Compact Liquid Cooling vs Air !!

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

TheThirdMan

Member
Jul 5, 2011
113
11
81
Because u dont have enterprise class fans blowing though the board like it was designed when fitted into a U rack, which is basically a standard ATX case laid flat.

U cant tell me 2 dinky 140mm will come anywhere close to a ducted Enterprise case setup...
...

There all built like that... even Dells.
Theres a big reason why Dell, HP, and all the other server vendors builds them like that.

Registered ECC ram is NOTHING like desktop unregistered ram... they get super hot as work is being done on them.
There is no Fancy heat pipe cooler on a supermicro / Tyan / intel enterprise class board... There is no blingy stock sink.


Intel-Xeon-E5-CPUs-Get-Five-New-LGA-2011-Motherboards-from-Tyan-2.jpg

(U see how everything is setup to be ducted in a wind tunnel down the board... the ram and NB sink Align in the same direction as wind would travel from high static fans)

So please dont tell people to build servers like you would a desktop.... your asking for FAIL b4 the project even started.



No its not AIO HATE, its the pure fact on the machine being built WRONG in design to how its heat sinks were made.



Again... LOOK at that server case designed to hold a board meant for more then 1 CPU.
Look at the fan layout on how its set.
Those fans are also 2200-2800RPM class, and have way more static pressure then most 140mm fan.

If u say its building a server wrong... then your telling DELL / HP and almost every other enterprise series vendor they are building there servers WRONG.

...

And i still needed some fans.


Stock cooling on a gaming board is much beefier then a server board.
Stock cooling on a gaming board has evolved a boat load... we used to get them like server boards.
Desktop != Enterprise...

Enterprise is about stability (they cant hear you complain about noise)...
Desktop is about Stability with Noise.... if you cant hear everything in uber bit audio over the fan noise... FAIL...

If u had a fan layout just on the RAM... and another on the ALU sinks... u can cool a enterprise board.... however with all those addition of fans... you could of plan'd a little bit better by replacing the stock sinks, and putting ones designed to fill the role your asking for.



just who do u think your talking to?
Theres a reason why everyone voted me the CnC Mod...
ROFL..

...
Expensive Lesson Aigo learned....
You DONT BUILD SERVERS LIKE U DO A DESKTOP!

1: Xeon ≠ server
2: Dell and all those other manufacturers build workstations (as i wrote in my first post) with slow spinning large fans. They also use small heatsink/fans on the CPUs, usually close to stock cooling. I have one at work. It doesn't use 80mm high RMP fans. You keep talking about rackmount servers as if that's what we're talking about. We're not. Which leads me to
3: Server ≠ rackmount- there are many servers which are not rackmount and use large slow spin fans instead of small high spin fans, in a desktop enclosure.

aigomorla said:
just who do u think your talking to?

I'm tempted to say a 14 year old boy on a power trip on an online forum where there's just friendly talk about cooling a CPU till you came along? But I don't want to offend so I'll just say someone who got the wrong end of the stick i guess...
 

TheThirdMan

Member
Jul 5, 2011
113
11
81
While the Toms didn't really push that 3930k (most do go to 4.4Ghz so 4.25 is a little low) the results do match with what I have seen elsewhere. That is the All in one coolers just don't seem to really perform like custom water. I have a bundle of theories as to why and its probably combination of a lot of things:
1) The radiators are aluminium and tend to be thin or too small. When we are talking about needing 2x 120mm thick radiators they are using 2x120mm thin or 1x140 thick. Its just not enough radiator space and the quality of the radiators is lacking both in material and probably in the manufacturer of the radiator itself. The radiators are often low FPI which means they need high speed fans and hence its quite a noisy solution.
2) The tubing is really very thin. While there isn't much drop in performance with thin tubing there is a notable drop from 1/2" to 10/8 and some of the AIO's are using this notably thin tubing. None of them seem to use 7/16" or 1/2" which is far more common in the custom world
3) The pump is right on top of the CPU which could well be heating that package up a bit. In addition its vibrating onto the CPU and that could reduce the thermal capabilities of the pump itself. The pump itself is pretty small and quite likely not achieving a good flow rate, which while relatively minor in the overall scheme of things compounds with all the other decisions made.
4) The quality of the block really makes a big difference in the cooling potential of a loop. Best to worst waterblocks on a CPU can differ by about 10C and these are often aluminium and in addition not made by the well known to be great water block manufacturers. There is every chance they just plain underperform.
5) The liquid used isn't pure distilled water, it has to contain all sorts of anti corrosives due to the aluminium in the loop. This does degrade cooling potential a little.
6) The fans they use often aren't really that great for radiators.

