 360 with a moderately overclocked 10-core Broadwell-E. I replaced the Predator's 3 stock fans by 3 Noctua NF-P12 which are slower than the stock fans, have a psychoacoustically optimized noise profile, and lack the bearing noises that the original fans exhibit. However, the P12s at full speed (circa 1200 RPM) are barely able to keep the CPU temperature in check under full load.

Recently I added a GTX1080 into this loop, and surprisingly, matters worsened only a little even when both CPU and GPU are under load. I.e. the EK-XLC 360 is able to keep up with the combined heat output of CPU and GPU even with the slower fans. But since 1200 RPM are very much audible, I am going to add at least one more large radiator into this loop when I find the time.

Also note that while the fans spin down to inaudible RPMs when the PC is idle, you cannot spin down the pump as far as the fans can be spun down. On my Asus motherboard, the lowest pump speed which I can achieve is 1200 RPM. (Actually I keep the pump at 1200 RPM even when CPU and GPU are under load; this is totally sufficient. Only fan speed matters; and radiator surface of course.) At already little more than that minimum speed, the pump would become audible for me.

I currently have this setup in a Fractal Design R5 PCGH edition (basically an R5 with a closed and dampened top side), whose closed design is beneficial to reduce the remaining pump noise, but is somewhat detrimental to air flow and, in the PCGH edition, lacks space for a second large radiator.

In my opinion, water coolers only make sense if provided with very large radiator surface, or/ and if it is not possible to place a sufficiently large air cooler directly on the CPU or GPU.



With this CPU and this aim for overclocking, core temperatures will probably depend more on whether or not you de-lid the CPU, and less so on how large a cooler you choose (air or water alike). Size of the air cooler or radiator would still have an effect on how fast = how loud the fans need to be, but less on how soon you are going to run into a temperature limit. (This is just my guess from what I have read about i7-7700k, not from own experience.)

Yeah. De-lidding and CLU-replacement is a very effective way to get some decent overclock temperatures. It just doesn't have the glitz and elegant complexity.

On the fans, I tolerate turbulence-white-noise at the high-load end, but choose to muffle the motors. That's not really hard to do, just tedious.

Everybody went to water in times bygone partly to eliminate the noise, but now people are saying that certain air-coolers are more quiet. I'll keep an eye on the water-kits over the next year. Would anyone complain about a 6700K at 4.7 and peak package T of 77C?

 

[Laststop311]

What's so strange about overclocking and keeping the pc long. I'm 33 years old and i still have my first PC I bought with my own money an athlon and it still works. Every PC I've ever built they all still work it's incredibly rare for a cpu to fail. I've literally never seen it happen.

I'm probably going to wait for next year tho because 6 core is coming to mainstream. or maybe even the year after that for pci-e 4.0. I think getting a 6 core and pci-e 4.0 should keep you right for a decade. my i7-980x still does the job well 7 years later and it's overclocked to 4.13ghz it's lasting no problem.

I think the 980x was a rly good buy. Incredible longevity. It's 7 years old but it's only 2 processes behind at 32nm. DX 12 even breathes new life into it as that works rly well on 6 cores. It's nice having the luxury to wait for the perfect time to strike. I think 6 core and pci-e 4.0 is the time to do it.

 

[StefanR5R]

What's so strange about overclocking and keeping the pc long. I'm 33 years old and i still have my first PC I bought with my own money an athlon and it still works. Every PC I've ever built they all still work it's incredibly rare for a cpu to fail. I've literally never seen it happen.

It may work, but IMO there remains a risk. I too have not heard of cases where a CPU was entirely failing after some time of overclocked use (aside from LN2 kamikaze runs). Neither had I issues with CPUs which I overclocked myself so far (but I never overvolted, or overvolted only a little).

I have heard and read several times from people who used e.g. overclocked Sandy Bridge for a long time that their CPUs degraded such that they lost stability eventually, and they had to compensate for this either by increasing their voltage, or by reducing the clock. Granted, it is not clear in such cases whether it was indeed the CPU which had degraded, or if it rather was the power delivery components of the motherboard that had degraded over a long time. Either way, obviously they had to accept a performance loss relative to the earlier part of the lifecycle of their overclocked PC.

And this was with 32 nm Sandy Bridge. Longterm behavior of 14 nm+ Kaby Lake, if operated outside Intel's specifications, is obviously still entirely unknown. Likewise, longterm behavior of Z270 boards, when overclocked, is entirely unknown too. So IMO there remains a risk, notably if a long lifetime at the order of 10 years is targeted.

What we do already know is that 5 GHz+ overclocked Kaby Lake will require overvoltage and will be accompanied by core temperatures very near T_junction.

