Water cooling vs. Fan cooling.

[Jason Rash]

So I'm new to water cooling and working on building my first pc. I Like the idea of water cooling and I think it looks cool. But a little confused on the practicality of it I guess. I know that water cooling is usually used on the cpu, but I have read about the water cooling unit being used on the gpu as well. I think for visual purposes fan cooling the cpu and water cooling the gpu would be cool. But like i stated before not sure about practicality.

Thanks
Jason.

 

[BonzaiDuck]

From the view of my illustrious colleagues here who are old hands at water-cooling, I'm probably the last person you'd want to hear.

Aigomorla, the King of Water-cooling gurus, might agree that, for years, I've been all talk and no walk. I've come into the forum here and there to discuss everything from evaporative cooling rigs to huge external floor-standing radiators.

So I began planning my Skylake system around last June. Between June and September, I was modding my computer case, with a plan to install either an EKWB Predator AiO "customizable" water-cooling kit, or a Swiftech H240 X2 customizable.

But I was also reading comparison reviews keen to inferential cross-references, and I was thorough psyched for the prospect of actually paying someone to de-lid my 6700K processor and then re-lid it with Liquid Ultra for a 12C temperature improvement.

With the case all ready for one of the water-coolers, I finally just decided to get a heatpipe cooler that only fell 5C behind the EKWB Predator kit. Of course, I might have added the 5C to the 12C, but I decided a net 7C improvement in that sort of comparison was good enough.

I feel humble and diminished after looking at rigs in the sticky on custom cases. Some of those rigs are works of art. I can't believe that the builders didn't spend extra time (and money) to reach their achievements in that dimension.

But when it came down to the bottom line -- not so much the money but factors including "number of failure points" and "maintenance," I can't complain about my modest rig. And I keep telling myself, as if I am short on confidence for conviction, that I did the simplest and wisest thing. But that would be the simplest and wisest thing for me, and from my own perspective.

 

[Billb2]

The "practicality" of water cooling is the ability to remove heat faster that fan cooling, though there are more than a few All-In-One (AIO) water coolers that won't do that.

Why would you want to do that? So you can run faster processor speeds that require higher voltages and thus generate more heat.

 

[Jason Rash]

@BonzaiDuck
Yeah. That's something I was worried about; the maintenance and failure points. I was also looking at Swiftech water coolers. If I did do water cooling they were the one I planed to go with. They seemed to simplify things over other brands.

@Billb2
Thanks. I didn't know about AIOs. Hopefully Swiftech does a good job. Really liked them overall. Not sure if I plan on doing much overclocking but I assume water cooling would work well for that.

 

[Billb2]

If you're not overclocking then why bother with water cooling?

 

[Jason Rash]

If you're not overclocking then why bother with water cooling?

I'm thinking about overclocking. So thats why im doing research and trying to get imput on water cooling. This is my first build and I have always been a console gamer. I currently own a labtop but it's been a while since I've owned a desktop. So really new to all this. Thanks for the input. I do appreciate it.

 

[Billb2]

Water Cooling Essential Threads

 

[BonzaiDuck]

I'm thinking about overclocking. So thats why im doing research and trying to get imput on water cooling. This is my first build and I have always been a console gamer. I currently own a labtop but it's been a while since I've owned a desktop. So really new to all this. Thanks for the input. I do appreciate it.

For me, Jason, it all boils down to this.

Am I building my computer to "create a great PC system," or am I building it in some nebulous competition to prove the absolute fastest benchmark results?

If I'm doing it for the fastest benchmarks -- playing devil's advocate here -- do I have any personally-chosen voltage limits beyond which I won't venture? Or do I feel comfortable with some POSSIBLE prospect of replacing my processor in six months or a year?

If there are researched limits, can I reach the highest most likely speed in Mhz/Ghz within those limits, or do I have to settle for "second best," and does second-best still give me the "overall great PC system?"

And if I had another cooling strategy -- de-lidding the processor and restoring the IHS after replacing the TIM with CLU -- can I meet or EXCEED non-CLU/undelidded results of a proven dual-fan water-cooler? See -- with a water-block, I could delid my processor and use CLU between the processor-die and block -- talk about giving it a chill!

But lower temperatures don't pay off in extra Mhz -- or if they do, not too noticeably. If I'm trying to prevent the processor from being heat-stressed, but almost all heat-stress I give it barely takes it over the advisory TCASE temperature spec for OEM builders, and for real-world uses the temperatures barely rise 20+C above idle, then why complicate things just to get another 5, 10 -- even 15C?

