Observations Re: Top Mount H100i as Intake

[Stringjam]

Recently assembled the PC in my signature, and before putting it together, I scoured the web to determine what sort of case fan / orientation setup I should use.

Countless times I heard things like "hot air rises...if you're going to top mount the H100i, you have to use it as an exhaust. Exhausting out the front of the case simply doesn't work."

In my situation, this turned out to be inaccurate.

After assembling according to recommendations, I was getting what I felt was poor performance.

At 4.6 GHz: 68-75C
At 4.8 GHz: 80C

I messed with all the fan speeds for quite some time, and even remounted the pump to make sure there wasn't an issue there. Nothing lowered the temps.

Pulled it all apart and remounted as intake. Remounted my other three 140mm Phanteks as exhaust - front and back.

The difference was huge. Now I can run at 4.8 GHz and only hit 54-60C. I was able to lower the curve on the radiator fans and significantly reduce the noise.

The GPUs are also happy, running at 58-62C (previously 65-70C).

So at least, in my case, intaking the H100i and using the remaining fans to scavenge the hot air from the case was the proper solution.

 

[DrMrLordX]

What was your prior setup? Top-mounted exhaust?

 

[Stringjam]

What was your prior setup? Top-mounted exhaust?

First configuration was top-mounted exhaust. For the case fans, I tried both intake front / exhaust back and all intake.

 

[BonzaiDuck]

I would think the delta such as you observe would depend on overall case airflow. Some earlier threads had me sparring with someone about "exhaust" versus "intake." While we'd all like to mount our radiators at intake, it would matter somewhat or proportionately less with smaller case volume and higher intake-to-exhaust airflow CFM.

I've only recently been surprised by how adding a second GPU will increase case interior ambients. I saw my peak load CPU temperature increase by 4 or 5C degrees with the second GPU. And I have a case with "High-Air-Flow."

On the "hot air rises" factor, that -- too-- would have less to do with anything with higher overall CFMs. I can't see that as an objection to top-mount with intake, but then -- I'd have to try it.

Anyway, higher CFMs can mean more RPMs and more dBAs, so . . . it's a balancing act.

Now looking closer, I see that you mounted the H100 at the top and reversed the flow. Here, my speculations are possibly "hot air" about hot air. I'd only guess with thermally controlled fans, the idle temperatures would be higher, but the load temperatures would be lower, such as those you show.

And the thing about radiator at intake: any strategy that makes for case pressurization could defeat some of that intake. So you'd want the exhaust CFMs to be a tad higher than the intake CFMs -- if you could measure such a thing.

 

[DrMrLordX]

First configuration was top-mounted exhaust. For the case fans, I tried both intake front / exhaust back and all intake.

I agree that this is probably a case of inadequate case fans. If you have poor intake, you are basically asking those rad fans to pull air through the rads and to pull air in through whatever openings exist for your case. You're also exposing the rad fins to all the heat generated by everything else dumping heat into the case air, though that is not that big of a deal with sufficiently-powerful rad fans and case fans.

Using rad fans as intake is probably the safest bet when it comes to taking care of CPU temps. Such a configuration hampers cooling everything else in the case (depending on how much actual heat your rad is dumping into the case), but with sufficiently-powerful exhaust fans, you'll be okay.

 

[BonzaiDuck]

I agree that this is probably a case of inadequate case fans. If you have poor intake, you are basically asking those rad fans to pull air through the rads and to pull air in through whatever openings exist for your case. You're also exposing the rad fins to all the heat generated by everything else dumping heat into the case air, though that is not that big of a deal with sufficiently-powerful rad fans and case fans.

Using rad fans as intake is probably the safest bet when it comes to taking care of CPU temps. Such a configuration hampers cooling everything else in the case (depending on how much actual heat your rad is dumping into the case), but with sufficiently-powerful exhaust fans, you'll be okay.

If you use the radiators for exhaust, then high intake airflow and pressurization would help with good fans with decent power or throughput. I wouldn't hesitate to get a couple Noctua iPPC fans for such a radiator, but we're talking about $60 -- nevertheless. More that I think about it, I might use the same fans for intake and exhaust.

If you make the radiator an intake affair, you have other cooling imperatives that can also be fixed, and you'd want the exhaust fans to help with the intake airflow as well.

 

[Stringjam]

I agree that this is probably a case of inadequate case fans.

My 140mm Phanteks case fans are 87 CFM, which I would think would equate to some pretty good flow. Hmm...

 

[DrMrLordX]

Hmm. That's odd. Got any significant airflow blockages inside the case? Poor cable management or something? Do the case fans develop any significant amount of static pressure?

The only way I can think of that would make your AiO perform significantly better as an intake rather than an exhaust is if your case was under low pressure, forcing the fans on the AiO rad to work really hard just to move any air. If the previous setup was rad exhaust + case fan intake, then the implication is that the fans on the rad were outperforming the case fans (or trying to anyway).

