Fan set up in modded C70.

[BonzaiDuck]

Installed the Aquaero 5 LT and had a thought: To ensure overpressure on the intake side, what if I mounted an additional 140mm PWM intake fan on the bottom like below.



I can control the H60 fans individually and the two intake fans as one, leaving the 4th fan header on the LT (PWM) for the added bottom intake.

I can build a curve in the software based on internal case temps or a virtual sensor with the CPU/GPU and run it from that. Quite at 800RPM at idle and then cranked up under load.

Now I need to find a monster airflow 140MM PWM fan!

I think I encouraged you to do that.

On the post you made just previous, it seems to me that you could increase CFM through a cooler, the exhaust air temperature would be cooler than at lower CFM, but the reduction in CPU temperature would be greater. It's a matter of how many air molecules are bouncing through the heatpipe fins per unit of time. The exhaust may be "cooler" but heat extraction would increase up to the point we all find somewhere when we get a Vantec Tornado pushing more and more CFMs with no further decrease in temperature.

 

[wgusler]

I think I encouraged you to do that.

On the post you made just previous, it seems to me that you could increase CFM through a cooler, the exhaust air temperature would be cooler than at lower CFM, but the reduction in CPU temperature would be greater. It's a matter of how many air molecules are bouncing through the heatpipe fins per unit of time. The exhaust may be "cooler" but heat extraction would increase up to the point we all find somewhere when we get a Vantec Tornado pushing more and more CFMs with no further decrease in temperature.

Correct. I believe that working with the fan speed on the 2 GT on the H60 will prove this. When I find that sweet spot, whatever number that is for both fans, I will set that as the max on the fan curve.

As far as the "air molecules are bouncing through the heatpipe fins per unit of time" that is a bit more complex.

Simply put, the radiator works by thermal conduction transferring the heat from the fluid inside into the mass of the radiator. Heat flows from a region of higher temperature to a region of lower temperature.

More specifically, thermal convection speeds the process; the tubes in the radiator move some heat directly to the air stream and transfer the bulk to the fins between the tubes, where they work just like a heat sink. They in turn transfer the heat to the airstream. Because the radiator has fans forcing air thru its core it is considered a forced convection device.

The faster the air flow over the core, the greater the transfer. But the limiting factors on the amount of heat transfer are the temp split between the ambient temp of the air flow and radiator and the ability of the radiator to conduct the heat from the fluid to the core.
At some point the flow thru the core will reach a point where it is removing all the heat that the radiator can move into the airstream.

The million dollar question is; will the fans move enough air to reach this point? If they can reach that at less than 100% speed, then you are using the radiators maximum ability to cool the fluid and can turn down the fans.

I would think that the H60 designers would have specified a fan that can come close to that point. So two higher static pressure fans sealed up in push/pull should exceed it. It is possible that the fan size, noise and power consumption requires a compromise on a retail product like the H60 and it is nowhere near the radiators limits. We shall see.

 

[BonzaiDuck]

Correct. I believe that working with the fan speed on the 2 GT on the H60 will prove this. When I find that sweet spot, whatever number that is for both fans, I will set that as the max on the fan curve.

As far as the "air molecules are bouncing through the heatpipe fins per unit of time" that is a bit more complex.

Simply put, the radiator works by thermal conduction transferring the heat from the fluid inside into the mass of the radiator. Heat flows from a region of higher temperature to a region of lower temperature.

More specifically, thermal convection speeds the process; the tubes in the radiator move some heat directly to the air stream and transfer the bulk to the fins between the tubes, where they work just like a heat sink. They in turn transfer the heat to the airstream. Because the radiator has fans forcing air thru its core it is considered a forced convection device.

The faster the air flow over the core, the greater the transfer. But the limiting factors on the amount of heat transfer are the temp split between the ambient temp of the air flow and radiator and the ability of the radiator to conduct the heat from the fluid to the core.
At some point the flow thru the core will reach a point where it is removing all the heat that the radiator can move into the airstream.

The million dollar question is; will the fans move enough air to reach this point? If they can reach that at less than 100% speed, then you are using the radiators maximum ability to cool the fluid and can turn down the fans.

I would think that the H60 designers would have specified a fan that can come close to that point. So two higher static pressure fans sealed up in push/pull should exceed it. It is possible that the fan size, noise and power consumption requires a compromise on a retail product like the H60 and it is nowhere near the radiators limits. We shall see.

" . . nowhere near the radiator[']s limit. . . " This is the same thing communicated to me by Noctua tech-support. They admit that their coolers can be made more effective, but they favor a customer base that wants less noise. If you can attenuate the noise yourself, it's less of a trade-off item.