For me -- yes -- creating a great PC -- doingk gurr-eat thingks!

I have a preconceived voltage limit: the "Auto" stock load voltages provided by several boards, whose makers may have been briefed on the prospects by Intel, and who want to assure the boards work with ANY processor. My limit for the Skylake is simply a solid 1.408V VCORE, 1,424V VID. Period. The . . . END . . .

And I get the highest, most likely speed for that voltage, while beating or matching in peak package temperature a dual-fan water-cooler at the same speed. More specifically, I come out ahead of an EKWB Predator 240 by about a 7C margin.

So when it comes to my wallet, do I want to put my tempered-glass/tinted-water masterpiece in a museum? Perhaps I have some prosperous person who wants to buy it. But do my friends care? Do they visit all the time, so we can adjourn to the game room and I can show and tell? There are several perfectly laudable scenarios that range widely. But I only got "into" this technology in the '80s because I wanted to do 2-stage least-squares and Box Jenkins Time Series analysis.

As of two days ago, I looked in the mirror and saw Butthead -- sidekick to Beavis -- muttering "I have been to the mountain . . . Huh-ah--Huh-huh. . . . Huh-huh -- huh-huh . .

But when I pick up anything from an x79 to x99 and beyond for a hexa-octo-deca-core processor -- You know damn well I'm going to the water-mod shop.

Ultimately, it came down to a decision that 4-cores is enough for me. Maybe the balance between hardware and software will change. If I want, I can drop a CLU-lidded Kaby into this puppy.

No hurry there, either. . . .

 

[Fatrod]

If you're not overclocking then why bother with water cooling?

I'm interested in water cooling for the noise reduction. But I'm unclear on how much quieter an AIO with say 140mm Rad will be for a GPU.

 

[BonzaiDuck]

I'm interested in water cooling for the noise reduction. But I'm unclear on how much quieter an AIO with say 140mm Rad will be for a GPU.

That was always the promise of it. But I don't need to go into detail about it, and I can't from any firsthand view. You'll see it throughout discussions about watercooling over the past few years. a 360mm radiator could be fitted with six 120mm fans, and push-pull is deemed most effective if you can find da space-in-da-case.

For this, water-cool advocates buy high-quality but lower CFM fans than I would use just for air-cooled intake of exterior air. Lower CFM also implies lower RPM, and lower RPM implies lower dBA -- the noise.

So I've noticed this admission in the discussions that water-cooling isn't all that quiet.

Instead, what I do for my air-cooled rig is to assure a higher total-of-specs CFM intake, a pressurized case, and carefully-chosen exhausts. Usually, you'd have a single exhaust for an air-cooled midtower, but I have two: a barrel-fan that sucks air off the motherboard covered by a restrictive duct, and the traditional case-rear exhaust.

Then, ducting the cooler-rear to the rear-exhaust also makes sure that no warm air is mixing with the case interiror as a whole. Intake air goes under the motherboard duct nearest the PCI backplates and is sucked forward past processor (underneath the lowest cooler-fin), the RAM and the VRM components and heatsinks, then exhausted from the barrel-fan to the right sidepanel. Thus, the heatpipe cooler intake and exhaust have no interaction with the ducted motherboard exhaust, but the air streams are going in opposite directions! If I wanted to -- I could allocate sidepanel and frontpanel intake fans between the two exhausts, and separate their airflow with something like a 140x140mm lexan plate added as a perpendicular extension of the motherboard duct. Don't need to obsess about that or pursue it.

So the only fan allowed a little dBA "liberty" is the rear-exhaust. The barrel-fan could be loud enough, but I isolated it and trimmed the speed vary between 1000 and 2500 RPM. In stress-testing, you can certainly "hear" the Noctua iPPC 3000 rear exhaust, but it's well isolated and covered with a noise-deadening rubber accordion duct.

I have no games or applications that push my system to heat up as much as LinX or Prime95. If it does -- it will stay as cool as needed for the sake of processor health and overclock, but it won't offend the ears. Otherwise, a quiet system.

 

[Fatrod]

Yes I am familiar with the fact that more heat = more rpm = more noise.

It also makes me wonder about 140mm rad vs 360...surely the single 140 is more quiet than 3x120...

From what I hear though its actually the pump that makes more noise than the fan does.

 

[BonzaiDuck]

That's also true. Not too long ago -- maybe a couple years -- there was a discussion with Aigomorla about noise-isolation of a pump. Again, a DIY project with Spire acoustic foam-rubber padding cut to size and even applied in layers would solve some of that.