 

[Stringjam]

Hmm. That's odd. Got any significant airflow blockages inside the case? Poor cable management or something? Do the case fans develop any significant amount of static pressure?

I don't think there are any significant blockages. Cables are all managed behind the back panel.

Just to note, I took this pic before I changed the fans out. I also removed one of the top drive bays and moved half of my hard drive bays up to the top so that at least one of the front fans would have an unobstructed shot at the GPUs.

 

[DrMrLordX]

Yeah, you're not dealing with any kind of cable clutter. That's a clean system compared to my box, ha ha. Ugh. But I digress.

Not sure why the intake config was so dramatically better for you, but hey, if it works, great. At least you discovered this by trial-and-error.

 

[Stringjam]

Yeah, you're not dealing with any kind of cable clutter. That's a clean system compared to my box, ha ha. Ugh. But I digress.

Not sure why the intake config was so dramatically better for you, but hey, if it works, great. At least you discovered this by trial-and-error.

Thanks for the input on the topic.

My best guess is that it has something to do with the glut of hot air these 980s are pumping out and filling the case with. Perhaps exhausting that hot air through the radiator just wasn't adequate to hit the CPU temps I was aiming for.

 

[guskline]

Stringjam, my guess is you answered the question yourself. On exhaust mode you were pulling a large amt of hot air from the 2 980s into your radiator. On intake you are pulling cool air.

 

[Stringjam]

Stringjam, my guess is you answered the question yourself. On exhaust mode you were pulling a large amt of hot air from the 2 980s into your radiator. On intake you are pulling cool air.

Yeah, it makes me wonder if going with the squirrel-cage fan GPUs that exhaust out the back of the case might be a better option.

My past experience tells me that design is usually a good bit noisier and less effective at cooling the GPUs, though, which is why I went with the non-reference design this time around.

 

[Cabezone]

I used a reference 980 for jsut that reason. I only have a single card but thought you'd be interested in my build. It's intended to be nearly completely silent not super efficient with the heat.

Copied and pasted from another forum:

I thought I'd do a brief write up, for those that like quite systems, on my final results in moving my puter to it's new home the Nanoxia Deep Silence 1:



That's a GTX 980 running at 1480 (reference cooler), a 3570k@ 4.4 (Corsair H110), so some fairly standard overclocks. My DB readings from my sound studio, late night silent apartment(hamster was not running on wheel), and my studio grade listening device, Galaxy S5 and the free noise meter ap, taken at 1 foot(30 centimeters) are 17db system off, 21db at idle, and 28db while stressing both the CPU and GPU. I am very pleased with those results.

I ended up really liking this case aside from a few nitpicks. The vents at the top are centered and not offset to the front to accommodate radiators and they went with round and not slotted fan holes. So I had to do some custom work up top to make the H110 fit. Behind the tray could use a bit more space for cabling, main power cable is a tight fit. I went with this case because of the covered top, I figured I could get it working well without having to pop the chimney up. I am running the cpu cooler pulling case air and exhausting out the top. This works well because the reference blower design on the 980 blows directly out of the case so the inside of the case never really warms up much. Power supply pulls from outside the bottom of the case and vents directly out also, so it doesn't impact case temps much.

The fans are all stock except the rear fan was replaced with a NF-S12A that I had in my other case. I could probably knock a couple more DB off by replacing all the fans with 140mm Noctura, I will probably do this later in the year for funsies. It's plenty quite enough as it is now.

EDIT: Not sure on case temps, CPU cores are in the low 60's, GPU is the 980 standard running temp of 79-81.

 

[DrMrLordX]

Yeah, it makes me wonder if going with the squirrel-cage fan GPUs that exhaust out the back of the case might be a better option.

My past experience tells me that design is usually a good bit noisier and less effective at cooling the GPUs, though, which is why I went with the non-reference design this time around.

Hmm, so your cards are set up to exhaust into the case? Ouch, yeah that could be it. I didn't realize that was the case.

What are the standard operating temperatures for your 980s during idle and moderate-to-heavy use? If the case temp is defined by your GPU temp (which is the maximum temperature of air leaving the cards as exhaust), then that same temperature is the minimum temperature for your rad fin surface (in reality, the fin surface will have to be hotter to move any heat off the rad).

Now, with strong enough intake fans, that shouldn't have been a big issue. All they had to do was saturate the case with enough ambient-temperature air to spread out the heat exhausted by the cards and normalize temperatures. That wasn't going to happen if the cards were cranking out too much heat, or if the rad was drawing hot air directly from the cards before it mixed with anything from the intake fans (creating a "cold" air cell at the bottom of the case that never mixed with air from the top of the case).

I almost wonder if AiO rad mounting for exhaust configurations should be done near the bottom of the case, where lurks the coldest available air.