Just to give an example of what you also said, I was pretty sure when I picked the PWM AP-30's that I would neither want -- nor need -- to run them full-bore at 4,200 RPM. Running a few 1-hour stress tests, I found that 3,200 was all that was needed, after which there would be no further improvement in load temperatures. With the 3-ply Spire-rubber-lined duct, 3,200 RPM may be "audible" but much harder to notice. 2,800 is a point where you start hearing anything at all. Oh -- I don't know. . . . maybe you can hear just a bit of "whoosh" out the rear of the case. No -- not at that speed, now that I think about it.

 

[wgusler]

Have the system up and running. The Aquaero LT is running but cannot access the MB sensors. Apparently Gigabyte are buttholes about sharing their sensor information with third parties. No matter the CPU/GPU info is transmitted to the LT.
I set up the Aquaero remote temp sensors in three locations in the case: upper area by the memory on the MB, a lower above the power supply and a one in the intake area of the H60. The 4th one is mounted under the case in the front for ambient. Also added the Notcua (?) 200 RPM industrial PWM fan in the bottom front for intake.
Not thrilled with the H60 so far, with all fans running at 100% and a 22C ambient the cores were 77C to 81C running the Intel burn test at maximum. Re applied some NT-1 and re seated the block with the same results.
I have moved the drives from in front of the upper intake fan to inside the cage to eliminate any flow restriction to the H60. I will test this and see if that helps the CPU temps. I ordered an H80I that should be here in a couple days and will install that with the GT-15s in a push / pull. Will update after.

 

[BonzaiDuck]

Have the system up and running. The Aquaero LT is running but cannot access the MB sensors. Apparently Gigabyte are buttholes about sharing their sensor information with third parties. No matter the CPU/GPU info is transmitted to the LT.
I set up the Aquaero remote temp sensors in three locations in the case: upper area by the memory on the MB, a lower above the power supply and a one in the intake area of the H60. The 4th one is mounted under the case in the front for ambient. Also added the Notcua (?) 200 RPM industrial PWM fan in the bottom front for intake.
Not thrilled with the H60 so far, with all fans running at 100% and a 22C ambient the cores were 77C to 81C running the Intel burn test at maximum. Re applied some NT-1 and re seated the block with the same results.
I have moved the drives from in front of the upper intake fan to inside the cage to eliminate any flow restriction to the H60. I will test this and see if that helps the CPU temps. I ordered an H80I that should be here in a couple days and will install that with the GT-15s in a push / pull. Will update after.

Did you say the Noctua iPPC fan you bought runs at TWO HUNDRED RPM? The models I saw were 2,000 and 3,000 RPM respectively. I wouldn't consider anything lower than that. I'd get the 3,000 RPM fan, isolate its mountings from the radiator, and wrap that sucker in Spire rubber. As many ply as you could fit around the radiator and duct. I think I had to stop with four layers surrounding my AP-30.

Comparing coolers on different platforms is difficult. For instance, if I said my processor is throwing out 135W under IBT and I get 65C degrees averaging the maximums recorded for my cooler, it would be different than putting it on your system running 135W under IBT if your processor is a 22nm as opposed to a 32nm lithography. If I remember, the die is even smaller, so less area means less heat transferred per unit time, and your core sensors are bound to show higher numbers than mine.

I'll come back and edit this post if I need to after I review your system components at thread beginning.

OK, WGusler . . . some other folks have firsthand experience with the 4790K, but I've been watching since it was released earlier this year. I've seen higher temperatures on the 4790K and yours aren't that bad. Of course, some folks are overclocking, but you can only go so far on those processors since they already have a turbo speed of 4.4 GHz.

Let me put it another way. You might find ways to make this work better. If you can trim another 10C off the load temperatures, I'd think that's pretty good. As it stands, it just about matches my 32nm SB-K running at 4.7 Ghz before I improved the ducting -- indeed -- before I tried another cooler. Put another way: I'd got it down to 75C, and it was still showing the lead core sometimes at 81C. Coupla degrees if you replace the NH-1 with diamond paste or CLU. Coupla degrees (maybe) if you lap off the nickel plate on the IHS. I think you could get your core peak-average to 75C -- maybe. Maybe 72C. I'm just guessing in a dim light.

At this point, I'd be interested in what others say about that particular processor, or whether your temperatures are good, average . . . worse.

You're not finished with this. I don't think so, anyway.

As an afterthought, consider what you're doing with a lower-tier AiO and what I've been doing with higher-tier heatpipes: we're exploring the limits of "adequate" to "better than adequate" so every little C improvement matters. IF you overwhelm your processor with custom water and bigger radiator -- maybe doing the same things -- you wouldn't be worried about the little Celsius nickels and dimes.