Any object that generates vibration in contact with another hard-surface will propagate noise. sound waves will propagate through the interior case air and echo within the case directly from the source, but again -- if you can cover the source of the noise with something that dampens the noise acoustically, the noise isn't going to bounce around and propagate so much in the case that way. So you'd first address contact points -- for fans, fans connected to radiators, pumps in contact with the case metal surfaces. Then turn attention to muffling sound transmitted through air as close to the source as possible.

The way Spire is sold in rectangles, either the intention of the manufacturer or the intention assumed by the user involved pasting it all over the metal surfaces of the case. This is an almost irreversible mess, because the stuff doesn't come off so easily, and the adhesive and rubber imbedded in it gets left all over the case surface.

But you can use the stuff to make "collars" that slip on over fan shrouds, or you can pad a foam-board duct with it. You can even stick layered rounds of it on the fixed frame center of the fan hub. If you can place something like a cork wrapped in spire close to the opposite side with the spinning fan hub so things don't get tangled up, you have muffling of sound according to Doppler -- coming and going. This latter idea can be tedious. Basically, though, in situations like an exhaust fan covered with a duct, you can deaden the motor noise of a high-CFM fan to the point where you mostly hear nothing but white-noise of air-turbulence. So if a threshold at which I might hear a fan at 2,000 RPM, I might be able to acoustically deaden that fan so the threshold moves up to 3,000 or 3,200.

My CPU has to have a package temperature of 70C before my AP-30 moves into a range of 4,000 RPM. It's never otherwise going to spin that fast. It's not likely to exceed 50C with a certain level of gaming activity.

 

[Billb2]

It's the fan(s) that make the noise. And the simple fact is that the more air they push through the radiator the more heat is removed - and the more noise they make. So, for quiet, it's a lot of rad faces and a lot of low performance, quiet fans.

 

[BonzaiDuck]

Actually, I'm conflicted. The CLU-re-lidding gets me a 12C advantage over unmolested retail-box i7-6700K.

The Kraken X62 otherwise beats my LG Macho by maybe 8C -- ceteris paribus to exclude the TIM advantage.

So I could get another 8C by swapping out the air-cooler for the Kraken, and I'm pretty much all set with the case preparation.

But since I don't intend to volt this processor more than the setting I currently have, the temperature range is currently acceptable for running LinX and Prime95. Getting another 8C temperature improvement won't make any difference. What to do? What to do?

 

[StefanR5R]

Yes I am familiar with the fact that more heat = more rpm = more noise.

It also makes me wonder about 140mm rad vs 360...surely the single 140 is more quiet than 3x120...

No, it is the other way around. More fans spinning slower, and slower moving air, give less noise than fewer fans spinnig faster and moving the air faster. In addition, the more radiator surface there is, the slower the air may move for same thermal performance.

From what I hear though its actually the pump that makes more noise than the fan does.

If the pump is slowed to the least necessary RPM and reasonably decoupled from the case, it can be quite enough. It then depends on the amount of radiator surface in relation to power consumption of the parts to cool whether the fans and air stream can be kept equally quiet.

As for the question of cooling CPU, or GPU, or both:

If noise is a concern, all it takes to cool a 120 W CPU extremely quietly is a big tower cooler and a reasonable flow path.

But to cool a 120 W GPU quietly, either a big aftermarket air cooler with at least 120 mm fans is required (IMO, from my experience with a GTX 1070 with one of the best factory coolers there is, the triple-slot cooler of Gainward and Palit branded cards) -- or a proper water cooling setup.

(Less than 120 W of the respective components would mean that air cooling will be quieter than watercooling if done right.)

 

[BonzaiDuck]

No, it is the other way around. More fans spinning slower, and slower moving air, give less noise than fewer fans spinnig faster and moving the air faster. In addition, the more radiator surface there is, the slower the air may move for same thermal performance.



If the pump is slowed to the least necessary RPM and reasonably decoupled from the case, it can be quite enough. It then depends on the amount of radiator surface in relation to power consumption of the parts to cool whether the fans and air stream can be kept equally quiet.

As for the question of cooling CPU, or GPU, or both:

If noise is a concern, all it takes to cool a 120 W CPU extremely quietly is a big tower cooler and a reasonable flow path.

But to cool a 120 W GPU quietly, either a big aftermarket air cooler with at least 120 mm fans is required (IMO, from my experience with a GTX 1070 with one of the best factory coolers there is, the triple-slot cooler of Gainward and Palit branded cards) -- or a proper water cooling setup.

