Airflow Recommendations

[Zarathustra[H]]

Hey all,

So I am working on a new build. Please see specifics of my build in my build thread.

Would appreciate everyone's take on what the lesser evil is of the below flow options.

I have dual 480's (60mm thick) in push pull up front, and a single 420 (45mm) in push pull up top.

(Image is stock corsair pic, not my build)

Here are the options I can think of:

Pro's: Flow works pretty well, hot air gets to rise the way nature intended.
Con's: Hot air from front radiators exits through top radiator, hot air from front radiators travels across internal components

Pro's: No hot air exits through any radiator, hot air gets to rise the way nature intended, no hot air from radiators travels across internal components
Con's: Poor flow, (inputs and outputs very uneven) very negative air pressure (results in extra dust).

Pro's: Positive pressure (less of a dust problem)
Con's: Flow not the best (inputs and outputs very uneven). Hot air forced downwards, counter to natural laws, hot air from both radiators travels across internal components.


Appreciate everyone's input.

 

[Dullard]

C

Cool air over the rads. After air cooling it might seem counter intuitive to blow the warm air in, but it's not going to be but a few degrees warmer than the outside air even on a really good radiator and still much cooler than the components. And don't worry so much about the warm air rising, the flow through the case will more than negate any rising air.

The idea behind the radiator is the temp difference between the intake side and coolant needs to be as great as possible, having warmed case air flowing outward over warm coolant is not as efficient.

Opinions on this are varied, there will undoubtedly be somebody with exactly the opposite as mine, but that's how I do it.

 

[Thunderdolt]

I second what Dullard said. When you're water cooling, what matters is getting the coolest air possible over the radiators. That means outside air.

That's also how I run my own 3-radiator 1000D build, so maybe I'm biased. I am running dual 140 exhaust fans instead of just 120s, but they rarely spin up over 600rpm based on motherboard temps.

 

[pendragon1]

^^^ thats why i picked C

 

[Smoked Brisket]

A is what I would do. I know in your build you have double rads in the front so A would give you the closest to balanced. Its not just about coolest air into every rad, its about the whole environment. The blocks arent all that generate heat.

 

[VanGoghComplex]

C for a number of reasons, but the main reasons being your concerns aren't really merited - hot air naturally rising becomes a moot point the instant a fan is involved, and air coming off a rad a system like yours is only going to be a hair warmer than ambient.

Strike those two off the cons list and C becomes the clear winner.

 

[Zarathustra[H]]

So,

I did B on my last build, and while it worked I found that I often had pretty poor airflow through my radiators, presumably due to them being starved for air. In other words my fans were likely often forced to spin faster than otherwise would be needed resulting in a system that is louder than desired.

Before starting this poll I was leaning towards A this time, but because the flow in and out are more closely aligned. I figured having good airflow was probably more important than the small temperature difference in the air going out through the top radiator.

That said C does have a strong following. It does have the same "uneven flow" issue, but maybe it doesn't matter has much in this flow direction?

I think my use of B all of these years has demonstrated that even static pressure fans are pretty bad at pulling negative pressure air through a radiator. At least with C you always have normal air pressure on the outside to pull from.

I still find myself leaning A though. Not decided yet.

 

[Thunderdolt]

Keep in mind that the glass sides aren't sealed at all. That provides a substantial leak path, which means you don't need as many exhaust fans in order to balance flow.

 

[hititnquitit]

I chose A, thats what ive always run and had the best overall results with. Top and back out. Front side and bottom in.
Id like to see if you have better results with C in that monster case. It may help keep your vrm temps in check better than A might.

 

[Tsumi]

A because more even flow, and that most cases have provisions for filtering on the front but not the top or rear. If you wanted to maximize cooling, coolant flow should go to top and rear radiators prior to the front radiators, essentially doing a counter-current flow scenario.

 

[Thunderdolt]

I may have to change my vote.

I've been running more CPU-heavy loads lately, and that means using my top loop. My 1000D is configured as C, and I'm thinking that this might be an under-exhausted setup. Because the exhaust fans are right up against the intake of the top loop, that pretty well chokes those off in terms of flow. When I run with the side panel open, I'm seeing CPU coolant temps that are 3C lower than when I have the side panel closed.