 

[wgusler]

2000 not 200. Move the drive cage and now the case temps are lower by 1.5C. Running the Heaven benchmark the GPU is 1-2 C cooler. Can't run the burn test now to see if the temps are better as it shuts down the pc after about 5 sec.

The only thing I can think of is when I shut down to move the cage windows did an update. Time for be I will fight it later.

 

[BonzaiDuck]

I can probably understand why they never produced an H60 with a 3"-thick radiator. But they probably should have tried, anyway.

It could be a challenge to fit a dual radiator in that case, but maybe it's possible. It would just be a lot easier for a very fat, single-fan radiator. Even if you had to convert exhaust to intake or vice-versa, it could all work the same way.

Who has 4790K temperature data with any cooler, fan configuration -- even water-cooling for certain OC settings, thermal wattage, voltage etc.? LIke I'd said, I was watching people post about that processor around here, and I think Gusler's H60 rig is promising.

 

[wgusler]

Ok, I think I have a handle on this now. The shutdown problem with Intel burntest was a windows update. That was only thing that changed when the shutdowns started, a service pack I believe. Uninstalled the update and the shutdowns stopped. Ran Burntest at very high for 1 cycle and passed @ 81C. Reconfigured the fans and re-ran it for 10 cycles........BSOD after 3.
Back to testing: Set the CPU OPT output in the BIOS to full speed to make sure the pump was at max RPM, updated the BIOS and ran BurnTest and Prime95. Burntest passed 5 cycles and BSOD. Prime95 for 30 min on blend was fine but max temp at 96C!!!!

Found a thread on Tom's Hardware Fourm with a post by Shikadensstating he had the same problem on the same MB and processor I have with shutdowns and BSOD's. He said that Gigabyte had by default pushed his core voltage to 1.4! He found a cure by lowering the voltage and pushed the 4790K to 4.4 G. with this config:


CPU CLOCK ratio: 44
I set Vcore manually to 1.13
CPU VRIN external override to 2.00
CPU VRIN loadline calibration: extreme
CPU VRIN current protection: extreme
PWM Phase Control: eXm Perf
and everything else on auto.


Looked and my Core Voltage was 1.29. I copied his config and BINGO! Passes Burntest x 10 and has been running Prime95 on blend for almost 5 hours now. And now to make the Duck happy...................Max CPU temp as of now, 70C running steady at 68C.

 

[BonzaiDuck]

Ok, I think I have a handle on this now. The shutdown problem with Intel burntest was a windows update. That was only thing that changed when the shutdowns started, a service pack I believe. Uninstalled the update and the shutdowns stopped. Ran Burntest at very high for 1 cycle and passed @ 81C. Reconfigured the fans and re-ran it for 10 cycles........BSOD after 3.
Back to testing: Set the CPU OPT output in the BIOS to full speed to make sure the pump was at max RPM, updated the BIOS and ran BurnTest and Prime95. Burntest passed 5 cycles and BSOD. Prime95 for 30 min on blend was fine but max temp at 96C!!!!

Found a thread on Tom's Hardware Fourm with a post by Shikadensstating he had the same problem on the same MB and processor I have with shutdowns and BSOD's. He said that Gigabyte had by default pushed his core voltage to 1.4! He found a cure by lowering the voltage and pushed the 4790K to 4.4 G. with this config:


CPU CLOCK ratio: 44
I set Vcore manually to 1.13
CPU VRIN external override to 2.00
CPU VRIN loadline calibration: extreme
CPU VRIN current protection: extreme
PWM Phase Control: eXm Perf
and everything else on auto.


Looked and my Core Voltage was 1.29. I copied his config and BINGO! Passes Burntest x 10 and has been running Prime95 on blend for almost 5 hours now. And now to make the Duck happy...................Max CPU temp as of now, 70C running steady at 68C.

I should've mentioned that. We were having a discussion here about the i7-5820K, and it became apparent that the wisest overclocking strategy was to reduce the voltage -- possibly even from auto / default. Whatever the particulars, that's what I remember. Should've told you.

My beliefs are flexible. I was really on track to ditch all this ducting stuff, except where it would offer improvement over water-cooling already installed. Then I got myself into this . . . old-tech . . . spare-computer build . . . motherboard tragedies etc. The least of my extravagance was purchase of an EVGA ACX cooler. For the SB-K, it chills!

But I suspected as you did that you could apply the techniques for high air-flow to an AiO cooler. If the 68C to 70C is from an Intel Burn-Test run, that's splendid. Even for Prime95 -- simply capital.