(Less than 120 W of the respective components would mean that air cooling will be quieter than watercooling if done right.)

Just a casual thought about this. I can clock my own GTX 1070 to 2,025-core/4,392-memory. I hadn't thought about the wattage, but so far the temperatures remain in the low to mid-60s C. I'm more fond of simulators than games, so my choices are fairly mild, pushing the CPU usage to around 45% and graphics to a mere mid-80s%. I'm still running with 1920x1080 resolution, until I sort out other priorities to accommodate a new 2560x1440 monitor .

This is "phase-2" of my Skylake system, expecting to double my RAM and add a 2TB MX300. If I were to add a second GTX 1070, I would guess that the top card and "master" unit would heat up more by maybe 10C. Since I'm covetous of usable PCIE slots, I'm not eager to get a card with a triple-slot cooler. I'm really wondering if I need to pay any more attention to graphics. The addition of a second 1070 would provide marginal gains.

On the CPU end, the monitoring software shows package wattage under stress load at close to 150W. That's when you can hear my iPPC 3000 exhaust fan behind the air-cooler. It's not really that loud, and I can attend to acoustic improvement if I choose. The system is quiet in conjunction with this one fan during real-world use.

Now . . . when it comes to a Skylake-X or Broadwell-E or any hex-or-more-core choices, I don't think I'd bother with an air-cooler, but some have, and some do . . .

 

[StefanR5R]

It's all very much subjective of course. Also, when I stress my graphics cards, it is not with game workload but with computing applications, i.e. a rather constant load and possibly entire days in a row.

Dual GPU setups cannot be quiet if air-cooled, due to very much restricted space for air flow, not to mention the doubled heat output.

when it comes to a Skylake-X or Broadwell-E or any hex-or-more-core choices, I don't think I'd bother with an air-cooler, but some have, and some do . . .

Noctua NH-D15S cool my dual Xeon E5-2690v4 super quiet.

 

[BonzaiDuck]

It's all very much subjective of course. Also, when I stress my graphics cards, it is not with game workload but with computing applications, i.e. a rather constant load and possibly entire days in a row.

Dual GPU setups cannot be quiet if air-cooled, due to very much restricted space for air flow, not to mention the doubled heat output.


Noctua NH-D15S cool my dual Xeon E5-2690v4 super quiet.

You know -- I'm only a spectator when it comes to the Xeon processor line, but I'm aware that they can be had from various sources and come with mucho cores. But I just looked at the Intel site's spec sheet on that processor.

Did you really pay that much for it -- something north of $2,000? Interesting, no? That it has a TCASE spec of 89C? That's 20C above the desktop/workstation "K" chips. You'd think it would be lower, for the number of cores on that puppy.

If I spent that much on the processor, I'd be willing to spend more on cooling. But if you could cool it to your satisfaction with a D15, D15S, Dark Rock, LG Macho or other upper-tier heatpipe, you'd have conviction for saying "Why spend the money? Why make things more complex than need be?"

And ALSO -- you say "dual Xeon." How would you find space on the motherboard for two D15S's? And only because I know what I read and nothing else -- we're talking about four kilobucks, aren't we?

You're really pulling my leg on this, aren't you? . . .

 

[StefanR5R]

Did you really pay that much for it -- something north of $2,000?

This is indeed close to the retail price which I paid less than a year ago in Europe. They were required for the job. Used Xeons can be had for much better prices, but at the time (and still today) I don't know my way around the market for used computer equipment, and something reliable was needed.

I hear OEM prices tend to be considerably lower.

That it has a TCASE spec of 89C? That's 20C above the desktop/workstation "K" chips. You'd think it would be lower, for the number of cores on that puppy.

My understanding is that desktop chips of the same generation require lower TCASE due to their higher frequency.
Since I put the Xeons into a midi tower case with those large dual-tower coolers, not into a rack with low-profile coolers, actual temperatures are far below that limit.

Spoiler: sensor data under moderate all-core workload, at all-core turbo

Room temperature is above 25°C, I might want to open the window now.