The more reviews of the AIO's I see the more I am disappointed with their cooling potential. I have a 3930k at 4.4 Ghz 1.3V and it tops out in prime 95 at 55C at the package level. I do that with fans running at 800rpm (basically silent, quieter than any hard drive). So a custom loop can achieve this high performance well above what air can, but the AIO's almost always seem to show they aren't really much better than high end air, they are often more expensive and less performant. I don't think Tom's is alone in this testing or that 0.15Ghz makes much difference in the performance of AIOs, I think the cheap components are really hurting the performance of the all in one loops.

I suspect this is true. But there's still lots of occasions you'd want to use it over air coolers. I'm using it on two 60w xeons at stock at the moment. I'm hoping to get an SR-2 when they're available on UK ebay and having two huge air coolers hanging off a motherboard doesn't sound like a great idea.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,894
3,247
126
1: Xeon ≠ server
2: Dell and all those other manufacturers build workstations (as i wrote in my first post) with slow spinning large fans. They also use small heatsink/fans on the CPUs, usually close to stock cooling. I have one at work. It doesn't use 80mm high RMP fans. You keep talking about rackmount servers as if that's what we're talking about. We're not. Which leads me to
3: Server ≠ rackmount- there are many servers which are not rackmount and use large slow spin fans instead of small high spin fans, in a desktop enclosure.



I'm tempted to say a 14 year old boy on a power trip on an online forum where there's just friendly talk about cooling a CPU till you came along? But I don't want to offend so I'll just say someone who got the wrong end of the stick i guess...

You specified dual xeon and you call that not a server?
in most likelyhood dual Xeon's also use Reg. ECC Ram... which are different then consumer non registered ECC ram... also heat up a lot more then non registered ECC ram.
2 Logical CPUS + ECC Ram => does not equal server?

I specifically told you the differences between a dual xeon board, as well shown u examples in which i undertook to cool and silence a dual xeon board and you think im having a power trip?

First off you know dual xeon boards are EATX case to begin with?

Your getting dual xeon confused with a single xeon workstation system which in all terms is exactly the same as a desktop.

Can i point you to the release of the EVGA's SR-2.. when people thought consumer means was sufficient to cool it down yet found out the hardway its fail?
 
Last edited:

TerryMathews

Lifer
Oct 9, 1999
11,464
2
0
You specified dual xeon and you call that not a server?

I specifically told you the differences between a dual xeon board, as well shown u examples in which i undertook to cool and silence a dual xeon board and you think im having a power trip?

First off you know dual xeon boards are EATX case to begin with?

Your getting dual xeon confused with a single xeon workstation system which in all terms is exactly the same as a desktop.

What about my dual Xeon? I don't know of many servers that advertise Quad SLI. Or overclocking.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,894
3,247
126
What about my dual Xeon? I don't know of many servers that advertise Quad SLI. Or overclocking.

SR-2 as i told you was a consumerized Enterprise platform.
And watch... if u read problems with it, it will state how hot the NB gets.
Also the VREG's controlling your memory voltages will get hot, which later limits your overclocking on the ram.
Also if u use ECC REG ram, it fails cuz ECC REG's cant overclock for squat....
(the list goes on and on.... )

This was fixed by block vendors... and EVGA pretty much left them to do it.
Because their methodology was.. the board was either going to have a LN2 pot over it keeping the entire area cold... or people with that much money will WATERCOOL it, which most did.
It was a 700 dollar board.... watercooling it was an accessory at that price point.
 

