Question About Mounting Water AiO Cooler

[flexy]

I have a Fractal Design Arc Midi R2 with currently one 140mm back, one 140mm side, one 140mm intake and one 120mm intake on the bottom.

I am still using an old-school Thermalright TRUE on my i4770K.

I am having an eye on the Coolermaster Seidon 120V R2 which seems to be a decent AiO for a low price.

My question is in regards to the mounting options I'd have with my existing fans.

* Mounting the Radiator behind the existing 140mm exhaust so it's a push-pull with the existing and the 120mm of the Coolermaster.

* Replacing the existing 140mm exhaust at the back with the 120mm of the Seidon?

* Putting the radiatior and the 120mm UP "at the ceiling" of the case which would probably be the best option BUT I have several concerns there and dont necessarily want to blow up hot air UP and out the case, besides I use the PC to put all kinds of stuff on top of it and my cats also sit there often. (Otherwise I would mount the radiator "on the ceiling" and wouldn't need to think twice there and can keep everything else).

My main concern with all those options is that the hot air from the GTX970 still needs to be exhausted from the case. One concern I have is that with the AiO I basically must "block" the existing 140mm back exhaust with the radiator...wouldn't this ultimately lead to less hot air being able to get out of the case as compared to the 140mm? Or would it ultimately be the same (or better even) in terms of case temps if I have a push pull with the existing back exhaust fan?

** Edit: Now in another thread, someone is AGAIN mentioning the "EVGA Superclock" aircooler which is basically a black TRUE. And this one beats the Seidon....makes me think whether just to keep my ancient TRUE instead...

 

[monkeydelmagico]

* Replacing the existing 140mm exhaust at the back with the 120mm of the Seidon? * ..

This

You will also find your choices to be somewhat dictated by the rather stiff tubing on the seidon series coolers. I tried both top and back mounting for mine and found back/rear to be most effective.

 

[XavierMace]

** Edit: Now in another thread, someone is AGAIN mentioning the "EVGA Superclock" aircooler which is basically a black TRUE. And this one beats the Seidon....makes me think whether just to keep my ancient TRUE instead...

Since this is a rare occasion where I'm getting to reply before he does....

Here's what you have to understand with some of the data you see on here. A lot of people spend a lot of time tweaking their air setup. Obviously there's nothing wrong with that. But after they spend all their time tweaking, trying to new fan, etc. They finally reach a point where they can match or slightly exceed some AIO water cooling setups. Here's the catch. First off, there's no data on how that AIO was installed. That makes a difference. Secondly, there's nothing stopping you from doing the same tweaks to the AIO. Personally, I haven't seen an air cooler that can match a decent AIO when all else is equal (meaning either no tweaks or both tweaked).

Radiators perform better when used as an intake. Period. There's a reason nearly all cars have their radiator on the very front of the car. How big of a deal does that make you ask? In my testing I've seen a 7c idle and 15c load increase moving the radiator from an intake to exhaust layout on the same hardware. Even with that increase though, it's still well below stock temps even overclocked. My 3770k @ 4.2Ghz, idles at 31c with radiator in a top exhaust position. The same rig idle at 25c with the same radiator mounted as a front intake. Some cases unfortunately limit your options on that. I found this picture of your case though showing a radiator mounted on the front:

I don't know if relocating the drive bay required any custom work though of if it's just a matter of some screws (Fractal's website gives me the impression it's the latter). The other limitation is the hoses and this brings me to why I would recommend against your choice of AIO. That setup has the same (or similar at least) hose setup as the old H100's. They are short, inflexible hoses. You pretty much have to mount the radiator right next to the processor. The newer model Corsair and NZXT's have much more flexible tubes. The NZXT's also have much longer tubing than my Corsair H100. I do not know if the newer Corsair's (H90/H110) have longer tubing or not.

That brings me to the second reason for recommending against the setup you recommended. Bigger is better when it comes to radiators. Looking at the specs for your case, you can mount a 240mm radiator either on the front or top (per Fractal). I'd rather have a 240mm setup as exhaust than a 120mm setup as intake. So, IF you do go AIO liquid, I'd strongly recommend a Corsair H100i. It's pricier but you're going to get much better performance.

IF you are set on a 120mm setup and IF you are set on using it in an exhaust setup, then with stock fans I doubt it's going to be much if any of an improvement over what you have now.

 

[BonzaiDuck]

[I think he may mean "me" when he says "he:"

Since this is a rare occasion where I'm getting to reply before he does....

Here's what you have to understand with some of the data you see on here. A lot of people spend a lot of time tweaking their air setup. Obviously there's nothing wrong with that. But after they spend all their time tweaking, trying to new fan, etc. They finally reach a point where they can match or slightly exceed some AIO water cooling setups. Here's the catch. First off, there's no data on how that AIO was installed. That makes a difference. Secondly, there's nothing stopping you from doing the same tweaks to the AIO. Personally, I haven't seen an air cooler that can match a decent AIO when all else is equal (meaning either no tweaks or both tweaked).

Yup. Absolutely -- Positively. But in those comparison reviews, they would've installed the AiOs without any "tweaks." It might be interesting to review again whether they were deployed as intake or exhaust. But with a lot of cases, it would seem a lot easier for the former -- given factors you mention.

Radiators perform better when used as an intake. Period.

Could we somehow approach an asymptotic coincidence of agreement that this statement is most certain when the radiator performance is taken "by itself?" With the radiator at intake, the radiator will perform some degrees-C better, but you're now pushing warmer air into the case. As exhaust, you've got warmer air pushing through the radiator. And these differences are going to be exacerbated noticeably if the graphics card(s) aren't included in any water-loop. In the exhaust case, pushing more CFM through the entire case will narrow the gap. In the intake case, you'd either want the other components included in the loop, or apply some isolated "intake-to-exhaust" solution.

For the overall case-environment, I'm not sure how these trade-offs cancel or overwhelm each other. It seems that a lot of water-based systems are attempted with lower overall case airflow -- for silence, appearance and other reasons.

