Why do AIOs never give any information about their cooling power?

darius123

n00b
Joined
Apr 17, 2021
Messages
6
Hi,

i'm new into watercooling and have a (maybe) stupid question: Why do AIO manufacturers don't give any specific information about the cooling power of their products? I am trying to find an AIO which is capable of cooling a LED Chip with 200 watts of heat power (TDP), so that temperature stays below ~60 °C. The only option i see right now to find out if a specific AIO is suitable for this case is reading other peoples experiences or trying it out, which seems stupid... Can someone explain?

Thank you,
Darius
 
Because AIO's are generally just a gimmick. I got one, but the reality is that they are mostly just a marketing/gimmick exercise. The only real benefit is just the cleaner look case compared to having the massive heatsink on the CPU. If you've got the room, a decent heatsink will perform just as well, if not better, for the money.
 
Because AIO's are generally just a gimmick. I got one, but the reality is that they are mostly just a marketing/gimmick exercise. The only real benefit is just the cleaner look case compared to having the massive heatsink on the CPU. If you've got the room, a decent heatsink will perform just as well, if not better, for the money.
I used to think AIOs were better until I used some Noctua coolers. The biiiiiiiig ones. They work just as well.

To answer your question, I suppose it would depend on the fans, on their speed, ability to push air... If you are running it in push vs pull.

I agree that there should be some sort of standard though. All fans in push, on automatic fan control, after x amount of time.
 
Most likely because without a good testing standard there would be no meaningful way to compare numbers and most of the audience would probably not understand the numbers. It depends on a lot of factors like how large is the warm surface (finstack coverage), what kind of tim do you use, what is the ambient, was it noise normalized etc. E.g. if one tests with the supplied paste and the other tests with the best liquid metal then there will be a significant difference in number given.

Another example is do you measure max capacity at settings that user never will run with or with the lower noise profiles the AIO is delivered with? In my experience the most agressive fan profile on corsair h150i pro xt is about 5-7 degrees worse than what the cooler is capable of on a 5800x, but it has a much more palatable noise profile. At 40dbA noise normalized fan profiles the 360 and larger AIOs will beat out the best air coolers every time, but those are generally higher fan speeds than what most users do use so the real world experience is different.

Anandtech has a good setup for measuring coolers and probably the most accurate measurements of capacity for AIOs and air coolers.

The temp you get from your chip depends on the heat transfer it gets with the chip etc. Pretty much all good air towers and 360 AIOs are capable of keeping it below 60 if they get optimal heat transfer, but that is an unknown until you try.
 
Yes, sounds reasonable that it's hard to come up with a testing standard. But i think that some numbers for orientation - even if not very accurate - would be better than nothing. For example enermax is stating that their LIQTECH TR4 II offers an cooling capacity of up to 500W+ TDP. That at least gives me the certainty that its suitable for my 200 watt LED. Sadly, this model seems to have problems with corrosion.

@Mchart I do not have the place for a big heat sink, thats why i need a water cooling solution :/

Anyways, thanks a lot for your answers guys!
 
Because AIO's are generally just a gimmick. I got one, but the reality is that they are mostly just a marketing/gimmick exercise. The only real benefit is just the cleaner look case compared to having the massive heatsink on the CPU. If you've got the room, a decent heatsink will perform just as well, if not better, for the money.

I'd argue there is a pretty big benefit in dumping the heat outside the case instead of inside it.
 
Hi,

i'm new into watercooling and have a (maybe) stupid question: Why do AIO manufacturers don't give any specific information about the cooling power of their products? I am trying to find an AIO which is capable of cooling a LED Chip with 200 watts of heat power (TDP), so that temperature stays below ~60 °C. The only option i see right now to find out if a specific AIO is suitable for this case is reading other peoples experiences or trying it out, which seems stupid... Can someone explain?

Thank you,
Darius

Tell us more about this LED chip! I'm kind of curious what you are trying to do. That's a lot of light!
 
I'd argue there is a pretty big benefit in dumping the heat outside the case instead of inside it.
That isn't always the case. Some people, like me, have a radiator pulling in outside air.

Though my gpu is looped too, so it isn't swallowing up the ejected heat. Also you would have case fans blowing outward to keep the heat from bottling up inside.
 
I'd argue there is a pretty big benefit in dumping the heat outside the case instead of inside it.
It would depend on the case. Most decent cases have both the top and bottom areas that you could mount a radiator. However, you could just as easily just mount normal fans in these areas, and have ample airflow through the front over the center of the board where the heatsink is and ejected on the top. If you had a radiator you'd either have to mount it in the front which means fresh air for the CPU AIO, but warm stale air going to the rest of the case. Or you'd have to mount it on top, but now you don't have fresh air passing through it. You could also mount in the normal back area, but this is typically only 240mm at most which isn't enough radiator for any higher-end CPU.