Maybe I just need fans on my CPU radiator that are higher flow than the ML120s.

 

[Zarathustra[H]]

I may have to change my vote.

I've been running more CPU-heavy loads lately, and that means using my top loop. My 1000D is configured as C, and I'm thinking that this might be an under-exhausted setup. Because the exhaust fans are right up against the intake of the top loop, that pretty well chokes those off in terms of flow. When I run with the side panel open, I'm seeing CPU coolant temps that are 3C lower than when I have the side panel closed.

Maybe I just need fans on my CPU radiator that are higher flow than the ML120s.

I suspect you might be right.

The flow in my loop is good resulting in good loop tempt to core temp delta T's, but the fan performance has been somewhat underwhelming.

Unless the Arctic specs for static pressure are inaccurate, which by subjective methods of feeling while they are blowing, I don't think they are, I'm going to guess that even on a case like this which has openings all over the place for air to escape, C is not a great flow orientation.

It's a major pain in the ass to re-orient the fans on this case though, as I will have to slide the top out, which means disconnecting some wires, and opening the loop and training it, so I am probably going to put it off for now.

There may be a 3080ti in my future if I like what they announce and the price is right, and Noctua has committed to sending me new fans, so once I get the fans and the new GPU and block, I may do everything all at once.

At that point I'll likely reverse the top fans and have them blow out, going to the A orientation.

 

[Thunderdolt]

It's a major pain in the ass to re-orient the fans on this case though, as I will have to slide the top out, which means disconnecting some wires, and opening the loop and training it, so I am probably going to put it off for now.

Is it any easier to slide out both top and front at the same time? If so, you could reverse fans on top by flipping one fan at a time from the rad-fan-tray sandwich, and then do the lower fans (currently pull, would become push) all at once. This way you could keep the top rad attached the whole time and not have to deal with disconnecting, draining, refilling, and re-purging. I think that's my plan actually, and I can just use my QDCs to yank the whole upper tray in 10 seconds. There was a killer Wiha tool deal the other day, so I'm just waiting on my order to arrive before I do this. (may as well play with new tools, right?)

 

[Zarathustra[H]]

Is it any easier to slide out both top and front at the same time? If so, you could reverse fans on top by flipping one fan at a time from the rad-fan-tray sandwich, and then do the lower fans (currently pull, would become push) all at once. This way you could keep the top rad attached the whole time and not have to deal with disconnecting, draining, refilling, and re-purging. I think that's my plan actually, and I can just use my QDCs to yank the whole upper tray in 10 seconds. There was a killer Wiha tool deal the other day, so I'm just waiting on my order to arrive before I do this. (may as well play with new tools, right?)

Yeah, I ma definitely planning on moving them together. It would be a total pain to touch the tubing between the top and front radiators.

Other things prevent the two from moving together.

My pump mount location blocks the drain lines from moving forward with the front radiators.

And since I put the Aquaero up top, my Aquabus Calitemp sensors cables are so short that I need to disconnect them from the temp sensors in order to slide the top forward.

So what I'd need to do is:

1.) Drain the loop.

2.) Remove the drain lines

3.) Remove the two reservoir return lines

4.) Disconnect the line from the CPU to the top radiator

5.) Disconnect all the carefully routed temp sensor wires

6.) Maybe disconnect power to the active PWM fan splitters and Aquaero and PWM lines to the pumps as well.

Once that is done I can slide both trays forward together, and start working on the fans.

Picture for reference:

 

[SticKx911]

Other. This feels like overkill. Get rid of the top radiator. Cool air in the filtered front. The rest set to blow out.

 

[primetime]

What kind of results so far? temps wise

 

[Zarathustra[H]]

What kind of results so far? temps wise

I'm achieving my target temps of 32C loop during full load, with an ambient of ~22C, but I am doing so with significantly higher fan speed than I anticipated needing based on the performance of my previous loop. Now with almost double the radiator capacity I expected the fans to be able to run slower.

Last night in Outer Worlds the fans briefly hit 100% in order to maintain my temps. I was not expecting that at all.

It must be an airflow issue. Either the fans are not performing as expected (I doubt it) or the way I have airflow set up is the problem.