There's a reason nearly all cars have their radiator on the very front of the car. How big of a deal does that make you ask? In my testing I've seen a 7c idle and 15c load increase moving the radiator from an intake to exhaust layout on the same hardware. Even with that increase though, it's still well below stock temps even overclocked.

Again, no real disagreement, except -- as I asserted before -- I think the gap can be narrowed. Either the OP or others may not particularly like the means of narrowing it, though. And also again, I'm still somewhat impressed with what WGusler did with an H80i at exhaust in a C70 case.

My 3770k @ 4.2Ghz, idles at 31c with radiator in a top exhaust position. The same rig idle at 25c with the same radiator mounted as a front intake. Some cases unfortunately limit your options on that.

But that's what you'd expect at idle with radiator at intake. Like I said already, there are trade-offs with the exhaust strategy and I'm not exactly sure how they play out, but there's no guarantee that the load temperatures would show a linear increase matching the 6C intake difference you cite.

But instead of initiating a nitpicking p***ing contest (again), I totally agree with all that follows below:

I don't know if relocating the drive bay required any custom work though of if it's just a matter of some screws (Fractal's website gives me the impression it's the latter). [BD : Probably would.] The other limitation is the hoses and this brings me to why I would recommend against your choice of AIO. [BD : Abso-freakin'-lutely!] That setup has the same (or similar at least) hose setup as the old H100's. They are short, inflexible hoses. You pretty much have to mount the radiator right next to the processor. The newer model Corsair and NZXT's have much more flexible tubes. The NZXT's also have much longer tubing than my Corsair H100. I do not know if the newer Corsair's (H90/H110) have longer tubing or not.

That brings me to the second reason for recommending against the setup you recommended. Bigger is better when it comes to radiators. [BD : The basic, ineluctable imperative] Looking at the specs for your case, you can mount a 240mm radiator either on the front or top (per Fractal). I'd rather have a 240mm setup as exhaust than a 120mm setup as intake. So, IF you do go AIO liquid, I'd strongly recommend a Corsair H100i. It's pricier but you're going to get much better performance.

IF you are set on a 120mm setup and IF you are set on using it in an exhaust setup, then with stock fans I doubt it's going to be much if any of an improvement over what you have now.

I guess the only departure from any of this is the assessment of the TRUE cooler. I had more than one of those. That's been eclipsed by later models of various manufacture -- I needn't sing praises for the "Superclock" or whatever it's currently called.

Personally, I'm trying to sort through all of these factors and considerations to come up with my own water-loop for a system yet-to-be-built. I keep postponing it. I'm not so inclined to include graphics card(s) in the loop, yet that would be the best answer to all of this . . . At some expense, additional complexity -- and I keep postponing it.

UPDATE: Actually, I'll revise just a tad what I said already. He could "try" WGusler's solution, but it involves more than just installing a "square" AiO at the exhaust. And rather than exchanging a 140mm (stock?) exhaust fan with a 120, I'd pick (at least!) a "square" 140mm AiO -- if such exists. And true to my own craziness, I'd pick beefier fan replacements. But squeezing some extra C's from a square AiO is going to involve as much tedium or tweaking as my own eccentric air solutions.

He's better off -- as Xavier says -- with a dual-fan AiO -- better, a dual 140 AiO, but he has to fit it into that case -- intake or exhaust -- whatever. Or he could simply scrap the AiO idea, get a "thicker" square radiator, a custom-choice of pump(s) and reservoir, and spend mo' money . . .

 

[Essence_of_War]

My 3770k @ 4.2Ghz, idles at 31c with radiator in a top exhaust position. The same rig idle at 25c with the same radiator mounted as a front intake. Some cases unfortunately limit your options on that.

Both of those measurements are so far away from Tjmax that it isn't reasonable to quote numbers to two sig figs. Error bars on that sensor measurement get larger the farther you are from Tjmax.

 

[BonzaiDuck]

Both of those measurements are so far away from Tjmax that it isn't reasonable to quote numbers to two sig figs. Error bars on that sensor measurement get larger the farther you are from Tjmax.

Also quite true. I can't even add anything to your observation there . . .

I've personally toyed with the thought of trying WGusler's H80+C70+4790K solution. I thought his load temperatures were quite good compared to what I'd seen in a number or configurations. But it involves some "modding" and "tweaking."

Another thought I've had: if it weren't for the Fractal Designs case selection, there must be plenty of midtower cases that have enough 5.25" drive bay space that would allow a "straight-through" fan and square AiO cooling solution with a couple TR accordion ducts or similar. With SSDs and large HDD-capacities these days, one doesn't need all those drive bays. And that would make Xavier's point somewhat moot: the cooler would be placed in a combined intake and exhaust strategy.

Given such a choice, you'd want at least a single side-panel or lower-front intake with a top-panel exhaust -- for the rest of the components and including the graphics card. But you could get by pretty well with just four fans, I think . . .

On the "straight-through" mod -- I'd mentioned in recent threads a "quickie" build to replace my brother's E8600 system. We used a spare case purchased in the $50 range maybe 10 years ago: A CM Centurion. Bro only needed the optical disk (installed in the top 5.25" bay) and the lower internal bays for 2.5" and 3.5" drives. So I stuck a 140mm Akasa Viper in the empty 5.25" bays, replaced the aluminum bay-covers with vented ones from another case. For that project, the priority was to add more intake to the bottom-front-panel 120mm fan. But the possibilities remain.

And by the way. For bay covers and no less for "securing" fans in such places, this stuff is wonderful:

http://www.amazon.com/Permatex-25831-6PK-Choice-Adhesive-Sealant/dp/B001RERDHG

But it's becoming less and less available. Last time . . . I had to order it . . . .from a motorcycle shop in Boston.

 

[Micrornd]

Radiators perform better when used as an intake. Period.

I very much agree, my testing confirms that also.

This is the result of my testing (originally tested 11/27-31/2013 and retested 3/26/2015).
Everyone's testing WILL vary, but hopefully, this will inspire some others to actually experiment, rather than simply quote the "status quo" as they've read it, rather than actually tested themselves.