You may want to pick up a package of this stuff:

http://www.coolerguys.com/840556014003.html

Don't stick it all over your case sidepanels! Make constructions of it that don't leave an adhesive contact surface on the exterior. Put it closest to the noise source: try to bury your fan with it except for the intake and exhaust. If you need to stick it on something (other than your ducting parts), use foam-board panels the same size and apply a dab of Pit-Crew or similar adhesive to secure it to those special locations.

Now you have me thinking. If I still plan to do any "custom-water" projects, one of those fat 120x120 or 140x140 radiators could actually be enough. I've been looking for the point of "maximum adequacy" without becoming an MO-RA3 extremist.

Keep posting on this project.

 

[wgusler]

While Prime95 tries to kill it, I will lay out the current (and most likely final) configuration on the Ammo Box.
I7 4790k OC to 4.4
Gigabyte Z97-UD5H
GSkill Sniper DDR3 1833 4GB x 2
EVGA GTX 780 Ti SC w/ Ref Cooler OC'ed 143/167 @ 106% voltage
Intel 530 SSD 120GB
Intel 730 SSD 480GB
Seagate 1TB HDD
Corsair H60 Cooler w/ 2 GT-15 in push pull
(Soon to be replaced with an H80I)
Aquaero 5 LT controller
2x Corsair QF140 intake fans
1X Notcua Industrial NF-A14 2000rpm 140mm bottom intake fan
Here is the airflow setup:

And the Aquaero sensor locations:


Bonzai, you were correct in recommending the Aquaero 5 LT, it really is a great unit and the software is even better. The downside is a bit of a steep learning curve and limited set up info on the web.


The combination of the software sensors I could access and the 4 sensor included with the unit gave me the ability to read the temps in different areas of the airstream and adjust the fan speed curves to the temps and then build software controllers to manage them.

They are setup as follows:
The push /pull GT-15's on the H60 are linked to the CPU core temp - Fan speed at 10% @30C to 100% @ 60C


The Lower front Corsair QF 140 and the Notcua NF-A14 both run on a single curve controller linked to a virtual sensor. The virtual sensor is a composite of the 3 internal hardwired temp sensors. Their average temp is then used to create a single reading for the curve controller to follow; Fan speed at 10% at 23C to 100% at 28C.


By measuring the temps under and over the GPU and up in the dead space by the memory as well as the intake area of the H60, they react very well to loads on the GPU and chipset. The Bottom fan is somewhat limited by the proximity of the case bottom to the floor so I cut out the case under the fan to help it breath. I may build some 1" spacers to raise the case off the floor. That being said it blows outside air directly into the 780Ti blower fan and drops the GPU max temps by more than 5C.


I found that the GPU under load radiates quite a bit of heat and because the H60 is drawing air from inside the case right above the card, the CPU temp gains 1-2C if the GPU is at 50-60C under load. For this reason, I control the upper front intake fan separately from the other intake fans.
The upper front Corsair QF 140 that blows directly at the GPU and the H60 is also controlled by a virtual sensor of the CPU and GPU temps on a single curve controller as follows; fan speed at 10% @30C to 100% at 60C. Testing shows this fan @100% has a small affect on the GPU (1.0 to 1.5 C) but a big drop for the CPU at 2.5 to 4.0C as again, it is blowing directly into the intake of the H60.


The box is dead silent during non gaming use; the only thing I can hear is the pump, just barely. Running CPU or GPU benchmarks it makes some noise, but less than the two XPS Dells I have. I believe that by upgrading to the H80I and dropping the 780Ti back to stock clocks (over clocking this beast is overkill anyways); I can lower the noise by pulling all the fan curves down by 15 to 20% and still maintain temps.


Observations:
Removing the side panel on the MB side actually RASIES the CPU temps while dropping the case temps. The GPU doesn't seem to change temps with the case open. I believe the loss of case pressure hurts more that the benefit of more outside air flow.


The H60 is useable, but not any great shakes over an air cooler. I think it's more of an " I have a water cooled PC" thing than a real world cost benefit over air cooling. I have mine running at 46-48 full load over ambient after a LOT of work. Push/pull, sealing , adding a shroud and lots of airflow/ pressure to feed it.


Hopefully, by adding the GT-15's and sealing the H80I, I can keep the 46-48 with lower fan speed and noise.
The Box has been running Prime95 for 9 hours now and seems fine. Yea!

Here are some numbers.
Intel BurnTest at very high 10 passes 22.23 ambient 68C max CPU temp
Prime95 blend for 9 hours 21.85 ambient 70C max CPU temp
FurMark! GPU Test 30 mins 22.73 ambient 68C max GPU temp with card overclocked
FurMark! GPU Test 30 mins 22.26 ambient 66C max GPU temp with card stock clocked
Heaven 4.0 Ultra/ 8xAA 1980 64.4 FPS stock/ 70.1 overclocked.