 

[Stuka87]

Corsair specifically states to always mount it as an intake if possible. Using cool external air works far better than sucking in hot case air.

I have mine mounted as an intake and the CPU stays quite cool under load. 45-50C under heavy gaming with all cores loaded to 85-90%. And it hits mid 60's with OCCT running. Thats at 4.5GHz at 1.2V.

 

[Stringjam]

Hmm, so your cards are set up to exhaust into the case? Ouch, yeah that could be it. I didn't realize that was the case.

What are the standard operating temperatures for your 980s during idle and moderate-to-heavy use? If the case temp is defined by your GPU temp (which is the maximum temperature of air leaving the cards as exhaust), then that same temperature is the minimum temperature for your rad fin surface (in reality, the fin surface will have to be hotter to move any heat off the rad).

Now, with strong enough intake fans, that shouldn't have been a big issue. All they had to do was saturate the case with enough ambient-temperature air to spread out the heat exhausted by the cards and normalize temperatures. That wasn't going to happen if the cards were cranking out too much heat, or if the rad was drawing hot air directly from the cards before it mixed with anything from the intake fans (creating a "cold" air cell at the bottom of the case that never mixed with air from the top of the case).

I almost wonder if AiO rad mounting for exhaust configurations should be done near the bottom of the case, where lurks the coldest available air.

The 980's sit around 40C. Medium use will bump them up to 50C or so, and heavy use will get them into the low 60's. Now mind you, this is with my current "all case fans as exhaust" method.

In my previous configuration, where the back fan and rad fans were exhausting, they would get up to 75C. That's a lot of hot air getting pulled up through the radiator. I even maxed all the fan rpms to see what effect it would have, and it did almost nothing except sound like jet taking off.

It seems to me like the GPUs pumping hot air into the case is key here. Flipping my case fans all to exhaust immediately relieved the GPU temps and required fan speed.

Bottom mounting (or front mounting) the radiator as intake as you mentioned seems like it would be ideal. Unfortunately, the H100i doesn't come with long enough hoses to even make that a possibility. :\

I think if I were doing this again, I might look at some other cooler options. I am certainly not impressed with the H100i mounting hardware, which I had to add washers to (the thumb-screw slots are so wide that they barely provide any surface area for the base of the screw).

Other than that, as a cooler, it seems to work great as an intake. Pushing all cores past 90% at 4.8GHz generates a lot of heat, and the temps I'm getting now are excellent.

 

[BonzaiDuck]

The 980's sit around 40C. Medium use will bump them up to 50C or so, and heavy use will get them into the low 60's. Now mind you, this is with my current "all case fans as exhaust" method.

In my previous configuration, where the back fan and rad fans were exhausting, they would get up to 75C. That's a lot of hot air getting pulled up through the radiator. I even maxed all the fan rpms to see what effect it would have, and it did almost nothing except sound like jet taking off.

It seems to me like the GPUs pumping hot air into the case is key here. Flipping my case fans all to exhaust immediately relieved the GPU temps and required fan speed.

Bottom mounting (or front mounting) the radiator as intake as you mentioned seems like it would be ideal. Unfortunately, the H100i doesn't come with long enough hoses to even make that a possibility. :\

I think if I were doing this again, I might look at some other cooler options. I am certainly not impressed with the H100i mounting hardware, which I had to add washers to (the thumb-screw slots are so wide that they barely provide any surface area for the base of the screw).

Other than that, as a cooler, it seems to work great as an intake. Pushing all cores past 90% at 4.8GHz generates a lot of heat, and the temps I'm getting now are excellent.

Hadn't time to check if you had one or two/SLI. What is your idle power consumption % on the gfx cards themselves? AfterBurner reports it. It should go into a lower P-state and show maybe 10%.

I think there's a bug in the driver for SLI. With my 970's, reboot after enabling SLI shows 40% power consumption on the cards, when they show <10% after re-enabling SLI. If it's not a driver bug, it's "something" I'll find through troubleshooting. But that puts my temperatures on the cards at 40C in idle, when they should show between 28 and 32C.

With this, I may have said it: once I installed the dual graphics cards and ran a CPU stress-test, I was stunned that my IntelBurnTest peak temperatures had increased 4 to 5C.

 

[Stuka87]

I think if I were doing this again, I might look at some other cooler options. I am certainly not impressed with the H100i mounting hardware, which I had to add washers to (the thumb-screw slots are so wide that they barely provide any surface area for the base of the screw).

Why and where did you have to add washers? I did not have to add any to mine.

 

[DrMrLordX]

Bottom mounting (or front mounting) the radiator as intake as you mentioned seems like it would be ideal. Unfortunately, the H100i doesn't come with long enough hoses to even make that a possibility. :\

I think if I were doing this again, I might look at some other cooler options. I am certainly not impressed with the H100i mounting hardware, which I had to add washers to (the thumb-screw slots are so wide that they barely provide any surface area for the base of the screw).