My current rendering rig (2xE5-2696 on Gigabyte GA-7PESH3 MB) uses 2 H110 AIO coolers side by side, in a custom polycarbonate frame with a single constant speed 300mm fan above and a single constant speed 300mm fan below (push-pull), spaced 1" from each rad core and designed so that all airflow into the case must go through the rads.

H110s were chosen for both their size AND the fin spacing of the rad.
No air leakage was allowed between the fans and rads, and proper fan spacing from the rads was used, this is the important part.
Spacing from the rad core matters and I have found that standard "on the rad" mounting is too close.
The highest airflow/velocity is not at the face of the fan, but rather a small distance from it, as with any fan.
This distance varies based on a number of factors (pitch, # of blades, distance from reflecting surface, etc.)

My case is an early server case from the late '90's, same line as this one -
http://www.xlr8yourmac.com/systems/SonofFrankenMac/son_of_frankenmac.html
Mine is 2 - 5.25" bays shorter, being listed as an 11u case on casters, but is also rack mountable.
The case has 2 sections/halves, 1 side for HDDs and the other for the MB.
My custom made rad "box" mounts on top of the case with the 2 H110 rads side-by-side, running across the case.
The top of the case is cut out under the bottom (pull) fan in a 300mm circle that spans both sides/halves equally.

The 300mm fans were selected for their airflow (125cfm) and noise level (22dB-A), not for SP.
SP actually needed is highly dependent on the rad used and way over emphasized in most cases, mostly because no one actually tests to see what will actually work adequately.
The absolute ultimate, perfect, lowest temp cooling setup is only needed by bench queens and not 99.99% of the real world.
Practical has it's place, as does quiet.

Airflow is from the room, above the PC, through the rads, into the case and then out the rear of the case by 2 Cougar Vortex PWM (70.5cfm, 17dB-A) (1 at the rear of each half of the case) and also out the MB side panel by 2 constant speed 220mm fans (110cfm, 21 dB-A).

The 2 - 220mm side fans were not original to the build as there was no problem cooling the CPUs, but were added when a heat build up was noted in the MB VRMs when running a simultaneous load test using IntelBurnTest and GPU-z to test the cooling of the 2 MSI 7970 Lightnings in this build.
As you know, Lightings exhaust the majority of their airflow into the case, rather than out the back and this was causing turbulence, disrupting airflow over the MB VRMs.
The VRM heatsinks are oriented on the MB as to require vertical air flow in a pedestal case.
Adding the side fans solved the turbulence problem, and allowed the the 7970s to not exceed 64c/72c, with secondary card being the higher temp (and indicating the need to check the secondary's OEM TIM install, which has still to be done)

And since airflow is into the case, a 300mm round DEMCifilter screen is used on the rad intake grille to keep the rads clean.
It only requires cleaning once a month and was sourced directly from the manufacturer.

In testing in the 2013 session, my dual E5-2696's (12 cores + 12 Hyperthreaded cores each, total 48 cores) never exceed 52c after 8 hrs (on the highest temp real core), fully loaded (using AIDA 64 (Finalwire)) in a room that started at 26.6c (80F) ambient and finished at 28.9c (84F).
That's 23-26c above ambient, as measured 12" above and to the side of the Rad intake, but not in it's intake air stream so as to keep the digital probe in a still area.
(Room temp itself is controlled by central A/C with the thermostat being located 2 rooms away and set at 26c (79F), so testing (increased room temp) did not start/stop the A/C unit, as the door to the room was closed during testing.)

A second testing session (during the 2013 session), running IntelBurnTest 2.54(LinPack64) (set on maximum stress 100 runs) for 8hrs did raise the temps slightly higher to 56c and room temp to 30c, but that's still only 26c above room temp.
The IntelBurnTest app was able to maitain all 24 cores at 100% and TDP at 118w-119w for each cpu, just under the 120w TDP Intel lists.
(As a side note, if you run the IntelBurnTest app in conjuction with other apps, it should be started first and ended last, BUT ALWAYS check the task manager to make sure it has shut down.
It has a nasty habit of closing the GUI, but leaving linpack64 still running !!)

And please note, these are 8 hour tests, allowing the entire PC and all it's components to totally heat soak and show any flaws, rather than some short 10-60min. test.

Also note that when full load testing of the dual 7970 Lightnings was done simultaneously with cpu testing (again for 8 hrs.), it had no effect on the cpu max temps, although it did raise the room temp an additional 3c.

All above tests were duplicated (during the 11/2013 testing) with all fans reversed, thereby exhausting all air through the rads.
This was a pain in the a**, as rads as an intake was done first, then all fans reversed to accommodate rads as an exhaust, and then reversing the fans again to return to rads as an intake after finding out that rads as an exhaust raised CPU temps 6-8c (the Lightning's temps stayed the same) in the 8hr tests

8hr tests were rerun in 3/2015 but only using the rads as an intake, which is how the PC has been running since 2013.
All temps for CPUs and the Lightnings were within 1-2c of the original 2013 session, which I consider "margin of error".

Additionally, the original tests (11/2013) were with AS5 as the TIM, and duplicated with Indigo Extreme as the TIM on the cpus.
Indigo Extreme was 3-4c lower in all tests.
(Perfect contact with Indigo Extreme is relatively easy to get with a 2011 socket, as long as the MB is level during the heating/curing process.)
The tests during the 3/2015 session were using only AS5 as the TIM (and naturally only compared to the 2013 tests that used AS5).

Since this was not an extreme cooling case, I decided AS5 works well enough, especially since the expensive Indigo Extreme is destroyed each time the CPU cooler is lifted.

The unit sits about 24" from my right knee and is for all practical purposes silent, only becoming slightly audible when the Lightnings take a heavy load.
Since no attempt at sound deadening was made, some sound absorption material in the case would probably take care of that.

Photos - http://s172.photobucket.com/user/Micrornd/library/Render Rig?sort=2&page=1

Note that the outside of the case is as-received 10+ years ago.
Please do not upset the flamingo, he's a mean drunk

 

[Micrornd]

Both of those measurements are so far away from Tjmax that it isn't reasonable to quote numbers to two sig figs. Error bars on that sensor measurement get larger the farther you are from Tjmax.