$ cat /sys/devices/system/cpu/cpufreq/policy*/scaling_cur_freq | uniq
3199929
$ sensors
radeon-pci-0200
Adapter: PCI adapter
temp1:        +38.0°C  (crit = +120.0°C, hyst = +90.0°C)

coretemp-isa-0000
Adapter: ISA adapter
Physical id 0:  +62.0°C  (high = +93.0°C, crit = +103.0°C)
Core 0:         +58.0°C  (high = +93.0°C, crit = +103.0°C)
Core 1:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 2:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 3:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 4:         +60.0°C  (high = +93.0°C, crit = +103.0°C)
Core 5:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 6:         +62.0°C  (high = +93.0°C, crit = +103.0°C)
Core 8:         +60.0°C  (high = +93.0°C, crit = +103.0°C)
Core 9:         +58.0°C  (high = +93.0°C, crit = +103.0°C)
Core 10:        +57.0°C  (high = +93.0°C, crit = +103.0°C)
Core 11:        +62.0°C  (high = +93.0°C, crit = +103.0°C)
Core 12:        +62.0°C  (high = +93.0°C, crit = +103.0°C)
Core 13:        +62.0°C  (high = +93.0°C, crit = +103.0°C)
Core 14:        +60.0°C  (high = +93.0°C, crit = +103.0°C)

coretemp-isa-0001
Adapter: ISA adapter
Physical id 1:  +65.0°C  (high = +93.0°C, crit = +103.0°C)
Core 0:         +60.0°C  (high = +93.0°C, crit = +103.0°C)
Core 1:         +62.0°C  (high = +93.0°C, crit = +103.0°C)
Core 2:         +64.0°C  (high = +93.0°C, crit = +103.0°C)
Core 3:         +63.0°C  (high = +93.0°C, crit = +103.0°C)
Core 4:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 5:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 6:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 8:         +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 9:         +60.0°C  (high = +93.0°C, crit = +103.0°C)
Core 10:        +61.0°C  (high = +93.0°C, crit = +103.0°C)
Core 11:        +63.0°C  (high = +93.0°C, crit = +103.0°C)
Core 12:        +64.0°C  (high = +93.0°C, crit = +103.0°C)
Core 13:        +63.0°C  (high = +93.0°C, crit = +103.0°C)
Core 14:        +61.0°C  (high = +93.0°C, crit = +103.0°C)

nct7904-i2c-9-2d
Adapter: SMBus I801 adapter at 0580
in1:          +0.93 V
in2:          +1.05 V
in3:          +1.48 V
in4:          +1.22 V
in5:          +1.23 V
in6:          +0.72 V
in7:          +1.79 V
in8:          +0.63 V
in9:          +1.79 V
in10:         +1.79 V
in11:         +1.20 V
in12:         +1.20 V
in13:         +1.20 V
in14:         +1.20 V
in15:         +3.22 V
in16:         +3.15 V
in20:         +3.36 V
fan1:           0 RPM
fan2:         804 RPM
fan3:         808 RPM
fan4:         858 RPM
fan5:         765 RPM
fan6:         876 RPM
fan7:           0 RPM
fan8:        1202 RPM
temp1:        +37.9°C
temp2:        +61.8°C
temp3:        +64.0°C

Fans 4 and 6 are the NF-A15 of the processor coolers. Fans 2 and 3 are NF-A14 PWM as intake at the case front. Fan 5 is a NF-F12 crammed at the case back as another intake, probably somewhat ineffective but with the aim to press some air towards the VRMs beneath the cooler of CPU2 at the back. Fan 8 is a NF-A9x14 PWM side fan for the PCIe cards.

The case has a meshed top side as exhaust. CPU1 (physical ID 0) sits towards the front and is therefore a little bit cooler than CPU 2 (physical ID 1), but not much.

The RPMs shown above, combined with the low restrictive case and the sound profile of Noctua fans, are just about at the limit what I personally tolerate in a moderately quiet room.

I am unclear about where the temp1 sensor is located; it may be a board sensor or it may be the PCH's sensor. I am also unclear which of the voltages represent VCORE.

If I spent that much on the processor, I'd be willing to spend more on cooling. But if you could cool it to your satisfaction with a D15, D15S, Dark Rock, LG Macho or other upper-tier heatpipe, you'd have conviction for saying "Why spend the money? Why make things more complex than need be?"

Cooling requirements of socket 2011(-3) Xeons are actually modest if compared to overclocked socket 2011(-3) desktop chips, but of course one or more notches above mainstream desktop chips, depending on SKU.

My home desktop had been a 65 W Phenom II under semi-passive Scythe Ninja for a long time; now it is a 84 W Haswell Xeon E3 under Thermalright Macho, with which I am also very happy with.

I have a single watercooled system, X99 platform, at which I slowly learn how to balance the various components.

And ALSO -- you say "dual Xeon." How would you find space on the motherboard for two D15S's?