We furthermore already know that delidding is very effective to reduce core temperatures of Kaby Lake. But on the other hand, longterm behavior of the self-made thermal interface between silicon and heat spreader is highly questionable again, even more so if we talk about a targeted lifetime of 10 years. So this speaks against delidding.

 

[BonzaiDuck]

It may work, but IMO there remains a risk. I too have not heard of cases where a CPU was entirely failing after some time of overclocked use (aside from LN2 kamikaze runs). Neither had I issues with CPUs which I overclocked myself so far (but I never overvolted, or overvolted only a little).

I have heard and read several times from people who used e.g. overclocked Sandy Bridge for a long time that their CPUs degraded such that they lost stability eventually, and they had to compensate for this either by increasing their voltage, or by reducing the clock. Granted, it is not clear in such cases whether it was indeed the CPU which had degraded, or if it rather was the power delivery components of the motherboard that had degraded over a long time. Either way, obviously they had to accept a performance loss relative to the earlier part of the lifecycle of their overclocked PC.

And this was with 32 nm Sandy Bridge. Longterm behavior of 14 nm+ Kaby Lake, if operated outside Intel's specifications, is obviously still entirely unknown. Likewise, longterm behavior of Z270 boards, when overclocked, is entirely unknown too. So IMO there remains a risk, notably if a long lifetime at the order of 10 years is targeted.

What we do already know is that 5 GHz+ overclocked Kaby Lake will require overvoltage and will be accompanied by core temperatures very near T_junction.

We furthermore already know that delidding is very effective to reduce core temperatures of Kaby Lake. But on the other hand, longterm behavior of the self-made thermal interface between silicon and heat spreader is highly questionable again, even more so if we talk about a targeted lifetime of 10 years. So this speaks against delidding.

I think the confidence with CLU between IHS and silicon is fairly high -- it doesn't pump out. As for the recent-gen motherboards, mid- to upper-tier boards are made for the overclock BIOS settings. Meanwhile, I have two Sandy Bridge systems -- one running constantly at 4.6/4.7Ghz for 5 years and still going. I give it an annual "shake-down" work-out, and nothing has changed since I built the system.

If you take care of cooling, the thing to worry about is over-volting the processor. It was easy to follow a discipline with Sandy Bridge so that load voltage never exceeded ~ 1.38V. Similarly, the newer processors -- my Skylake for instance -- was touted by Intel as being "more resilient" to voltage, while board-partners have likely been told that 1.4V was safe.

So it depends on the quality of the motherboard, and the prudence of the user for volting the processor.

 

[DrMrLordX]

When I retired my A10-7700k that I had delidded and relidded, I noted no "pump out" from the CLU. There was one bit of the die that was not coated with CLU, though it might have sloughed off when I removed the lid (it stuck to the bottom of the HSF, oops).

 

[BonzaiDuck]

Well, I'm very satisfied with the CLU-re-lid for my i7-6700K. I think I have maybe a 7 or 8C advantage over some watercoolers installed on out-of-the-box retail processors. No hoses, no fittings, no failure points, less noise. Be gentle moving my Stacker 830 mod around, because I fitted a Le Grand Macho to it with a single 140mm pusher attached -- not too weighty, but an Akasa Viper "Round" 140R, rated at 100 to 110 CFM. The puller is the case exhaust with a Noctua iPPC 3000 thermally controlled. If I remember, the LGM with no fans weighs in around 900gm and 80 less than the bare D15.

I've yet to exceed 65C peak package temperature doing anything. IBT will take it to maybe 68 to 69C, and Linx-Affinitized max-problem-max-memory to around 75C @ 4.7 Ghz. 1.424V is beyond my self-imposed limit, and the temperature would probably just exceed 80C. In fact, I can test that, but not now. The CPU was binned as stable at 4.8 and 1.424V [SL says "possibly less" but it's probably within a narrow range of that if you look at affinitized GFLOP stats to find the optimal voltage. ]

Gaming, I've seldom seen it reach 45C on any of the four cores. The GTX 1070 is still part of an unfinished duct and vent refinement, but I don't expect much from finishing it. It may ride around 60 to 65C gaming; 59C with Valley -- jus-a-minute . . . No -- it's 59 to 60C gaming with something between 90+ and 98% GPU usage. The graphics is set so that gaming pushes the core clock to a steady 2038, and the memory to 4,452 or the double-data equivalent of 8,904 Mhz.

. . . . Looking at the Afterburner record, CPU cores may touch 50C but mostly ride in the 30's and 40's. One small spike for one core hits 52C, but for all the cores, the variation goes from 27 to the high 40's C. That's the gaming CPU load that I mean.