While that is a valid point, it is what everyone uses for comparisons, because there isn't another way to read core temps.

My gun can only read surface temps, not core temps.

What would you suggest be used?

 

[BonzaiDuck]

. . . but . . . you hardly need to test it -- that radiators will perform better at intake . . . it's just logically deducible.

Micrornd's test involves a huge case, and there are two very large fans in push-pull. The rest doesn't seem entirely clear to me.

This is all very different from a situation with a small mid-tower case (or even a mid-mid-tower). The only point I make about it: it would all likely be a matter of compromise: the case and cooler itself might constrain placement.

Suppose you have a case with overall airflow of 50 CFM -- all going through the radiator(s) at EITHER intake or exhaust? Suppose, with the exhaust placement, you increase overall airflow to 100 CFM? How does the DIFFERENCE between the intake and exhaust strategies change?

It would only be a smaller difference.

I think a lot of AiO solutions applied in this forum used an exhaust/radiator strategy, only because of the cooler, hose-length and case-design constraints.

Either way, the most significant improvement would likely be an application of larger radiators in a smaller case. And that presents a dilemma which is obvious.

 

[XavierMace]

Yup. Absolutely -- Positively. But in those comparison reviews, they would've installed the AiOs without any "tweaks." It might be interesting to review again whether they were deployed as intake or exhaust. But with a lot of cases, it would seem a lot easier for the former -- given factors you mention.

Could we somehow approach an asymptotic coincidence of agreement that this statement is most certain when the radiator performance is taken "by itself?" With the radiator at intake, the radiator will perform some degrees-C better, but you're now pushing warmer air into the case. As exhaust, you've got warmer air pushing through the radiator. And these differences are going to be exacerbated noticeably if the graphics card(s) aren't included in any water-loop. In the exhaust case, pushing more CFM through the entire case will narrow the gap. In the intake case, you'd either want the other components included in the loop, or apply some isolated "intake-to-exhaust" solution.

But that's what you'd expect at idle with radiator at intake. Like I said already, there are trade-offs with the exhaust strategy and I'm not exactly sure how they play out, but there's no guarantee that the load temperatures would show a linear increase matching the 6C intake difference you cite.

Indeed. It was an assumption, but I thought a safe one. Just some quick responses to the bolded parts.

These points are why I don't like the majority of reviews of AIO's. There's other factors beyond just the mounting position of the radiator. The newer, higher end AIO's are including fans more suited to radiator duty but the older/cheaper one's included just basic case fans. Unfortunately fans that may be OK case fans aren't necessarily good radiator fans. That's why Corsair now sells AF and SP product lines.

Radiators are also affected by ambient temperature. Again, cars (namely modified cars) provide a good analogy. In winter the car handles everything they throw at it, but in summer it overheats. It's the same car, at the same performance level, but the radiator can't cope with the higher intake temps. The solution is a better radiator (either larger or better materials).

Regardless where you mount your radiator, it's not sucking 100% of the heat from the CPU (assuming that's what it's cooling). The CPU is still affecting the case temperature. But the cooler the CPU, the cooler the air around it and the cooler the air being exhausted out the radiator. At 25-30c, that's less than the idle temp of the GPU so it's a negligible affect on the GPU. More importantly, it still cooler air than a traditional heatsink would be dispersing.

But, lastly, having a radiator doesn't automatically mean you have to have less airflow. It just means you may have to give it more thought and buy appropriate components (fans and case). Just like you would probably tend to avoid a case with all 80mm fan mounts when trying to build a good air cooled setup, I would try to avoid a case that prevents me from mounting a radiator like I would like. Obviously both of these "restrictions" limit your choices to varying degrees. This does not mean you NEED to have a huge case though. Right now, I've got 4 systems with AIO setups in the house.

System 1. NZXT H440 with 3770K @ 4.4Ghz, GTX970, and Kraken X61. Currently it has 2x 140mm NZXT intake fans on the front, 1x 120mm NZXT intake on the rear and 2x Noctua iPPC 3000rpm 140mm on the top mounted radiator as exhaust. The 140's are all controlled by the X61 which runs in Silent mode. Right now I show 28c CPU, 33c GPU, 26c Liquid. 2520rpm on the pump, 1020rpm on the fans. Case sits on the floor within arms reach. With nothing else running in the room, all I hear is a very slight hum. The PS3 makes far more noise at any of it's fan level. I just woke up so my room hasn't warmed up yet, so infrared thermometer shows 17.5c at the intake and 19.5c at the exhaust.

System 2. Corsair Carbide 300R with Xeon X5650, GTX460, and Corsair H100. This is the least optimized of the 4 systems as it's my "guest PC" so I didn't feel like putting much (or any) work into it. Currently it has 2x Corsair basic 120mm intakes on the front, 2x no-name 120mm intakes on the side, 1x 120mm intake on the rear, 4x Corsair basic 120mm exhuasts on the top mount intercooler in a push/pull setup. It sits on the desktop right next to the mouse pad. With the fan selection, I would describe it as uncomfortably loud. If I was so inclined though, that would be easy enough to solve with fans that don't suck. 43c CPU, 32c GPU. H100 doesn't show the additional info the X61 does, one of the many reasons I'm not happy with it. Thermometer again says intake 25.6c, exhaust 27.2c.

System 3. HAF EVO XB with Xeon L5640, GTX960, NZXT Kraken X31. Given the large amount of ventilation on this case, I decided to use it as an experiment. It's only got a 2x 120mm Corsair SP120 intake's on the rear mount radiator in a push/pull configuration. 22c CPU, 31c GPU. Again, no additional info in regards to pump/fan RPM's but I'd describe the computer as silent at idle. GPU fan noticeable but not terribly so under load. Thermometer shows 21.2c intake, 22.8c exhaust. However since it's not really "exhausting" from one place that's not a terribly useful data point.