I can't speak to the mounting hardware, but the h240-x is a powerful AiO unit that is actually somewhat moddable. You can actually swap out the block and possibly other bits if you like. It's pretty expensive, though.

 

[Stringjam]

Hadn't time to check if you had one or two/SLI. What is your idle power consumption % on the gfx cards themselves? AfterBurner reports it. It should go into a lower P-state and show maybe 10%.

Wow....I'm glad you mentioned this. Mine were at 40%! I disabled / reenabled SLI and it dropped to 8%. My electric bill thanks you.

Why and where did you have to add washers? I did not have to add any to mine.

Underneath the thumb screws on the top. The base of the thumb screw is barely larger than the slot the stud sits in on the bracket.

I can't speak to the mounting hardware, but the h240-x is a powerful AiO unit that is actually somewhat moddable. You can actually swap out the block and possibly other bits if you like. It's pretty expensive, though.

Really nice looking unit (with some great reviews). If this one ever shuts down, I'll definitely put that one on the list.

 

[BonzaiDuck]

Wow....I'm glad you mentioned this. Mine were at 40%! I disabled / reenabled SLI and it dropped to 8%. My electric bill thanks you.

Mr. Stringjam! Sir! You have provided me immense insight without realizing it!

Once again, I'm rocking two GTX 970's in SLI; you have two 980's -- correct me if I'm wrong.

What driver version are you using? Is it 347.52 ??

I think it is now fairly clear that there is a bug in the NVidia driver pertaining to P-states and power-saving clocks. I was still troubleshooting this, when I was interrupted by a dead monitor. The monitor's problem had nothing to do with either my GTX 780 or my 970's. It was a s***ty 2009 Hanns-G, with customer-reviews reporting early failure and symptoms I had this last week rendering it totally unusable.

If you were running them in SLI, found them at 40% power, disabled then re-enabled SLI to see 10% power -- it's the same symptom as the 970's I have.

Please tell me if otherwise or either way, but reboot the system and check the NVidia cards' power consumption again after the system is at after-boot idle. In my case, it goes back to the 40% power-consumption and higher P-state, and you have to play this disable-re-enable game with SLI again.

Either way, I think there's a driver bug with the new 9xx cards. NVIDIA MUST BE TOLD! THEY MUST FIX IT! IT'S NOT THE HARDWARE! IT'S NOT THE 970 "RAM" ISSUE. IT CAN ONLY BE THE DRIVER.

 

[Stringjam]

What driver version are you using? Is it 347.52 ??

...

If you were running them in SLI, found them at 40% power, disabled then re-enabled SLI to see 10% power -- it's the same symptom as the 970's I have.

Please tell me if otherwise or either way, but reboot the system and check the NVidia cards' power consumption again after the system is at after-boot idle. In my case, it goes back to the 40% power-consumption and higher P-state, and you have to play this disable-re-enable game with SLI again.

I am also running on 347.52, and just as you, rebooting brought the power usage up to 40% again. I think there is a definite issue!

 

[Corsair Joseph]

Underneath the thumb screws on the top. The base of the thumb screw is barely larger than the slot the stud sits in on the bracket.

Not quite sure why you have to add washers underneath the thumbscrews.. Are you saying that thumb screw head is not wide enough to hold the pump bracket in place? If that's what you mean, then this is the first claim ever to be reported regarding that. Typically, we'd hear complain about the stand offs having a small gap from the mb so people would add washer between the board and the stand off to compensate for the gap to get a more solid base before installing the pump.

 

[BonzaiDuck]

I am also running on 347.52, and just as you, rebooting brought the power usage up to 40% again. I think there is a definite issue!

Ah! I just stumbled back into this thread!! You're the one! You're the one who confirms the same problem with the 980 cards -- the same problem I'm having with my 970s!

Did you put in a customer-support ticket on this with NVidia?

In recent days, they seem "all over it," while cautiously failing to conclude that it's their driver bug. I've now been elevated to "Level 2" support -- whatever that means.

What software are you using? Are you using AfterBurner to at least set the fan curve? It's always possible that you, I and anyone else who has mentioned this P-state power-consumption anomaly is using some particular type of software that causes the 40% power-consumption after startup, but has no effect when we cycle through SLI disable-enable.

But it is just as likely this is a driver bug that nobody noticed, or which too many ignored.

Maybe I can look at your rig-specs after I finish this post, but I'm interested in what processor you're using and whether your PCI-E slots used in SLI are PCI-E 2.0 or PCI-E 3.0 compliant. Mine would be the latter otherwise, but the Sandy Bridge only runs as PCI-E 2.0.

Nah! the PCI-E version ain't got nuthin' to do with it!