Custom Water Loop Performance Increases

[KarathKasun]

8 minutes ago, Enderman said:

Air coolers are limited by the size of the heatsink, that's what determines the surface area for heat dissipation.

A custom loop is not limited because you can simply add more radiators.

And when you use the same fans on both, obviously the cooler with more surface area wins.

 

It's not hard...

same airflow + more surface area = lower temps.

You seem to have trouble understanding that.

You seem to have no experience with cooling.  More surface area does not = lower temps beyond a specific point.  Beyond ~480 worth of rads with normal "PC type" fans and you are not significantly changing your Δ T for a CPU.  With extreme high flow fans you just need less rad to get there.  You cant keep adding loop cooling capacity and actually realize meaningfully lower CPU temps.

 

The practical limits for both result in pretty much the exact same Δ T.  The only benefit with water is that your PC isnt circulating ~1000CFM of air and sounding like a freight train.

 

As a bonus, ALL temps are lower with massive air setups.  Board temps can be 20c lower or more.

[Ttnuagmada]

23 minutes ago, Enderman said:

Which is why you use liquid metal if you're going for the highest possible overclock.

Regardless of whether you use a heatsink or custom loop or whatever.

 

That doesn't change the fact that more surface area = more heat dissipation.

 

If the heat isn't getting there quickly enough to dissipate then the surface area isn't going to matter past a certain point. Liquid metal isn't a cure-all.  This is why GPU temps drop by a massive amount under water cooling vs a CPU.

[KarathKasun]

18 minutes ago, Enderman said:

Which is why you use liquid metal if you're going for the highest possible overclock.

Regardless of whether you use a heatsink or custom loop or whatever.

 

That doesn't change the fact that more surface area = more heat dissipation.

 

 

I never said anything about it being worth the money or effort for the performance improvement, that's up to the enthusiast.

The point is that a custom loop allows for more radiators which means more surface area, which means lower temperatures.

Go take a look at jayztwocents and singularitycomputers on youtube for the numbers.

It is limited.  Liquid metal is not infinitely better than TIM.  Not to mention the thermal resistance of the silicon itself.

 

Because the thermal interface is still the same for both, the limit is the same.

 

You CAN hit the TIM limit for LM with both given enough effort.

[Enderman]

30 minutes ago, KarathKasun said:

You seem to have no experience with cooling.  More surface area does not = lower temps beyond a specific point.  Beyond ~480 worth of rads with normal "PC type" fans and you are not significantly changing your Δ T for a CPU.  With extreme high flow fans you just need less rad to get there.  You cant keep adding loop cooling capacity and actually realize meaningfully lower CPU temps.

 

The practical limits for both result in pretty much the exact same Δ T.  The only benefit with water is that your PC isnt circulating ~1000CFM of air and sounding like a freight train.

 

As a bonus, ALL temps are lower with massive air setups.  Board temps can be 20c lower or more.

As I said before, the 'just add more cfm to a heatsink and it will be like a custom loop with lower cfm' is a stupid argument, because you can always add more cfm to a radiator too.

Obviously at a certain point the difference between them will be negligible but nobody is going to put 5000cfm fans i their computer (except linus, he did that already in a video)

So yeah, a custom loop with more surface area allows you to be able to get better temps than any air cooler or AIO with the same fans/airflow, which is what I've literally been saying since the beginning.

 

The variable here is the surface area, stop trying to change other things like fans or TIM to cater to your argument.

Nobody would choose between a crappy fans + custom loop vs good fans + air cooler.

People decide whether they want to go with a custom loop or air cooler or AIO, THEN choose thermal paste and fans.

The choice of thermal paste and fans depends on what kind of noise and performance the person wants, not "oh I bought a heatsink so I need 300cfm fans to beat a 420mm radiator with 100cfm fans" lol that's just ridiculous

 

26 minutes ago, Ttnuagmada said:

If the heat isn't getting there quickly enough to dissipate then the surface area isn't going to matter past a certain point. Liquid metal isn't a cure-all. 