Ultimately, from every realistic test i've seen, the AIO's don't perform any better for the money. A lot of people arrive at false conclusions of the performance of AIO's as well, because they'll just quick run a test and not actually heat-soak in the coolant. So for a short time the AIO is better, but once it's heatsoaked the difference is negligible. Water cooling really only makes sense if you are doing it for the looks, or you are doing a real setup where the radiator is exterior to anything happening in the case. I like doing AIO just because it removes that massive heatsink in the middle of the board, but i'm not fooling myself into thinking my 360mm AIO is any better than would I could have achieved with the same amount of fans and a heatsink.
 
Last edited:
Tell us more about this LED chip! I'm kind of curious what you are trying to do. That's a lot of light!





op, some give an average wattage rating. i think that people used to say 100w per 120mm, so a 240/280 should be plenty. could be wrong, someone will correct me...
 
If you had a radiator you'd either have to mount it in the front which means fresh air for the CPU AIO, but warm stale air going to the rest of the case. Or you'd have to mount it on top, but now you don't have fresh air passing through it.

While the norm, you really don't have to adhere to this design. For instance, I have a top mounted radiator sucking in outside air.
 
Tell us more about this LED chip! I'm kind of curious what you are trying to do. That's a lot of light!
Hey! It's a 200w COB LED, you can find these on ebay or amazon if you look for "COB LED CRI 95+ 100/200/300w"

Im trying to build an "artificial sunlight", similiar to this one right here (but more compact and transportable for filming purposes):
 
Seems like an interesting project :)

Make sure the radiator side that doesn't have fans with protective grills or something so it is hard for stuff to damage the water channels. Optionally doing push pull as the fans would provide protection.
 
Because AIO's are generally just a gimmick. I got one, but the reality is that they are mostly just a marketing/gimmick exercise. The only real benefit is just the cleaner look case compared to having the massive heatsink on the CPU. If you've got the room, a decent heatsink will perform just as well, if not better, for the money.
May be true for poorly built AIO's. But generally a well built 240/280mm AIO is better than any air cooler.

MSRP Arctic Freezer 280mm is ~$115; Noctua NH-D15 is $100.

AF 280mm is a good 5-6C cooler @ 200W, noise normalized.

 
Because AIO's are generally just a gimmick. I got one, but the reality is that they are mostly just a marketing/gimmick exercise. The only real benefit is just the cleaner look case compared to having the massive heatsink on the CPU. If you've got the room, a decent heatsink will perform just as well, if not better, for the money.
It's not a gimmick, it's a convenience. Also, it has been proven well enough that water cooling is better when factoring in the size of the cooling apparatus. Sure, an air cooler can be effective, but it takes up a lot of space and given that PCs are for the most part standardized this becomes an issue. Water cooling makes up for this by being a bit more diverse in terms of size and placement. Custom loops are even better in terms of maximizing cooling potential. There is a reason beyond 'it's a gimmick' that water cooling is relevant.
 
The one main area that AIOs excel in over normal heat sinks is time to thermal saturation. As you have more thermal mass to heat it takes longer to heat them up. This is good when you have transient heat load like with a cpu, not so much with constant load.
 
Last edited:
I think a big part of the general lack of hard dissipation specs on AIOs (and coolers in general more and more) has to do with the range of variables involved in the heat transfer with modern CPUs (and GPUs)

With process densities and die sizes all over the place and chiplets thrown in the mix, a given AIO can perform very differently with different chips. In my own experience, the exact same AIO installed on a Ryzen 1600 @140W, Ryzen 3600 @90W, and a Vega 64 @300W resulted in about the same temps on each because the chip with the lowest dissipation was the densest and the chip with the highest was the least dense (and was also direct-die cooled)
 
This is from Alphacool and it’s a rough guide only - info is from https://forum.alphacool.com/threads...much-radiator-you-need-to-cool-280w-tpd.1525/

But there is a simple rule you can follow
80W TDP per 120mm radiator space with 600rpm fans
100W TDP per 120mm radiator space with 800rpm fans
120W TDP per 120mm radiator space with 1000rpm fans
150W TDP per 120mm radiator space with 1500rpm fans
180W TDP per 120mm radiator space with 2000rpm fans

i hope it make sense to you, but as others have stated there are many variables

there has been lengthy discussions on this forum with regards to how much cooling you need thesmokingman has a very large build where he has to accommodate for for very long periods of full loads, hence need more cooling power. Where the above table isn’t useful, you may need to double or triple the numbers to get good cooling

I hope this helps
 
May be true for poorly built AIO's. But generally a well built 240/280mm AIO is better than any air cooler.

MSRP Arctic Freezer 280mm is ~$115; Noctua NH-D15 is $100.

AF 280mm is a good 5-6C cooler @ 200W, noise normalized.



Thanks for the video, good information! But according to this one none of the tested AIOs were capable of cooling 200 watts down to 60 °C, even with full speed. Or did i get something wrong?
 