System 4. Silverstone PS07 with AMD A10-5800K, Corsair H55. This is my latest experiment. It's a small midtower case and, very airflow limited. With the stock AMD cooler and stock clocks, AMD Catalyst would start giving CPU temp alerts almost immediately when gaming even with the front "door" open on the case to try to let more air in. So, I figured might as well have some fun with it. It's got a 2x 120mm Corsair SP120 intakes on the rear mount radiator in a push/pull configuration, and a 2x 120mm Corsair AF120's as exhaust on the front. Unfortunately since it's AMD rather than Intel like the rest, that will skew the data a bit. I should have some numbers on this build in the next week or so.

The point in all of that was there's more than one way to skin the cat. At first glance, you'd think system 2 would run the coolest, after all it's got almost as many fans as the other 3 systems combined. But, like a traditional air setup, fans make a huge difference. That combined with the hoses on the H100 preventing a front mount, I would describe it as the least satisfactory of the bunch. I was quite surprised by the results of system 3 given it's only got the radiator fans and it's in one of the warmest rooms of the house, sitting on the bottom of a Home Theater setup.

This brings me to your last question.

Suppose you have a case with overall airflow of 50 CFM -- all going through the radiator(s) at EITHER intake or exhaust? Suppose, with the exhaust placement, you increase overall airflow to 100 CFM? How does the DIFFERENCE between the intake and exhaust strategies change?

Why is it one or the other? This was a sticking point for me in some of our other debates (which I do enjoy). I'm not aware of any data regarding how exactly much the radiator itself is restricting CFM's. Obviously it's going to have some affect, just like a filter does but it's going to restrict regardless if it's intake or exhaust. Therefore, moving the radiator shouldn't be affecting the overall airflow CFM's. It's either restricting the incoming air lowering the CFM's to begin with, or it's restricting the outgoing air, lowering the exhuast CFM's. Either way, the result is the same.

 

[Micrornd]

.
Micrornd's test involves a huge case, and there are two very large fans in push-pull. The rest doesn't seem entirely clear to me.

The fans are huge in the sense they are 300mm, but pay attention to their cfm.
They are only 125cfm@22dbA and turn a max of 620rpm, that's 250cfm total for 2 H110s.

Each H110 comes from Corsair with 2 94cfm@35dbA at (1650rpm max)fans, that's 376cfm for 2 H110's.
So my system has 2/3rds the cfm, while still handling all the cooling of 2 120w CPUs and remains much, much quieter.
(And that 250cfm total doesn't include the cfm restriction added by the DEMCifilter, making the actual cfm lower)

As to the case, look at the pictures carefully, the case is divided into 2 halves with the side that houses the MB not much bigger than the MB itself.
There is no air intake at the front of the case (where the power and data cables are, with 2 SSDs and an I/O bay above that), so that area is about equal to where most cases have HDDs.
All air comes in through the rads.
The area that the MB is in is roughly the same size as a mid tower.

Was there anything else I wasn't clear enough on :$
I'd be happy to explain ^_^

 

[flexy]

Sorry, didnt have time to read all your replies yet.

The FD Arc Midi thankfully has a ZILLION of ways to mount radiatiors (god I love this case!!)....so you're saying putting HD cage back and mounting it in front might possibly be the best?

Also..I am aware that the 110i etc. would be better (maybe I find a used one), but in this case the Seidon just caught my attention due its very low price. I am note expecting super temps..but at least as it is as good as my current setup I would be fine with it.

 

[BonzaiDuck]

Why is it one or the other? This was a sticking point for me in some of our other debates (which I do enjoy). I'm not aware of any data regarding how exactly much the radiator itself is restricting CFM's. Obviously it's going to have some affect, just like a filter does but it's going to restrict regardless if it's intake or exhaust. Therefore, moving the radiator shouldn't be affecting the overall airflow CFM's. It's either restricting the incoming air lowering the CFM's to begin with, or it's restricting the outgoing air, lowering the exhuast CFM's. Either way, the result is the same. .

It wouldn't be "one or the other" if it were a comparison of lower CFM at either intake or exhaust to higher CFM at either intake or exhaust. But if a radiator had to be mounted at an exhaust location, and CFM through the entire case and particularly through the exhaust radiator is increased as opposed to lower CFM at either intake or exhaust, the loss of efficiency for the higher CFM exhaust radiator would narrow -- up to the point where pushing more air through the radiator wouldn't do any more to lower temperatures of the component (CPU.)

The problem, as you say, is going to be either the restriction of airflow through closely-packed radiator fins, or the noise of either air-turbulence or fans that might increase CFM at the radiator.

. . .

Was there anything else I wasn't clear enough on :$
I'd be happy to explain ^_^

If you tested cooling "effectiveness" at intake versus a different configuration at exhaust, it wasn't clear to me what the differential in CPU temperatures was. I only saw you cite a CPU temperature of 56C. Maybe I missed something.

The FD Arc Midi thankfully has a ZILLION of ways to mount radiatiors (god I love this case!!)....so you're saying putting HD cage back and mounting it in front might possibly be the best?

Also..I am aware that the 110i etc. would be better (maybe I find a used one), but in this case the Seidon just caught my attention due its very low price. I am note expecting super temps..but at least as it is as good as my current setup I would be fine with it.

It's nice to see that the Arc Midi has those fan and drive-cage options.

But you may be forgetting something -- a point of discussion and agreement between other posters.

What is the hose-length of your chosen AiO as compared roughly to a distance between the front intake fans and the square hole in the motherboard pan where the CPU will be located? If the water-block can't reach the CPU, then you may have to consider custom-water parts if you plan on mounting the radiator at the front-intake fans.

Xavier and Micrornd might have a good idea for that, or may already know that the cooler will fit. Their examples may have already proven it feasible.

. . . In other news today, Maximum PC is touting the Silverstone TD02-E. Hardware Secrets has posted a comparison review of the Silverstone showing its performance just equal to that of the H110:

http://www.hardwaresecrets.com/article/SilverStone-TD02-CPU-Cooler-Review/1810/6

They also include the EVGA "SuperClock" ("ACX") in the mix -- 8C warmer than either of the AiOs. To me, that means all my tedious efforts with the ACX just equal a stock installation of either AiO.