23 minutes ago, KarathKasun said:

It is limited.  Liquid metal is not infinitely better than TIM.  Not to mention the thermal resistance of the silicon itself.

 

Because the thermal interface is still the same for both, the limit is the same.

 

You CAN hit the TIM limit for LM with both given enough effort.

If you think the thermal interface is a bottleneck when maybe you should google what phase change cooling is.

If it was a bottleneck then it wouldn't be possible to get lower temps than liquid or air.

Spoiler alert, phase change can do -50C or below.

Oh, and did I mention LN2 and liquid helium?

...

 

 

[KarathKasun]

3 minutes ago, Enderman said:

As I said before, the 'just add more cfm to a heatsink and it will be like a custom loop with lower cfm' is a stupid argument, because you can always add more cfm to a radiator too.

Obviously at a certain point the difference between them will be negligible but nobody is going to put 5000cfm fans i their computer (except linus, he did that already in a video)

So yeah, a custom loop with more surface area allows you to be able to get better temps than any air cooler or AIO with the same fans/airflow, which is what I've literally been saying since the beginning.

 

The variable here is the surface area, stop trying to change other things like fans or TIM to cater to your argument.

Nobody would choose between a crappy tim + custom loop vs good tim + air cooler.

People decide whether they want to go with a custom loop or air cooler or AIO, THEN choose thermal paste and fans.

The choice of thermal paste and fans depends on what kind of noise and performance the person wants, not "oh I bought a heatsink so I need 300cfm fans to beat a 420mm radiator with 100cfm fans" lol that's just ridiculous

 

If you think the thermal interface is a bottleneck when maybe you should google what phase change cooling is.

If it was a bottleneck then it wouldn't be possible to get lower temps than liquid or air.

Spoiler alert, phase change can do -50C or below.

Oh, and did I mention LN2 and liquid helium?

...

 

 

Sub ambient is a wholly different ball game.  ΔT is generally HUGE when you go sub ambient, because you are just driving down the "ambient" side of the equation so much that a higher ΔT simply does not matter.

 

The fact that you are grasping for straws outside of the thermodynamic scope shows your lack of data.

[Enderman]

1 minute ago, KarathKasun said:

Sub ambient is a wholly different ball game.  ΔT is generally HUGE when you go sub ambient.  The fact that you are grasping for straws outside of the thermodynamic scope shows your lack of data.

Try reading the first half of the post again, instead of skipping the parts that prove you wrong.

[KarathKasun]

3 minutes ago, Enderman said:

Try reading the first half of the post again, instead of skipping the parts that prove you wrong.

You addressed nothing.  You changed the TIM in the argument, not me.  I even gave you that, but you have presented nothing other than "nuh uh".

 

Are you stepping back from your claim and saying WC is only about noise now?

[Ttnuagmada]

13 minutes ago, Enderman said:

As I said before, the 'just add more cfm to a heatsink and it will be like a custom loop with lower cfm' is a stupid argument, because you can always add more cfm to a radiator too.

Obviously at a certain point the difference between them will be negligible but nobody is going to put 5000cfm fans i their computer (except linus, he did that already in a video)

So yeah, a custom loop with more surface area allows you to be able to get better temps than any air cooler or AIO with the same fans/airflow, which is what I've literally been saying since the beginning.

 

The variable here is the surface area, stop trying to change other things like fans or TIM to cater to your argument.

Nobody would choose between a crappy fans + custom loop vs good fans + air cooler.

People decide whether they want to go with a custom loop or air cooler or AIO, THEN choose thermal paste and fans.

The choice of thermal paste and fans depends on what kind of noise and performance the person wants, not "oh I bought a heatsink so I need 300cfm fans to beat a 420mm radiator with 100cfm fans" lol that's just ridiculous

 

If you think the thermal interface is a bottleneck when maybe you should google what phase change cooling is.