Thanks for the video, good information! But according to this one none of the tested AIOs were capable of cooling 200 watts down to 60 °C, even with full speed. Or did i get something wrong?
The issue is the heat transfer from the chip to the cold plate of the cooler. E.g. the chips are very small and has to transfer the heat through solder and the heat spreader. Each layer partially acts as insulation even though they do transfer a good portion of the heat. The heat is located in a tiny area while the finstack of the cooler is typically 80-100% of the heat spreader. If you were running a chip that was the size of the heatspreader and evenly distributing the heat across it then the results would be quite different. The CPU has 2 CCDs and each CCD is aprox. 72mm^2 while the heatspreader itself is aprox. 1380mm^2 so the heat is very concentrated and it is impossible to get optimal heat transfer.
 
The issue is the heat transfer from the chip to the cold plate of the cooler. E.g. the chips are very small and has to transfer the heat through solder and the heat spreader. Each layer partially acts as insulation even though they do transfer a good portion of the heat. The heat is located in a tiny area while the finstack of the cooler is typically 80-100% of the heat spreader. If you were running a chip that was the size of the heatspreader and evenly distributing the heat across it then the results would be quite different. The CPU has 2 CCDs and each CCD is aprox. 72mm^2 while the heatspreader itself is aprox. 1380mm^2 so the heat is very concentrated and it is impossible to get optimal heat transfer.

Yep, you can't compare 200 watts from one CPU to 200 watts from a different one. I have a Ryzen 1700 that puts out around 110 watts in P95 small FFT and a 5960X that puts out around 220, both on their 24/7 overclocks. The Ryzen has a 280mm AIO while the big Haswell is on a Phanteks TC14PE air cooler. Load temperatures are about the same, but it's not because there's anything wrong with the AIO. It's because a 14nm chip is harder to cool than a 22nm chip, all things being equal.

Sometimes it's not even the size of the die, but how the heat is distributed. When I yanked the 5820K from my main rig and swapped in a 5960X, I was expecting it to run hotter, but it was basically the same. Even though there's more heat from 8 cores, it's spread out across more active silicon on the 5960X
 
The issue is the heat transfer from the chip to the cold plate of the cooler. E.g. the chips are very small and has to transfer the heat through solder and the heat spreader. Each layer partially acts as insulation even though they do transfer a good portion of the heat. The heat is located in a tiny area while the finstack of the cooler is typically 80-100% of the heat spreader. If you were running a chip that was the size of the heatspreader and evenly distributing the heat across it then the results would be quite different. The CPU has 2 CCDs and each CCD is aprox. 72mm^2 while the heatspreader itself is aprox. 1380mm^2 so the heat is very concentrated and it is impossible to get optimal heat transfer.

Alright, makes sense. As the size of the CPU (or LED) and its heat distribution is something i can't change, what i need is a good cooling plate that can transfer the heat most effectively.

The size of my 200 watt LED is 784 mm² (the surface that has contact to the cooling plate), the LED module itself is around 500 mm². Do you think that this LED is easier to cool than for example the AMD R9 3950X, which was used in the video above? (i can't find any information about it's size, one site says that it's DIE size is 74 mm²). And do you therefore think that AIO's might be capable of cooling my 200 watt LED to under 60 °C?

Thanks already for all the help, i'm completly new to all of this so sry if i don't get it right away.
 
Alright, makes sense. As the size of the CPU (or LED) and its heat distribution is something i can't change, what i need is a good cooling plate that can transfer the heat most effectively.

The size of my 200 watt LED is 784 mm² (the surface that has contact to the cooling plate), the LED module itself is around 500 mm². Do you think that this LED is easier to cool than for example the AMD R9 3950X, which was used in the video above? (i can't find any information about it's size, one site says that it's DIE size is 74 mm²). And do you therefore think that AIO's might be capable of cooling my 200 watt LED to under 60 °C?

Thanks already for all the help, i'm completly new to all of this so sry if i don't get it right away.
I am not willing to guarantee it, but I do think it is quite likely that a 360mm AIO is sufficient. You may have to turn up the fans a bit and set pump to full speed compared to "auto" settings that a lot of AIOs come with. Buy some additional cooling paste as well, in case you have to remount the unit.

The alphacool nexxos should be similar to what was used in the video you posted with a 500w LED (custom loop) so you can use that as a reference. Chosing the 200w graphs should give you an indication of the delta in temps under very good heat transfer conditions. Do keep in mind that the results there are at max fans so some can be quite noisy. https://www.anandtech.com/show/1642...iquid-freezer-ii-240-420-aio-coolers-review/3

There are 2 dies each of around 72mm^2 (CCDs) in the 3950x.
 
500mm^2 is a big chip, up there with large GPUs, so you've got that going for ya. Under 60*C sounds plausible but still depends on ambient temperature, airflow, and how good the heat transfer between LED->substrate-AIO coldplate is.

Seems it's likely enough to work that it's worth trying
 
What also doesn't help is that Corsair released a hunk of garbage AIO that required you to install Windows based software to control the pump speed via USB. Corsair, STOP. Keep it simple, we don't need SOFTWARE to control our pump speeds. Whatever happened to a simple PWM connector?

BTW this was for his Linux server, so where's Corsair's Linux support? Yeah ok.
 
Last edited:
Back
Top