The H80 gets here tomarrow..............back to work!

 

[BonzaiDuck]

While Prime95 tries to kill it, I will lay out the current (and most likely final) configuration on the Ammo Box.
I7 4790k OC to 4.4
Gigabyte Z97-UD5H
GSkill Sniper DDR3 1833 4GB x 2
EVGA GTX 780 Ti SC w/ Ref Cooler OC'ed 143/167 @ 106% voltage
Intel 530 SSD 120GB
Intel 730 SSD 480GB
Seagate 1TB HDD
Corsair H60 Cooler w/ 2 GT-15 in push pull
(Soon to be replaced with an H80I)
Aquaero 5 LT controller
2x Corsair QF140 intake fans
1X Notcua Industrial NF-A14 2000rpm 140mm bottom intake fan
Here is the airflow setup:

And the Aquaero sensor locations:

Bonzai, you were correct in recommending the Aquaero 5 LT, it really is a great unit and the software is even better. The downside is a bit of a steep learning curve and limited set up info on the web.


The combination of the software sensors I could access and the 4 sensor included with the unit gave me the ability to read the temps in different areas of the airstream and adjust the fan speed curves to the temps and then build software controllers to manage them.

They are setup as follows:
The push /pull GT-15's on the H60 are linked to the CPU core temp - Fan speed at 10% @30C to 100% @ 60C


The Lower front Corsair QF 140 and the Notcua NF-A14 both run on a single curve controller linked to a virtual sensor. The virtual sensor is a composite of the 3 internal hardwired temp sensors. Their average temp is then used to create a single reading for the curve controller to follow; Fan speed at 10% at 23C to 100% at 28C.


By measuring the temps under and over the GPU and up in the dead space by the memory as well as the intake area of the H60, they react very well to loads on the GPU and chipset. The Bottom fan is somewhat limited by the proximity of the case bottom to the floor so I cut out the case under the fan to help it breath. I may build some 1" spacers to raise the case off the floor. That being said it blows outside air directly into the 780Ti blower fan and drops the GPU max temps by more than 5C.


I found that the GPU under load radiates quite a bit of heat and because the H60 is drawing air from inside the case right above the card, the CPU temp gains 1-2C if the GPU is at 50-60C under load. For this reason, I control the upper front intake fan separately from the other intake fans.
The upper front Corsair QF 140 that blows directly at the GPU and the H60 is also controlled by a virtual sensor of the CPU and GPU temps on a single curve controller as follows; fan speed at 10% @30C to 100% at 60C. Testing shows this fan @100% has a small affect on the GPU (1.0 to 1.5 C) but a big drop for the CPU at 2.5 to 4.0C as again, it is blowing directly into the intake of the H60.


The box is dead silent during non gaming use; the only thing I can hear is the pump, just barely. Running CPU or GPU benchmarks it makes some noise, but less than the two XPS Dells I have. I believe that by upgrading to the H80I and dropping the 780Ti back to stock clocks (over clocking this beast is overkill anyways); I can lower the noise by pulling all the fan curves down by 15 to 20% and still maintain temps.


Observations:
Removing the side panel on the MB side actually RASIES the CPU temps while dropping the case temps. The GPU doesn't seem to change temps with the case open. I believe the loss of case pressure hurts more that the benefit of more outside air flow.


The H60 is useable, but not any great shakes over an air cooler. I think it's more of an " I have a water cooled PC" thing than a real world cost benefit over air cooling. I have mine running at 46-48 full load over ambient after a LOT of work. Push/pull, sealing , adding a shroud and lots of airflow/ pressure to feed it.


Hopefully, by adding the GT-15's and sealing the H80I, I can keep the 46-48 with lower fan speed and noise.
The Box has been running Prime95 for 9 hours now and seems fine. Yea!

Here are some numbers.
Intel BurnTest at very high 10 passes 22.23 ambient 68C max CPU temp
Prime95 blend for 9 hours 21.85 ambient 70C max CPU temp
FurMark! GPU Test 30 mins 22.73 ambient 68C max GPU temp with card overclocked
FurMark! GPU Test 30 mins 22.26 ambient 66C max GPU temp with card stock clocked
Heaven 4.0 Ultra/ 8xAA 1980 64.4 FPS stock/ 70.1 overclocked.


The H80 gets here tomarrow..............back to work!

Your curiosity is tapping your wallet!

But . . . it's the Big Build. This could be innerestin'.