If it was a bottleneck then it wouldn't be possible to get lower temps than liquid or air.

Spoiler alert, phase change can do -50C or below.

Oh, and did I mention LN2 and liquid helium?

...

 

 

Of course sub-ambient is going to offer improvements over ambient solutions. Not rocket science.  You aren't making the point that you think you're making. You're comparing apples to oranges.

[Enderman]

15 minutes ago, KarathKasun said:

You addressed nothing.  You changed the TIM in the argument, not me.  I even gave you that, but you have presented nothing other than "nuh uh".

 

Are you stepping back from your claim and saying WC is only about noise now?

Well since you refuse to go look at the youtubers that have actually done tests, like I said 5 times before, I'll try explaining it to you one last time as simply as possible.

 

1) In your magical theoretical world where you can have infinite cfm, yes the performance of a heatsink and many large radiators would be basically identical, since the determining factor of heat transfer in this case is the airflow velocity and not the surface area.

 

2) In the real world, however, where normal people have their PC beside them and use normal computer fans with <100cfm, surface area is the main determining factor.

 

3) So, when using the same fans and airflow, the larger surface area of many radiators offers lower temperatures, because heat dissipation is (primarily) a function of surface area and airflow velocity.

 

4) A custom loop allows for this option of larger/more radiators, while a heatsink and AIO does not, they have fixed surface area.

 

5) Therefore, for maximum performance, larger surface area is the benefit offered by a custom loop. Fan choice, tim choice, etc. is all applied as the secondary factor on top of that.

 

eg. (A)*(B) will perform worse than (2A)*(B), where B is the airflow and A is the area. Not hard at all, just simple math.

If the user desires more performance and they are ok with louder fans, they can do (2A)*(2B) which will perform better than (A)*(2B)

 

Hopefully that helps you understand, have a good night

[KarathKasun]

Just now, Enderman said:

Well since you refuse to go look at the youtubers that have actually done tests, like I said 5 times before, I'll try explaining it to you one last time as simply as possible.

 

1) In your magical theoretical world where you can have infinite cfm, yes the performance of a heatsink and many large radiators would be basically identical, since the determining factor of heat transfer in this case is the airflow velocity and not the surface area.

 

2) In the real world, however, where normal people have their PC beside them and use normal coputer fans with <100cfm, surface area is the main determining factor.

 

3) So, when using the same fans and airflow, the larger surface area of many radiators offers lower temperatures, because heat dissipation is (primarily) a function of surface area and airflow velocity.

 

4) A custom loop allows for this option of larger/more radiators, while a heatsink and AIO does not, they have fixed surface area.

 

5) Therefore, for maximum performance, larger surface area is the benefit offered by a custom loop. Fan choice, tim choice, etc. is all applied as the secondary factor on top of that.

 

Hopefully that helps you understand, have a good night

Custom heatsinks are a thing.  And you ignore the thermodynamic limits, falling back to just repeating "more surface area" with your fingers in your ears.

 

Gotchya.

[Ttnuagmada]

En

7 minutes ago, Enderman said:

Well since you refuse to go look at the youtubers that have actually done tests, like I said 5 times before, I'll try explaining it to you one last time as simply as possible.

 

1) In your magical theoretical world where you can have infinite cfm, yes the performance of a heatsink and many large radiators would be basically identical, since the determining factor of heat transfer is the airflow velocity and not the surface area.

 

2) In the real world, however, where normal people have their PC beside them and use normal coputer fans with <100cfm, surface area is the main determining factor.

 

3) So, when using the same fans and airflow, the larger surface area of many radiators offers lower temperatures, because heat dissipation is (primarily) a function of surface area and airflow velocity.

 

4) A custom loop allows for this option of larger/more radiators, while a heatsink and AIO does not, they have fixed surface area.

 

5) Therefore, for maximum performance, larger surface area is the benefit offered by a custom loop. Fan choice, tim choice, etc. is all applied as the secondary factor on top of that.