What I'd LIKE to see: an AiO that can offer a 12 to 16C temperature reduction over some of these air-coolers. No air-cooler will ever match that standard. But one might also "get there" with custom water parts. And I'm skeptical that anything you do to "tweak" an AiO will have much of a positive effect -- short of using beefier fans.

For instance, Max PC notes that a "thinner" radiator for the TD02-E may be a better fit for some cases. But a thicker radiator of otherwise same size would beat it.

 

[monkeydelmagico]

2 things to consider: 1) will the seidon even reach for front mount? Mine would not on a CoolerMaster N400 mid ATX. 2) When you front mount the radiator you dump the heat directly at your GPU. I don't know about you but my load profile is MUCH higher for the GPU than the CPU. A slightly higher average temp for CPU is a better trade off, for me, than a significantly higher GPU load temp.

 

[BonzaiDuck]

2 things to consider: 1) will the seidon even reach for front mount? Mine would not on a CoolerMaster N400 mid ATX. 2) When you front mount the radiator you dump the heat directly at your GPU. I don't know about you but my load profile is MUCH higher for the GPU than the CPU. A slightly higher average temp for CPU is a better trade off, for me, than a significantly higher GPU load temp.

Hey - Yo! Monkey-Man!

I jumped into this thread because of an "oblique reference" to . . . lil' ol' me.

This is what I've been saying all along, though. Nothing in this world escapes the notion of "trade-offs" and "constraints-in-optimization." I also think I spoke to this matter of hot-air on GPUs, but it would be less relevant (or not significantly at all) to water-cooled GPUs.

There are all sorts of cooling solutions: air-cooling gives only options about fans, CFMs -- nothing about locating the cooler -- that's a done-deal; Water is either AiO or custom, and either CPU and GPU, CPU-not-GPU, or even GPU-not-CPU. Personally, when I swing to H2O, I'm likely to leave the GPU(s) air-cooled. That could be "worse" for CPU and radiator-at-exhaust, but radiator-at-intake will be a "downer" for GPU cooling.

Lotta points on a map of "constrained optimums" depending on what one chooses to do! Makes you wonder . . . if you're going to "do water" -- whether you should expand the dollar-outlay, take your GPUs apart, and do the whole enchilada. And the $$$ is another constraint!

 

[XavierMace]

It wouldn't be "one or the other" if it were a comparison of lower CFM at either intake or exhaust to higher CFM at either intake or exhaust. But if a radiator had to be mounted at an exhaust location, and CFM through the entire case and particularly through the exhaust radiator is increased as opposed to lower CFM at either intake or exhaust, the loss of efficiency for the higher CFM exhaust radiator would narrow -- up to the point where pushing more air through the radiator wouldn't do any more to lower temperatures of the component (CPU.)

I don't want to side track this thread too much (like we usually do) but since every time we do our thing, this gets mentioned and I'm still not sure what you're getting at although I will say my original statement of it makes no difference in regards to CFM wasn't correct but shows a second reason why intake is superior. Unless, again, I'm not thinking this through correctly.

Let's say the radiator restricts airflow by 10CFM and you've got 80CFM fans. Those numbers are straight out of my backside, but it will do for the point I'm trying to make. At the most basic level, you have two options.

X = Radiator, O = Regular Fan, | = Case Wall

Intake -> | OX (70CFM) -> O | -> Exhaust (80CFM)
Intake -> | O (80CFM) -> OX | -> Exhaust (70CFM)

While having the radiator as an intake does technically pipe warmer air into the case, it frees up the exhaust to work at peak performance thereby removing said hot air faster and lowering overall system temps.

Agree/Disagree?

What is the hose-length of your chosen AiO as compared roughly to a distance between the front intake fans and the CPU...

I'll put money on it not reaching. Like I said, unless the OP has already purchased the kit, I would strongly recommend against it if for no other reason than the tubing (both length and material). My H100 has the same tubing and I hate it with a passion.

. . . In other news today, Maximum PC is touting the Silverstone TD02-E. Hardware Secrets has posted a comparison review of the Silverstone showing its performance just equal to that of the H110

I'd put money on it being the same Asetek pump as the H110 just with a metal casing. A good portion of the AIO kits at the various price ranges are just re-branded Asetek kits with slight differences in fans/packaging. Sometimes slight radiator changes.

They also include the EVGA "SuperClock" ("ACX") in the mix -- 8C warmer than either of the AiOs. To me, that means all my tedious efforts with the ACX just equal a stock installation of either AiO.

To play devils advocate here, they have the radiator top mounted (not mentioned but you can see it in the picture) and in a pull only configuration. I can't speak to Silverstone's manual, but page 1 of the Corsair manuals recommend an intake configuration and the pictures show the fans being installed in a push configuration.

I'd love to see them take that exact same rig and do two more tests. One with front mount push only, and one with front mount push pull. I'd put money on you getting your other 8c you want with the front push/pull.

Like I said, this is why I dislike AIO reviews. If you're going to review a product, at least follow the manufacturers instructions/recommendations when installing it.

/my .02. Or more like .25 probably

 

[BonzaiDuck]

Oh, I'd most certainly agree with that. My thoughts about the issue were not refined to that level of detail. We're going to have all sorts of restrictions to airflow in the "real-world" situation, and your "model" reflects that.

Some folks always make a point that "rated" fan CFM doesn't reflect the actual deployment; that the vents are restrictive, filters are restrictive and so on.

My response to that is to simply say that I choose how I deploy fans in my air-cooled rigs as a matter of "relative" CFM defined by the specs. If one tried to match intake and exhaust to the single CFM, it could be a disappointment of one sort or the other. And in my own air-cooled strategy, it's important to me that the case is otherwise sealed for anything other than air intake and exhaust, so the objective is to at least -- mildly -- overwhelm exhaust with intake. In the machines I've built in the last four years, the guesswork has been . . . adequate. And it's all a sort of dynamic situation given thermal fan-control: the balance is different at idle than it is under severe load conditions.