Automotive silicon adhesive-sealant? Yeah. That works.

Oh. Take a look here, for lifting up the case.

http://www.dlawlesshardware.com/casterwheels.html

You could get by with smaller wheel-diameter with that case. Or you could do something really bizarre and put 3.5" braked double-wheeled casters on it. Maybe -- a fashion statement. I use 1/4"x2"x12" aluminum bars to mount them in pairs, and tap screw holes for the models with square mounting plates. But you shouldn't have any compulsion now about making physical modifications to the case and tapping the holes for each wheel without using the 6-32 holes for the original rubber feet -- if that's the way the case is made.

The other possibility comes again from the electronics jobber warehouse/store. They've got 2"-high rubber feet -- maybe even bigger.

 

[XavierMace]

The H60 is useable, but not any great shakes over an air cooler. I think it's more of an " I have a water cooled PC" thing than a real world cost benefit over air cooling. I have mine running at 46-48 full load over ambient after a LOT of work. Push/pull, sealing , adding a shroud and lots of airflow/ pressure to feed it.

I would still put money on your setup having a lot to do with that.

 

[wgusler]

I would still put money on your setup having a lot to do with that.

No doubt that it does. But the intake air at the H60 is running about 2C over ambient max. So with the fan direction reversed you pick up 2c cooler air making a 44 to 46c drop. You would also have to factor the loss of flow from case pressure also.
Not bashing the H60, I think is really neat that now days you can 35 to 65 bucks and get self contained water cooling system. My point was that the same amount of effort on air will give the same results from my experience.
I hope I can lose another 3-5 C with the H80's larger radiator and cut fan speed. At that point, I can see an improvement over air coolers as far as pure performance.
Fact is, I'm one of those " Hey look at me, I have a water cooler" guys!

The Ammo Box in its natural habitat.

 

[BonzaiDuck]

I took a look at the H80 specs, but still not sure how it differs much from the H80i. The radiator is 50% wider -- definitely a plus. Whether the cooling capacity increases by as much, by more or by less, we'll only find out here unless a water-cooling veteran can enlighten me -- or us. My first thought would be to assume it's just linear.

The stock fan is also a 2,000 RPM affair. I still wonder whether a 3,000 RPM iPPC fan would've bought you some extra C's.

It used to be the case that the more innovative ducting projects were made with simple fin-type heatsinks sitting in an orientation like the stock cooler. I saw one guy who used parts of a kitchen-rubber-glove on one of those Zalman "flower" coolers.

With heatpipes, you're mostly increasing the CFM you wouldn't have without a duct. YOu can eliminate the dead spot on the cooler or radiator -- whichever.

One of the things WGusler proved here if it hadn't already been proven: a pressurized case can enhance a liquid cooling system. I had just naturally assumed it would, and I thought someone had been skeptical when I suggested it in another thread. It's almost like WGusler proved it by accident when he removed the case sidepanel!

I would never have looked at an H60 or H80 as a viable cooling solution for myself. This changes everything. Makes me look more closely at something like this:

http://www.frozencpu.com/products/1...40mm_Radiator_-_80mm_Thick.html?tl=g30c95s929

But in Gusler's configuration, something like that would be more troublesome with an exhaust duct that keeps the fan separate from the rad. Hmm. I'd used those Evercool aluminum fans to mount outside the case. No reason you couldn't do it with any other fan, but it changes the problem of noise-control.

Yeah -- I want to wait and see what this 38mm H80 will do . . . .

 

[XavierMace]

No doubt that it does. But the intake air at the H60 is running about 2C over ambient max. So with the fan direction reversed you pick up 2c cooler air making a 44 to 46c drop.

Are you trying to say your ambient air is 44c?

 

[ClockHound]

Are you trying to say your ambient air is 44c?

Just another day in tweaker hell...

 

[BonzaiDuck]

Just another day in tweaker hell...

A mild day in Death Valley?

I'm sure he'll clarify this. He mentioned that there's a steep learning curve with the Aquaero controller.

Even in the smaller C70 case -- with Gusler's fan-deployment -- I wouldn't have thought that the ambient measured inside the case would be that high.

This had always been a minor point of contention for analyzing reviews. In my cases, I usually find the motherboard and CPU varying in a range between 25 and 31C. I take the "room ambient" from the wall-thermostat, having measured it from time to time with a calibrated digital thermometer to satisfy myself that it's "reasonable."

 

[wgusler]

Are you trying to say your ambient air is 44c?

No, sorry 230am post and I wasn't clear. The H60 was running at 46C to 48C OVER ambient at full load and XavierMace commented that the one of the reasons the setup was running at that was not pulling outside air into the H60 as recommended by Corsair.