 

Hopefully that helps you understand, have a good night

Since you're basing your argument on Youtube videos, let me give you a real world scenario that I've dealt with personally:

 

3770K@4.9ghz 1.416v cooled by an NH-D14. Max temps with IBT max: 72c. (delided with TGC between IHS and die)

 

Now it's being cooled by: HWL GTX 560, 480 and SR2 280 with 22 Be Quiet! SW3's and 2 D5's: max temp: 68C

 

That's the kind of cooling reduction you're looking at going from a top-end air cooler to a water loop bigger than anything you'll ever build. Where was the bottleneck?  The delid.

 

Understand?

[KarathKasun]

2 minutes ago, Ttnuagmada said:

En

Since you're basing your argument on Youtube videos, let me give you a real world scenario that I've dealt with personally:

 

3770K@4.9ghz 1.416v cooled by an NH-D14. Max temps with IBT max: 72c. 

 

Now it's being cooled by: HWL GTX 560, 480 and SR2 280 with 22 Be Quiet! SW3's: max temp: 68C

 

That's the kind of cooling reduction you're looking at going from a top-end air cooler to a water loop bigger than anything you'll ever build.

Back to thread topic, what kind of fans were you using on the NH-D14?

 

Was it with the stock, nearly silent, fans?

[Enderman]

5 minutes ago, Ttnuagmada said:

En

Since you're basing your argument on Youtube videos, let me give you a real world scenario that I've dealt with personally:

 

3770K@4.9ghz 1.416v cooled by an NH-D14. Max temps with IBT max: 72c. 

 

Now it's being cooled by: HWL GTX 560, 480 and SR2 280 with 22 Be Quiet! SW3's: max temp: 68C

 

That's the kind of cooling reduction you're looking at going from a top-end air cooler to a water loop bigger than anything you'll ever build.

Just fyi, the SW3s have much lower cfm (anbd mmh2o) than the noctuas on the NH D14, so essentially you're trading performance for lower noise.

 

Also the 3770K isn't a very high TDP CPU which is why the temp difference isn't big. Higher heat output would make better use of that custom loop and would make a bigger temp difference.

 

Either way, I never said the performance difference was huge. My argument was that a custom loop with more area would get lower temperatures, which you just proved.

Thanks  

[Ttnuagmada]

9 minutes ago, Enderman said:

Just fyi, the SW3s have much lower cfm than the noctuas on the NH D14, so essentially you're trading performance for lower noise.

 

 

Wow it's almost like comparing 2 NF-F12's on a heatsink to 22 BQ SW3's in p/p on 3 60mm radiators is the dumbest thing someone could possibly come up with. Also, the fact that you think CFM is what matters just goes to show that you have no business trying to tell anyone anything about water cooling.

 

9 minutes ago, Enderman said:

Also the 3770K isn't a very high TDP CPU which is why the temp difference isn't big. Higher heat output would make better use of that custom loop and woud make a bigger temp difference.

 

Actually, even my lowly 3770K is limited by the heat transfer from the core to the loop. Way to actually prove my point. Like I've already said, if you want to see how heat transfer matters, look at the temp drops that GPU's get from water cooling due to the much larger surface area.

 

9 minutes ago, Enderman said:

Either way, I never said the performance difference was huge. My argument was that a custom loop with more area would get lower temperatures, which you just proved.

Thanks

 

No, like more than one of us has told you already, surface area isn't the bottleneck. My XE 360 cooled my CPU just as well as my monster loop using the same fans. See if you can figure out why that might be.

[Ttnuagmada]

7 minutes ago, KarathKasun said:

Back to thread topic, what kind of fans were you using on the NH-D14?

 

Was it with the stock, nearly silent, fans?

just the stock fans

[KarathKasun]

13 minutes ago, Ttnuagmada said:

Way to actually prove my point. Like I've already said, if you want to see how heat transfer matters, look at the temp drops that GPU's get from water cooling due to the much larger surface area.