 

[primetime]

I'm achieving my target temps of 32C loop during full load, with an ambient of ~22C, but I am doing so with significantly higher fan speed than I anticipated needing based on the performance of my previous loop. Now with almost double the radiator capacity I expected the fans to be able to run slower.

Must be an airflow issue. Either the fans are not performing as expected (I doubt it) or the way I have airflow set up is the problem.

Hos does it do in a real world heavy task that pushes all cores and the gpu to get the job done? speaking of which what kind of work is this monster going to do?

 

[Zarathustra[H]]

Hos does it do in a real world heavy task that pushes all cores and the gpu to get the job done? speaking of which what kind of work is this monster going to do?

I edited my post above with a gaming example.

Honestly, I never needed 24 cores. I just do a lot of I/O so I bought Threadripper for the extra PCIe lanes. I don't know how anyone survives with the base Intel and AMD lane counts. Just barely enough for a GPU and maybe something else small. I just bought this as it was the cheapest Threadripper I could buy without sacrificing per thread performance.

I've done some occasional video encodes, but that's probably the highest load the CPU will see.

The goal is to be a "no compromises" build. I want to hit the absolute max clocks on my GPU water cooling can provide, but I want it to be as quiet as possible.

Even gaming is not something I do a whole lot of though, but when I do, I want no compromises. I have so little time for games, that when I do play one, I want the experience to be as good as possible.

 

[SticKx911]

I'm achieving my target temps of 32C loop during full load, with an ambient of ~22C, but I am doing so with significantly higher fan speed than I anticipated needing based on the performance of my previous loop. Now with almost double the radiator capacity I expected the fans to be able to run slower.

Last night in Outer Worlds the fans briefly hit 100% in order to maintain my temps. I was not expecting that at all.

It must be an airflow issue. Either the fans are not performing as expected (I doubt it) or the way I have airflow set up is the problem.

Why is the fan curve set so low? 32c is not far over ambient. Your hardware can handle near 100c without a fit. I'm probably just looking at this from a different perspective, but no noise and medium temps is better to me than crazy noise and crazy low (albeit awesome) temps.

 

[Zarathustra[H]]

Why is the fan curve set so low? 32c is not far over ambient. Your hardware can handle near 100c without a fit. I'm probably just looking at this from a different perspective, but no noise and medium temps is better to me than crazy noise and crazy low (albeit awesome) temps.

At full load my overclocked Pascal Titan maintains a core temp of about 3-7C above loop temp, depending on the title/workload.

It maintains its maximum boost clocks only if I can keep it below 40C at full load. So that is my target temp.

Since 7C delta T is the worst case, and I want it to not exceed 39C, I set it to 39-7= 32

Since I game at 4K, and like Ultra settings, and the Pascal Titan X is just barely able to maintain 60fps at 4k in most demanding titles, I cant afford to lose even a single Mhz

No compromises!

 

[primetime]

I edited my post above with a gaming example.

Honestly, I never needed 24 cores. I just do a lot of I/O so I bought Threadripper for the extra PCIe lanes. I don't know how anyone survives with the base Intel and AMD lane counts. Just barely enough for a GPU and maybe something else small. I just bought this as it was the cheapest Threadripper I could buy without sacrificing per thread performance.

I've done some occasional video encodes, but that's probably the highest load the CPU will see.

The goal is to be a "no compromises" build. I want to hit the absolute max clocks on my GPU water cooling can provide, but I want it to be as quiet as possible.

Even gaming is not something I do a whole lot of though, but when I do, I want no compromises. I have so little time for games, that when I do play one, I want the experience to be as good as possible.

just imagine how bad ass this build will be in the winter cold months...Looks like a solid build to me and i highly doubt it be to noisy to my ears. I can understand always wanting to improve results, its just the way we are.

 

[Thunderdolt]

Yeah, I ma definitely planning on moving them together. It would be a total pain to touch the tubing between the top and front radiators.

Other things prevent the two from moving together.

My pump mount location blocks the drain lines from moving forward with the front radiators.

And since I put the Aquaero up top, my Aquabus Calitemp sensors cables are so short that I need to disconnect them from the temp sensors in order to slide the top forward.

So what I'd need to do is:

1.) Drain the loop.