The temps readings at the intake of the H60 are 1.5 to 2.0C higher than ambient, so changing the fan direction to draw ambient instead of inside the case air would, in theory, drop the CPU full load temp over ambient by 2C. This would be 44-46C over ambient. To clarify, here is data from my spreadsheet:


12/9/14 1:42AM
H60 sealed / push-pull with GT-15
Ambient 20.17
Case temp Average 21.43
H60 Intake temp 22.04
CPU package idle temp 28.0
IBT @Veryhigh CPU 68.20---48.03
Prime95 1 hour CPU 66.93----46.76


Another thing to consider, by changing the fan direction is the loss of airflow thru the cooler via case pressurization. As I said, removing the side panel raises the temps under load by 1.0Cto 1.5C. Replacing the panel the temp drop back down. Didn't believe it at first, but I have replicated it 3 or 4 times.


So again, changing the fan direction in this case might not make more than a .5C improvement.


From looking at reviews and tests of H60, H80, and other 120mm CLCs, the numbers over ambient run around the 50C range. So my cobbled up mess of an H60 is doing a good job IMO. The point was that good air coolers can do the same numbers with a little work. The H80 has some reviews showing better cooling than the H60 by 3-5C at full load with stock fans. If that is the case, I should be around 44-46C over ambient with the H80. We will see.

 

[BonzaiDuck]

First -- yes -- a trade-off between one positive cooling factor and another.

Second -- also yes -- doing a good job IMO. I offer "verification."

Does your summary of the "50 C range" include dual fan or 240mm radiators? I suppose I should look more closely. Either way -- we'll find out and post.

I would still try an AP-30 or a 3,000 RPM iPPC Noctua. Foam-rubber -- nothing to be ashamed of. The biggest likelihood that it would ever spin up to the maximum is in the stability benchtesting itself, or some really demanding games. With thermal fan control and attention to acoustic muffling, I don't think it would depart much from "quiet."

Add another few C's to your load-temperature reduction delta, and you should be able to play with some decent clock settings. Also, it presents me with some alternative radiator options to use with a reservoir and dual pumps. Depends on the difference between Haswell and the E.

 

[wgusler]

Installed the H80 I with the GT-15's in push/pull and seal gaskets from the rad to fans. First Intel Burn test the temps went to 98 within 1 sec. Shut it down and re applied NT-1 paste and reseated the cooler. Saw that the CPU was at 58 just after boot up..................Sum Ting Wong!


Pulled it apart and noticed that I could move the cooler in the mount and the thermal paste wasn't compressed like it should be. The backplate allowed too much slack between the mount/MB and the cooler. So I used two 1.0 mm plastic washers on all 4 back plate standoffs and reinstalled. Can't make it budge now.


Booted up at 28C and I started running burn in apps. All I can say is wow! The H80 is a great deal more effecient than the H60. I did not hook up any of the USB data cables to the H80I, just the SATA power connector, the fans and the RPM plug to the CPU OPT. The Auqaero LT will control the fans.


Had some random lockups and bluescreens starting yesterday with the H60 still installed so I did up the VCORE on the 4790 to 1.20V from 1.13V and the weirdness stopped. Ran Prime 95 for 4 hours and it's all good. With everything else the same in the setup ( I didn't put the shroud on the H80) I started testing.



H60 Baseline numbers.
4790K OC to 4.4K with Vcore @ 1.13V Idle temp at 31C
Intel Burntest / Very High/ single pass/ 66C with ambient at 18.25= 47.75
Prime 95 Blend x 30 min/ 67C with ambient at 18.40= 48.60

H80I Baseline numbers.
4790K OC to 4.4K with Vcore @ 1.20V Idle temp at 30C
Intel Burntest / Very High/ single pass/ 56C with ambient at 18.58= 37.42
Prime 95 Blend x 30 min/ 57C with ambient at 18.87= 38.13

Pretty good numbers so I installed all the fan dust screens in the H70 and ran Prime 95 for 4 hours. The numbers stayed about the same.

58C with ambient at 19.57= 38.43 both GT-15 running at 75 to 80% on the fan controller. You can hear the fans under max load, but this thing is silent ( other than the pump noise) under normal use.

I will play with it some more, but I am very pleased with the setup as of now.

 

[BonzaiDuck]

Lookin' better all the time!

I still think the Devils Canyon has a different temperature profile than my Sandy-K's, even for the same thermal wattage.

HWMonitor reports the "package" wattage under "Powers." I'd be interested in what IBT ("Very High" or "Maximum") shows in this regard for your processor and cooling solution.