TerryMathews

Lifer
Oct 9, 1999
11,464
2
0
SR-2 as i told you was a consumerized Enterprise platform.
And watch... if u read problems with it, it will state how hot the NB gets.
Also the VREG's controlling your memory voltages will get hot, which later limits your overclocking on the ram.
Also if u use ECC REG ram, it fails cuz ECC REG's cant overclock for squat....
(the list goes on and on.... )

This was fixed by block vendors... and EVGA pretty much left them to do it.
Because their methodology was.. the board was either going to have a LN2 pot over it keeping the entire area cold... or people with that much money will WATERCOOL it, which most did.
It was a 700 dollar board.... watercooling it was an accessory at that price point.

Nice stealth edit.

ETA: I've had no overheating problems and my system is fairly fully loaded and overclocked quite a bit.
 
Last edited:

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,894
3,247
126
Nice stealth edit.

ETA: I've had no overheating problems and my system is fairly fully loaded and overclocked quite a bit.

touch and feel your alu sink while under load. :D
 

YBS1

Golden Member
May 14, 2000
1,945
129
106
While the Toms didn't really push that 3930k (most do go to 4.4Ghz so 4.25 is a little low) the results do match with what I have seen elsewhere. That is the All in one coolers just don't seem to really perform like custom water.
I never meant to give the impression they perform as well as quality custom water. I honestly can't imagine a 3930K that would have trouble hitting 4.25, that's why I commented on it. That is well within the thermal performance envelope of an air cooler and wouldn't allow a decent AIO to differentiate itself much from the air cooler.

4) The quality of the block really makes a big difference in the cooling potential of a loop. Best to worst waterblocks on a CPU can differ by about 10C and these are often aluminium and in addition not made by the well known to be great water block manufacturers. There is every chance they just plain underperform.
I agree, the block can make a world of difference and in this case I have first hand experience of an AIO outperforming custom water. I don't mean this to sound as if I think the AIO had overall higher thermal ability than the custom loop though. My W3520 runs cooler under real world conditions with it's current H80i than it ever did under my custom loop. I believe this is due to sheer luck of the draw, getting a block that just happened to be concave/convex/flat in just the right spots to match up with the cpu. Now if I were to push it to it's max overclock and leave it there indefinitely under full load I'm sure the small radiator would be overwhelmed long before the triple 120/reservoir custom loop would be. However, due to the excellent block wicking heat away from the cpu as well as it does, it also runs about ~200MHz higher than it ever did before under normal use conditions.

So a custom loop can achieve this high performance well above what air can, but the AIO's almost always seem to show they aren't really much better than high end air, they are often more expensive and less performant. I don't think Tom's is alone in this testing or that 0.15Ghz makes much difference in the performance of AIOs, I think the cheap components are really hurting the performance of the all in one loops.
I've yet to see an SB-E above 4.6 on air, not to say they don't exist, but I've not seen reports of it. I do know of plenty on higher quality AIO's though. To be honest I like the AIO's even without any performance benefit. I have an H60 sitting in a box currently that I had previously used and I wouldn't hesitate to throw that on anything I wasn't planning to clock to the moon. It's certainly not a fantastic performer, but I think the installation looks far more tidy than a massive air cooler. It allows easy access to areas around cpu socket, you don't have to worry about clearance issues, you don't have to be concerned with the substantial weight of an air cooler hanging from your mobo, and it dumps all of it's heat outside the case.
 

TheThirdMan

Member
Jul 5, 2011
113
11
81
You specified dual xeon and you call that not a server?
in most likelyhood dual Xeon's also use Reg. ECC Ram... which are different then consumer non registered ECC ram... also heat up a lot more then non registered ECC ram.
2 Logical CPUS + ECC Ram => does not equal server?

I specifically told you the differences between a dual xeon board, as well shown u examples in which i undertook to cool and silence a dual xeon board and you think im having a power trip?

First off you know dual xeon boards are EATX case to begin with?

Your getting dual xeon confused with a single xeon workstation system which in all terms is exactly the same as a desktop.

Can i point you to the release of the EVGA's SR-2.. when people thought consumer means was sufficient to cool it down yet found out the hardway its fail?