2.) Remove the drain lines

3.) Remove the two reservoir return lines

4.) Disconnect the line from the CPU to the top radiator

5.) Disconnect all the carefully routed temp sensor wires

6.) Maybe disconnect power to the active PWM fan splitters and Aquaero and PWM lines to the pumps as well.

Once that is done I can slide both trays forward together, and start working on the fans.

Picture for reference:

Oh yeah, that definitely does look like a pain.

They wouldn't be pretty, but have you considered putting service loops into the tubing to let you move the radiators without disconnecting? I've done that with my wiring so that I don't have to disconnect it when I want to move the rads out.

 

[Thunderdolt]

Why is the fan curve set so low? 32c is not far over ambient. Your hardware can handle near 100c without a fit. I'm probably just looking at this from a different perspective, but no noise and medium temps is better to me than crazy noise and crazy low (albeit awesome) temps.

He's talking about the temp of the coolant in the loop, not the component temps. When the coolant hits 100C, the components would be baking at 130+.

 

[Zarathustra[H]]

He's talking about the temp of the coolant in the loop, not the component temps. When the coolant hits 100C, the components would be baking at 130+.

Not to mention that you don't want your loop temp to ever exceed 50C as that is not good for the pumps and can compromise fitting seals.

 

[Zarathustra[H]]

Oh yeah, that definitely does look like a pain.

They wouldn't be pretty, but have you considered putting service loops into the tubing to let you move the radiators without disconnecting? I've done that with my wiring so that I don't have to disconnect it when I want to move the rads out.

I have considered it. I can't seem to figure out how to fit it though. There really isn't any other way to route the drain lines or to place the pump.

(And yes, I know how ridiculous it sounds to have a space problem in the 1000D )

 

[Smoked Brisket]

I have considered it. I can't seem to figure out how to fit it though. There really isn't any other way to route the drain lines or to place the pump.

(And yes, I know how ridiculous it sounds to have a space problem in the 1000D )

Uh, yea, you are the only person that could possibly have a space problem in this freaking cavernous case. I mean that as a compliment, brother.

 

[Thunderdolt]

I have considered it. I can't seem to figure out how to fit it though. There really isn't any other way to route the drain lines or to place the pump.

(And yes, I know how ridiculous it sounds to have a space problem in the 1000D )

For the tubing, you'd just make a 4-6" diameter coil for each mobile section. Between front rads and res, instead of a straight shot like you have now, you'd have a pair of coils. Between top rad and CPU, another coil. With swivel fittings, you could unfurl the coils all the way to straight without having them kink as long as you're careful when moving the rads. It's very messy looking but it ends up being functional.

And yeah, I actually hear you about running out of space in this case. It's huge, but it is shockingly inefficient. It is basically a 500D that someone scaled up in CAD but without adding in any extra features to handle the extra stuff that goes inside. For example, if I'm going to run twice as many rads up front, then they should have provided twice as much space to route the wiring. And there's also stuff that doesn't simply scale, such as providing space for mounting up multiple fan controllers as just one example. Or how about the lack of exhaust options? It has space for 16 filtered 120mm intake fans, but there's only space for 1.5 120mm exhaust fans. And can someone remind me why they put that 3.5in drive cage in there again? Oh, and I almost forgot the double doored stuff on the back of it. Without that, the entire motherboard firewall thing could have been set 2" deeper and that would have created enough space for two more exhaust fans.

 

[SticKx911]

He's talking about the temp of the coolant in the loop, not the component temps. When the coolant hits 100C, the components would be baking at 130+.

Ahhhh. I totally missed that. I was thinking a cpu/gpu temp of 32 at load. My bad. Carry on.

 

[Thunderdolt]

I ran a limited test today. I ran Folding sets on the CPU alone and did that with the GPU fans stopped and both chassis fans maxed (minimal internal pressure) and then again with the GPUs at full blast and only one chassis fan running (maximum internal pressure). CPU was pulling the same 260-280W in both states.

What I discovered was that the CPU coolant temp was identical in both configurations. This now makes me think that configuration C is actually ok.

I will try again with full GPU loads as well. The reason I skipped those for today was that there aren't many GPU Folding sets and I wanted something reasonably consistent.