Also, the matter of "static pressure" is not incidental. There are good reasons to use 200mm fans for certain things, but if it's a matter of a restricted airflow situation, nothing is going to beat a smaller fan with good static pressure.

Your point about test conditions for AiO reviews is also probably spot-on. And it's likely going to make a difference. I was best able to compare my own setup to an H110 only because a member cooperated with passing his numbers to me for identical tests, but as I recall -- it was an H110 deployed at exhaust. On the other hand, he'd water-cooled both his CPU and GPU in two, separate loops, so that may have mitigated some of the difference, except for the fact that the same heat is being forced into the case from the GPU radiator and then blown out through the exhaust CPU radiator. Of course, if the GPU(s) remained at idle while the CPU was being stressed, it would've had less of an effect.

For that, it was interesting when I added a second gfx card to my otherwise-identical second rig -- the one fitted with the ACX cooler. IBT load temperatures rose ~4C over what they'd been.

 

[Micrornd]

If you tested cooling "effectiveness" at intake versus a different configuration at exhaust, it wasn't clear to me what the differential in CPU temperatures was. I only saw you cite a CPU temperature of 56C. Maybe I missed something.

You must have missed it.
I did not test with different fans, only those I have outlined as being in the build.
There was no need to, as other fans would only have guaranteed more noise, every thing else being equal, as well as higher cost per fan.

All above tests were duplicated (during the 11/2013 testing) with all fans reversed, thereby exhausting all air through the rads.
This was a pain in the a**, as rads as an intake was done first, then all fans reversed to accommodate rads as an exhaust, and then reversing the fans again to return to rads as an intake after finding out that rads as an exhaust raised CPU temps 6-8c (the Lightning's temps stayed the same) in the 8hr tests.

The purpose of the tests were to determine which way the air should flow through the radiators for maximum cooling effect of the entire PC.

I tested with all airflow into the case entering through the radiators and all airflow exiting out the side and rear fans.
(This is the setup I use now and was determined to be the best)

I also tested with all airflow into the case entering from the side and rear fans and exhausting all air through the radiators.

The test duration was 8hrs for each of the above tests.
These tests were conducted at 100% load on all 24 real cores and all 24 hyper-threaded cores, with all of those 48 cores running @ 3.1ghz for all 8hrs. in both of the tests.
TDP was logged as 118-119w for each of the 2 CPUs for the entire duration of both tests.
Additionally both MSI 7970 Lightnings were also loaded to 100% GPU load at the same time in each test for the full 8hrs. of the test.

The test with all airflow entering through the radiators showed 6-8c cooler temps for the CPUs (according to the logs) compared to using the test with all airflow using the radiators as an exhaust.
At the end of the 8hr. tests the differences were clear, 56c max (both cpus) using the radiators as an intake vs. 62c (cpu1) and 64c (cpu0) when using the radiators as an exhaust.
(VRM temps were within 1-2c of each other in both tests, which I consider "within margin of error", so I did not mention them. As I noted in my OP, these temps stabilized when the 2 8" side fans removed the turbulence from the case. And apparent they are unaffected by the direction of airflow.)

Interestingly the Lightnings were not affected by the way which the air entered or exhausted the case.
Their logs showed the same temps over the 8hr period in both tests, regardless of whether the radiators were used as intakes or as exhausts.
Just a testament to the good cooling Lightnings are famous for.

I also noted that using the radiators as exhaust produced more noise, probably since the airflow and noise was pointed up and started below ear level.
But since I forgot shoot it with the db meter, I didn't mention it, as it's just subjective.

 

[BonzaiDuck]

At the end of the 8hr. tests the differences were clear, 56c max (both cpus) using the radiators as an intake vs. 62c (cpu1) and 64c (cpu0) when using the radiators as an exhaust.

That's in the order of magnitude I would expect -- on the high side, but you have two, octo-core Xeons in that rig. Or is it 12 cores each? 12 -- I think you say. And nearly 250W of combined thermal wattage.

If one builds something like that, the idea of a small case profile or footprint is not going to be in the cards. I'd say that's a major accomplishment -- whether 56C or 62C -- for that kind of computing power.

 

[XavierMace]

I forgot to address the GPU side of it in the other equation. Converting your CPU to liquid cooled in my experience has very little affect on the GPU assuming the PC had a reasonable fan setup to begin with.

Converting the GPU can be a different story. I had a reference R9 290X. That POS would hit 85-90c underload, per AMD "by design, nothing to worry about". I don't care if it's designed to run that hot. I don't want that much heat in my computer. Even if the GPU is designed to exhaust outside the case, that heat still radiates into case. I converted it to liquid (Kraken X31) which lowered it's temps by an average of 50c. Overall system temps dropped by 15c. To say nothing of the fact my computer no longer sounded like an airplane at take off.

The GTX970 I have now is a different story. It's not idling much above system temps anyways and it has a far more effective heatsink. I don't feel it's worth the risk to convert it to liquid at this point.

The other issue with cooling the GPU in addition to the CPU is it adds complexity to your setup. You typically don't just take your existing setup and pipe the GPU into it. Most of the time you end up with a second radiator and second pump. I've been there, done that. Most of the time, personally, I just don't find it worth it any more.

 

[BonzaiDuck]

I forgot to address the GPU side of it in the other equation. Converting your CPU to liquid cooled in my experience has very little affect on the GPU assuming the PC had a reasonable fan setup to begin with.

Converting the GPU can be a different story. I had a reference R9 290X. That POS would hit 85-90c underload, per AMD "by design, nothing to worry about". I don't care if it's designed to run that hot. I don't want that much heat in my computer. Even if the GPU is designed to exhaust outside the case, that heat still radiates into case. I converted it to liquid (Kraken X31) which lowered it's temps by an average of 50c. Overall system temps dropped by 15c. To say nothing of the fact my computer no longer sounded like an airplane at take off.

The GTX970 I have now is a different story. It's not idling much above system temps anyways and it has a far more effective heatsink. I don't feel it's worth the risk to convert it to liquid at this point.

The other issue with cooling the GPU in addition to the CPU is it adds complexity to your setup. You typically don't just take your existing setup and pipe the GPU into it. Most of the time you end up with a second radiator and second pump. I've been there, done that. Most of the time, personally, I just don't find it worth it any more.