Yeah, its difficult to get a NH-D14 on a GPU, but it is possible.  And temps are pretty much the same as water with only a modest upgrade in fans.  Most people will not go through the riser + case modding to do this though.

 

GPUs are primarily space constrained with air cooling.

[JT241]

So.....liquid is better...? lol For clarity, I'm not doing anything SUPER crazy by any measure. I'm only looking at like a simple custom loop with standard parts not unlimited rads vs like a hefty noctua cooler or something. Thanks again for theinformation despite how sidetracked this may have gotten...

[KarathKasun]

3 minutes ago, JT241 said:

So.....liquid is better...? lol For clarity, I'm not doing anything SUPER crazy by any measure. I'm only looking at like a simple custom loop with standard parts not unlimited rads vs like a hefty noctua cooler or something. Thanks again for theinformation despite how sidetracked this may have gotten...

NH-D14 is pretty much as good as it gets on air, and it will last forever with minimal maintenance.

 

Temp difference and OC difference will be minimal.

[Mick Naughty]

I’d take a custom loop. Better temps and less noise and not as many limitations. 

 

Compared to a mediocre aio. 

You want better performance, get a better chip. 

[Enderman]

6 hours ago, JT241 said:

So.....liquid is better...? lol For clarity, I'm not doing anything SUPER crazy by any measure. I'm only looking at like a simple custom loop with standard parts not unlimited rads vs like a hefty noctua cooler or something. Thanks again for theinformation despite how sidetracked this may have gotten...

Well the performance will depend on how many rads you use, but the more rads you have the slower you can run your fans for lower noise if you prefer quiet over max performance.

 

Unfortunately some people seem to think that it performs the same as a heatsink, which is completely wrong if you take a look at any video of a custom loop with more than a 360 rad. Just ignore them, they don't understand how cooling works. Both surface area and airflow will determine performance so make sure you choose good fans regardless of what cooling you use.

 

Don't do liquid cooling just for performance though since it isn't a huge improvement (unless you do somethingf like jayztwocents or singularitycomputers), do it if you have money to blow and want to make something that looks cool and is fun to build.

[Blai5e]

Thanks for the entertaining thread people.

[WoodenMarker]

19 hours ago, KarathKasun said:

you are not significantly changing your Δ T for a CPU.

A static temperature difference isn't a consistent way to measure cooling performance. As load rises, the temp difference increases when comparing the same set of coolers.

At the end of the day, coolers don't perform exactly the same. One is better and the other is worse.  

18 hours ago, Ttnuagmada said:

My XE 360 cooled my CPU just as well as my monster loop using the same fans. 

Just because the load isn't heavy enough to separate the performance doesn't mean they're the same.

It's similar to how the temp difference at idle could be the same for a Hyper 212 Evo and the NH-D14 yet the NH-D14 is still better.

You can see an example here from CoolingTechnique: http://www.coolingtechnique.com/recensioni/air-cooling/dissipatori/1446-recensione-noctua-nh-d15.html?start=6

Spoiler

When the load is increased, it becomes clear that custom loops outperform the highest end air coolers. The better performers can keep cooling where weaker coolers fail to provide adequate cooling. 

https://youtu.be/_I--zROoRws?t=46

19 hours ago, KarathKasun said:

It is limited.  Liquid metal is not infinitely better than TIM.

I'm assuming you meant thermal paste since TIM is just thermal interface material which includes liquid metal. 

18 hours ago, Enderman said:

If you think the thermal interface is a bottleneck when maybe you should google what phase change cooling is.

Oh, and did I mention LN2 and liquid helium?

Cooling is entirely determined by thermal interfaces which includes anything that can transfer heat such as LN2.

When it comes to importance, the cooling interfaces closest to the heat source are exponentially more important than the cooling interfaces further away. 

So interface between the die and IHS > interface between IHS and cooler base > interface between cooler base > rest of cooler > etc.