I thought you were having a power trip when you asked me "Who do you think you're talking to?"

2 xeon CPUs does not equal server. A server is (to quote google) "A computer or computer program that manages access to a centralized resource or service in a network." Having a xeon cpu doesn't make a computer a server. You can have an Intel Atom server or and AMD Athlon XP server. The hardware makes no difference to it's intended use. The conversation was about cooling a desktop style computer, some of which have dual CPUs (like dell precisions, HP z600/z800 etc)

I get what you're saying about the cooling and of course you're right, but that's about rackmount enclosures, not desktops. It's simply not true to say you can't put two CPUs in a desktop case and cool it as you would a single CPU. And though ECC ram may run hotter than standard ram at the same speeds, high end cooling in a single CPU desktop is never going to be twinned with low speed ram is it? The ram and CPU in an overclocked computer will surely run hotter than stock speed ram and CPU in a dual cpu desktop.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,894
3,247
126
I thought you were having a power trip when you asked me "Who do you think you're talking to?"

nonono...

u said it would be near suicidal..

lol... i was implying i been there... :D
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,894
3,247
126
i grew out of overclocked desktops about 3 yrs ago...
from there i went to servers... real big servers with more then 1 logical cpu.

i also went into mini's as they are becoming more versatile to use then a full bloated atx desktop.

in my transformation, i had to relearn a lot of cooling methodology, because of the space restraints.

i also got a lot more war scars from handling itx cases....
 

TerryMathews

Lifer
Oct 9, 1999
11,464
2
0
i grew out of overclocked desktops about 3 yrs ago...
from there i went to servers... real big servers with more then 1 logical cpu.

i also went into mini's as they are becoming more versatile to use then a full bloated atx desktop.

in my transformation, i had to relearn a lot of cooling methodology, because of the space restraints.

i also got a lot more war scars from handling itx cases....

Ironically, workstations with server-class chips is what got me back into overclocking.

Intel Xeon 1600LV @ 3200
Opteron 270 @ 2500
Intel Xeon L5639 @ 2930
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,894
3,247
126
Ironically, workstations with server-class chips is what got me back into overclocking.

that isnt ironic...

because the reason why i got into servers was 3 yrs ago intel decided to bring all the overclockers into enterprise territory.

Hence u got flooded with all the new people who only had overclocking on there mind.
 

Agent11

Diamond Member
Jan 22, 2006
3,535
1
0
I ran a little ole Kuhler 620 till recently with push/pull fans, on a 955BE@3.8 and a 1090T@4ghz, the latter I swapped back to air via a Noctua DH-14 with a pair of PWM 120mm Noctua fans. I saw within a degree or so of the same temps, and it's quieter, and cost more then twice as much. It's working just as well now on an FX-8350 at stock speeds(cause it's plenty damn fast). I liked the AIO cooler, it was tidy, but I was afraid of it crapping out and/or leaking when I was away or asleep, the faith in the quality just wasn't there, nor was it in any of the other AIO coolers. I was real near $150 retail but the DH-14 with PWM fans is quiet, and even if both fans die I have enough case airflow it isn't going to melt the chip at mild loads if I leave it running overnight or something. That was what brought me back to air anyway. I'd try a custom loop at some point, but not on a computer I need as near to 100% reliability from as is reasonably possible from a non-server.

Cpu's have built in thermal shutdown these days.
 

Nec_V20

Senior member
May 7, 2013
404
0
0
Aigomorla,

There all built like that... even Dells.
Theres a big reason why Dell, HP, and all the other server vendors builds them like that.
Have you ever been a NetAdmin in charge of productive servers? I have. From what you have written I assume you have not.

The very first thing you will ever notice when you go into a server environment is that the server room is bloody cold. It is also a dust free environment.

This is the reason why rack mounted servers can do their job.

Put a rackmount case in your living room with a dual Xeon running in it and you can watch it die if you put it under the load it would be expected to handle in an enterprise environment.

I would also agree with the poster who wrote that a server is not the hardware per se but rather how the hardware is used.

A NAS running on an Intel Atom for instance is a server.