A parallel experience -- with the newer nVidias. I mentioned an increase of ~4C in CPU load IBT temperatures from adding a second GTX 970 (see last paragraph here). Yet, the 970's -- excluding anything other than power consumption and thermal dissipation from the discussion -- are amazingly "green" in their low-power state, and at stock core and memory speeds never exceed 70C under full Kombustor load testing with both GPU cores tweaked to run up to 98-100% usage.

So the GPU "dilemma" with these cards leads me to the same reticence about water-cooling coverage. Fact is, and I don't think I'm "deaf" -- the total of four fans with two 970s @ 50-60% duty-cycle just don't seem that noisy to me -- or actually -- not at all. I just wish I could adjust the fan-curves without changing gfx card BIOS' or using a utility like AfterBurner. The fans drop to 0% as soon as the GPU temperatures fall below 50C in what appears to be their default configuration, so it takes longer for them to cool to around 35C.

Also, I may run another IBT test on my CPU. It now occurs to me that I had tweaked the CPU fans for lower speeds after I'd reached the stable overclocks. So there may not be a "real" 4C difference, or at least part of it is due to the final CPU fan profiles.

 

[Micrornd]

That's in the order of magnitude I would expect -- on the high side, but you have two, octo-core Xeons in that rig. Or is it 12 cores each? 12 -- I think you say. And nearly 250W of combined thermal wattage.

Each Xeon has 12 real cores and also 12 Hyper threaded cores, so the rig shows up in Windows as having 48cores.

Each Xeon used 118-119w and are rated for 120w each, so about 240w for them.

The 2 MSI 7970 Lightnings are rated on most sites as consuming 300w each @ full load, so about 600w for them.

Assuming about 25w for the MB, SSDs, and fans, that's a load of 865w on the PS.
Since the PS used is about 91% efficient @ that load, that pretty much jives with the Kill-A-Watt reading of 950-960w @ the wall that I logged.

If one builds something like that, the idea of a small case profile or footprint is not going to be in the cards.

Yes, but how many AIO or liquid cooled rigs have you actually seen in small cases?
Almost all performance rigs that are AIO or liquid are in mid towers or bigger.
But then again the principles of my rig, proper fan placement, fans that have adequate cfm without being overkill or high SP, intake radiators, and good cooling paths, can be applied to any size case using AIO or liquid and maintain a very quiet noise level.

I'd say that's a major accomplishment -- whether 56C or 62C -- for that kind of computing power.

I wouldn't call it that.
People do it every day with air or liquid cooled system and a bunch of fans, I just think they've been conditioned to do it in what I feel is the easy way and put up with the excess noise and expense.

My intent was to just verify common sense -
1) That using radiators as the intake, rather than the exhaust, is always the best way.
2) High cfm, noisy, expensive namebrand, or high SP fans are not necessary for liquid cooling, if proper fan placement is used along with good airflow paths.
3) It is easy to build a quiet (not silent, though) high performance rig.
4) That on the above 1), 2) and 3), most websites that insist on the opposite are wrong and just pass on hearsay, without bothering to actually test for a better way.

Anyway, I accomplished what I set out to do so I am satisfied with the results.

 

[BonzaiDuck]

. . . My intent was to just verify common sense -
1) That using radiators as the intake, rather than the exhaust, is always the best way.
2) High cfm, noisy, expensive namebrand, or high SP fans are not necessary for liquid cooling, if proper fan placement is used along with good airflow paths.
3) It is easy to build a quiet (not silent, though) high performance rig.
4) That on the above 1), 2) and 3), most websites that insist on the opposite are wrong and just pass on hearsay, without bothering to actually test for a better way.

Anyway, I accomplished what I set out to do so I am satisfied with the results.

Actually, your major points here are parallel to what I had in mind last year when I posted a thread of inquiry about the MO-RA3 radiators. "Old hands" posted discouragement to the use of large fans with low static pressure (most all have that as a relative characteristic). I think when this flap about intake versus exhaust emerged, it owed something to the use of smaller cases -- even with smaller radiators.

I posed one idea about rebuilding a midtower case side-panel to fit an MO-RA3 to it, even if it meant adding extension bars to the bottom of the case which would also include caster-wheels for mobility. And I think I tried arguing for the use of 200mm fans.

If that's what actually happened, then your design may have proven what was only a guess on my part. But you needed a larger case enclosure, and I was also fishing for how one would use 200mm fans in push-pull on a single radiator and make it fit.

That leaves the possibility of a dual-radiator arrangement -- supposing they are in the same loop, with one at intake and one at exhaust. It would most surely increase cooling capacity, but not quite by a factor of 2, given the simple fact that you're pushing warmer intake-radiator exhaust through the exhaust-side radiator.

BY THE WAY. If I win a contract with NSA, I'll be coming to you in hopes of buying your 24-core Xeon rig. :biggrin:

 

[Micrornd]

But you needed a larger case enclosure, and I was also fishing for how one would use 200mm fans in push-pull on a single radiator and make it fit.

I didn't actually "need" a larger case, this is just the case I've used for all my multi-cpu builds since the early 00's starting with a quad 200mhz pentium pro's

My radiator housing is just sheet metal and poly-carb added to the
top of the case as a mod, which could be easily done to any case.
Mine blows into the case, but if the rads were positioned vertical (instead of horizontal as mine are), any size fan(s) can be used and the airflow can be strictly through the rads, independent of what goes on in the case.
I think this would actually cool better than my setup.

I originally looked at a vertical housing, as well as one using the side of the case, since the hose length is long enough for either on the H110s, but for my use, horizontal on top of the case means the rad intake is within 2 inches of being the same height as my desk, so it fits nicely along side my desk, without looking too out of place.

During the day, it earns it's keep remotely rendering for my son's business, and at night, I use it for my work (and play)

 

[BonzaiDuck]

During the day, it earns it's keep remotely rendering for my son's business, and at night, I use it for my work (and play)

That makes practical sense. I does earn its keep, then.