I'm tentatively concluding that your H80i, mods and configuration will trump my ducted EVGA ACX cooler hands down, but the actual wattage will tell more. If you're clocking the 4790 to 4.4 and have it set to "by all cores," the wattage would exceed the spec.

 

[BonzaiDuck]

I have been to the mountain . . . heh-heh . . .heh-heh. . . .

And the mountain is good . . . heh-heh . . . heh-heh . . .

http://www.newegg.com/Product/Produc...-052-_-Product

The cubic-meters/hour translate to about 110 CFM. I used it to replace the red LED fan that came with my EVGA ACX cooler. Definitely -- less noise, but I haven't seen it peg to top-end yet under stress-testing. Currently passing OCCT:CPU and now OCCT:CPU-Linpack @ 4.6 ghz, ~1.33V with maybe 0.02 or 0.03 droop. Average maximum temperatures around 64C, ASUS BIOS fan-control set to "Turbo." Significantly more quiet than the stock fan, but spinning at least as fast.

Then there's this Bad Boy:

http://www.newegg.com/Product/Produc...-048-_-Product

About 160 CFM at top-end, with lower dBA than the 120.

Gentle-Typhoon AP-30's? We don' need no AP-30's. Whadda we need stinkeen AP-30's for?

Anyway, the Package Power Maximum under the above settings with affinitized LinX 0.6.5 is 134.25W.

So I'd be interested in seeing your temperatures and your wattage. It's going to explain something about the Haswell K chips -- and probably the E's as well. Reasoning behind it: your H60 and H80 experiments are likely equal or better to my ACX cooler.
*******
*******
Just an update to this fan focus: You'd pretty much expect this of Noctua, and less so for other fan-makers. Their spec is usually stated as +/- 10% of the top-end RPM. I suppose this is within the spec, though: the Noctua PWM 3K-RPM iPPC fan tops out at 2,800 and change. On the plus side -- you simply can't hear it. I can't hear it.

 

[wgusler]

Ok, I am done playing with the Ammo Box. O/C it up to 4.6 and seemed to be stable in limited testing but reached 71C during Intel Burn test. Ran BF4 and COD AW and did not see any change in frame rates.
Don't like those temps and don't need 4.6 to play games. Put it back to stock clocks and the core volts to 1.15V. Ran 95 for 30 min and IBT on very high with a max temp of 55C @ 20.97 Ambient and 74.5 watts max.
So at stock clocks, max temp over ambient was 34.03. Pretty good cooling with my current setup and quite as a mouse. Thank you guys for your imput and advice.

 

[BonzaiDuck]

Ok, I am done playing with the Ammo Box. O/C it up to 4.6 and seemed to be stable in limited testing but reached 71C during Intel Burn test. Ran BF4 and COD AW and did not see any change in frame rates.
Don't like those temps and don't need 4.6 to play games. Put it back to stock clocks and the core volts to 1.15V. Ran 95 for 30 min and IBT on very high with a max temp of 55C @ 20.97 Ambient and 74.5 watts max.
So at stock clocks, max temp over ambient was 34.03. Pretty good cooling with my current setup and quite as a mouse. Thank you guys for your imput and advice.

What was the peak "package power" wattage with those 4.6Ghz IBT runs?

Certainly if you're referring to a package temperature, and no less if either the lead-core's temperature or the peak average -- I'd say that's darn good.

You're not going to do any thermal damage to the processor at that speed and temperature. That would only leave you the possibility of damage by overvolting.

Is there even an extant opinion about a "sane" voltage for the i7-4790K or similar?

You'd wonder how it would work with a de-lidded 4790K. But you'd also wonder if it's worth the trouble, worth the risk, or any advantage at all to anything but overvolting the processor.

 

[wgusler]

Peak power at 4.6 / 1.255V was 81.3 watts on the CPU max. At stock clocks and 4.4 this thing runs just fine a 1.13V. I bumped it up to 1.15 V just to be safe . The intel specs say 1.25V and the Gigabyte MB has it defaulted to 1.295V.

No clear answer on Vcore max numbers, but I read a thread where someone has a 4790K at 4.8 using 1.44V! Said it was stable but peaked at 90C with a H100I. Yikes!

I thought about making a 90 degree duct to the top of the case just to see what the draw of outside air will do vs the current set up. FYI, it I power down both GT-15's on the H80 and leave all the case fans on the controller, they spool up to max when the CPU reaches 68 to 70C and the temp maxed at 74C.

You can feel air still passing thru the H80 just from case pressure. I don't have the balls to kill all the fans and run burntest, but I wonder what it would do. Then I can compare using no fans and case pressure only to dead air radiation on the rad.