They fit just so on a Supermicro X10DAX. I measured before I bought.

You're really pulling my leg on this, aren't you? . . .

No. It is pretty much run-of-the-mill work equipment, only a little bit unusual maybe as it is DIY from parts, rather than built by a boutique or from a server/workstation OEM.

 

[Hans Gruber]

I am quite new to water cooling myself. I have a 240mm AIO running my 3570k @4.5ghz and a 280mm AIO still in the box for my next build. In the past I have always used 120mm air coolers. I still have the thermalright Ultra 120 and recently bought ($12) an upgrade kit for it to use on 115x motherboards. I have two ultra 120's circa 2007/08, one in use on a G4560 (max temps in low 40's).

The Coolermaster 212+ is based on the original Thermalright Ultra 120 and most of the current 120mm air coolers based on that design as well. They work well for OCing CPU's.

In my opinion custom loops are cool but not cost effective. Something a person who wants to have a custom loop builds for the sake of saying they have it or they did it. Not the most reliable parts for cooling solutions.

I personally think any 240mm AIO cooler is superior to any of the Air coolers. I am not a big fan of Noctura air cooling simply for the cost of the cooling solution. Based on my extensive experience with air cooling, I have found AIO liquid cooling to be superior. My 3570K @4.5ghz has never touched 60C on any core. With air cooling my temps on a dust free Xigmatek Gaia 120mm air cooler were mid 60's to high 60's. All with Arctic Silver 5 thermal paste. My AIO 240mm liquid cooler (3570K @4.5ghz) typically has temps at 50C or slightly below with a few cores in the low 50's. during heavy gameplay of BF1. I use Core Temp for my readings and the highest a core has hit on my 3570K was 59C. Typically the hot core is 53-56C in heavy gaming.

One of the arguments about water cooling is that the best air coolers can catch up to AIO air coolers after marathon gaming sessions 4 hours or more. I suspect that could be the case with 120mm air coolers but the 240mm air cooler seems to efficiently pump out the warm liquid to the radiator. In my liquid cooler it seems whenever the temps appear to get high mid 50's the liquid cooler pumps out the warm liquid filtering in much cooler liquid from the radiator.

If a person doesn't OC, there is no reason for liquid cooling solutions unless you are doing it for the cool factor. The useful life of a GPU is shorter than the life of a CPU. After a GPU is retired from a gaming system, you would have to reinstall the air cooling heatsink and fans. Problems typically occur in this process. Again all based on the skill of the computer builder. Taking off the good air cooling of GPU's could result in damage to the GPU, it happens. Typically those who do liquid cooling with CPU and GPU do it because they have a custom loop in their system.

When I installed my 240mm AIO cooler. I crafted a back plate for my GTX 970 to catch any leaks that may occur.

 

[StefanR5R]

If a person doesn't OC, there is no reason for liquid cooling solutions unless you are doing it for the cool factor.

There is another corner case for custom loops: Multi GPU systems will be loud if air cooled, but can be very quiet if cooled by a large enough custom loop. Granted, multi GPU systems are rare these days. (I have one for number crunching.)

 

[Hans Gruber]

I looked into building a system with a custom loop for both my CPU and GPU. The cool factor was definitely there but the price of admission was between $300 and $400. I could not justify the cost. I still enjoy looking and custom loop systems.

 

[bigboxes]

It's all very much subjective of course. Also, when I stress my graphics cards, it is not with game workload but with computing applications, i.e. a rather constant load and possibly entire days in a row.

Dual GPU setups cannot be quiet if air-cooled, due to very much restricted space for air flow, not to mention the doubled heat output.


Noctua NH-D15S cool my dual Xeon E5-2690v4 super quiet.

I'm planning out a dual Xeon build for the next year. You got any pics of those coolers clearing the ram? My research shows that fan barely clears a plain memory module without a heatsink. I'd prefer to go air for this build and was thinking about using a pair of NH-D15's.

 

[StefanR5R]

I have got NH-D15S with only one fan in the middle, hence no RAM conflict.
(The symmetric NH-D15 wouldn't fit on my board anyway, at least not the way I want them: Blowing upwards in parallel, not blowing front-to-back in series.)

 

[bigboxes]

I have got NH-D15S with only one fan in the middle, hence no RAM conflict.
(The symmetric NH-D15 wouldn't fit on my board anyway, at least not the way I want them: Blowing upwards in parallel, not blowing front-to-back in series.)

I'd still like to see a pic of your setup, if you get a chance.