17 hours ago, JT241 said:

So.....liquid is better...? lol For clarity, I'm not doing anything SUPER crazy by any measure. I'm only looking at like a simple custom loop with standard parts not unlimited rads vs like a hefty noctua cooler or something. Thanks again for theinformation despite how sidetracked this may have gotten...

A custom loop can be better if you have enough surface area. Whether that's what you're aiming for or will utilize is up to you. 

For most users, there isn't any need to go further than a good air cooler like the NH-U12S. 

[Ttnuagmada]

22 hours ago, WoodenMarker said:

 

Just because the load isn't heavy enough to separate the performance doesn't mean they're the same.

It's similar to how the temp difference at idle could be the same for a Hyper 212 Evo and the NH-D14 yet the NH-D14 is still better.

You can see an example here from CoolingTechnique: http://www.coolingtechnique.com/recensioni/air-cooling/dissipatori/1446-recensione-noctua-nh-d15.html?start=6

I was specifically talking about non-HEDT chips and specifically talking about anything much more than an NH-D14/D15. I think my point is still valid. Heat transfer becomes the main bottleneck because you're dealing with small, high frequency silicon, and getting the heat away from it becomes the main bottleneck once you're in NH-D15 territory.  This is why delidding has such a massive effect. By using liquid metal, you are improving the situation by a large amount, but it is still ultimately the bottleneck. That's why going from an NH-D15 to the biggest custom loop you can make is only going to give you minor improvements. 

 

 

[WoodenMarker]

4 hours ago, Ttnuagmada said:

I was specifically talking about non-HEDT chips and specifically talking about anything much more than an NH-D14/D15. I think my point is still valid. Heat transfer becomes the main bottleneck because you're dealing with small, high frequency silicon, and getting the heat away from it becomes the main bottleneck once you're in NH-D15 territory.  This is why delidding has such a massive effect. By using liquid metal, you are improving the situation by a large amount, but it is still ultimately the bottleneck. That's why going from an NH-D15 to the biggest custom loop you can make is only going to give you minor improvements. 

Yes, it's only a minor improvement if you're referring to the oddly specific situation with a custom loop on non-HEDT without delidding. The 9900k is technically not HEDT. 

If you're actually taking advantage of said cooling for oc'ing, it's either going to be much quieter with a custom loop due to the NH-D15 barely being able to keep up or failing outright. In that case, a custom loop being the only adequate solution for cooling would be much more than a minor improvement over an inadequate cooler. 

[Ttnuagmada]

1 hour ago, WoodenMarker said:

Yes, it's only a minor improvement if you're referring to the oddly specific situation with a custom loop on non-HEDT without delidding. The 9900k is technically not HEDT. 

If you're actually taking advantage of said cooling for oc'ing, it's either going to be much quieter with a custom loop due to the NH-D15 barely being able to keep up or failing outright. In that case, a custom loop being the only adequate solution for cooling would be much more than a minor improvement over an inadequate cooler. 

No, I'm talking about the "oddly specific situation" of every non-HEDT Intel chip made in the past several years delided or not. The delidding merely improves the situation, it does not remove it as the primary bottleneck, while the act does confirm what the bottleneck is.  The 9900K is probably the exception here, but the NH-D14/D15 has been doing perfectly fine with maxed out Intel CPU's for years. I dropped all of 3C going from an NH-D14 to custom water with a delidded 4.9ghz 3770K running 1.416v. I'm not sure why you think there are many non-HEDT CPU's using more power than that would be. Not that it matters, as it quickly becomes the bottleneck in a custom loop for a 9900K as well. 

 

GPU behavior under water is easy confirmation of what I'm talking about. I mean it's not rocket science. a maxed out GPU using significantly more power is going to have like an 8-10C temp delta while a maxed out, delided CPU is going to be 30C+ and will vary largely based on how much heat/power it's putting out. You don't get that kind of delta variance with a GPU.