Are we witnessing the slow demise of the air cooler?

[xtreemchaos]

But your tub of veg oil is also cooled by air.

i was going to say not if the tub was cased in ice but the ice would be cooled by air so i messed up there where's the delivery guy with my meds.

 

[Space Lynx]

I've used both cooling types extensively they seem to gain about the same within margin of error going with some 3000 RPM jet engines on them (which is pointless my 360 aio running fans at less than 1k rpm already keeps my 5950X below 70C under full load as it is) but 63-65c looks good vs 67-69c I guess. Maybe that will change once I upgrade to a next gen cpu....

To me the whole air cooler vs aio argument has always been dumb there are valid reasons to use either and they don't always come down to raw performance.

Aren't you the one desperately looking for a budget case that fits a 420mm aio?

yeah I really want the Arctic 420mm AIO. i'm big fan of Noctua too though. i just think the gains are less than people realize.

 

[Count von Schwalbe]

Why not just buy a cheaper CPU with lower operating temperature or the previous gen if you would plan on having 7900X undervolted for example. Seems like a logical choice. 5900X runs at 67 °C under full load.

7900X @ 80W with a voltage offset loses almost nothing for power, trouncing the 5900X.

 

[Sithaer]

We're on an enthusiast site so there is considerable enthusiast thinking about things and that's fine but most PCs are used in businesses where MS Office is one of the biggest needs. They don't need high performance CPUs and don't overclock at all and may only get an i3 or equivalent. They don't need anything but the stock cooler that comes with the basic CPU. Before I retired I worked for a company that employed around 7,000 people. Almost all of them needed a PC or laptop and none of them used anything but stock Dells. I can't imagine what my boss would have said to me if I had suggested that we need to toss in the trash those perfectly good stock coolers that had already been paid for and put in water coolers. I'm pretty sure he would have laughed at me and he would have been right to do so.

Yeah on the lower end-mid range air is all good and doubt that will change anytime soon.
I'm using a 12100F and the stock cooler was surprisingly usable, better than the prev gen stock intel coolers for sure.
I threw a ~30$ tower cooler on it and thats more than enough to keep it under 60 celsius while gaming with zero noise.

On the high/enthusiast end yea I can see air becoming an issue if the current trend continues but luckily that doesn't affect most of the users.

Personally I was only ever interested in AIOs is cause I like the aesthetics a lot more than a tower cooler.
Btw I can run a lot more than MS Office on this i3.

 

[caroline!]

If people would just set stronger fan curves and have good fans that don't make loads of noise at higher rpm like Noctua or Aer-P, we would be witnessing the demise of all AIO's.

Also, my 7700x will be getting a -30 undervolt, only hits performance 0.5% across the board, but comes with 30 celsius temperature drop.

Doubtul decent fans like ML Pro, Noctua a12x25, Phanteks T30 help aio as well.

Both cooling types have a purpose and aren't going anywhere anytime soon.

You want decent fans? Delta, Sunon, Mechatronics...

Barely known in the prosumer/gamer RGB+tempered glass world but FAR better than Noctua or new Corsair fans. Strap a 120x38 to a large air cooler and that CPU will be ice cold.

 

[Andrea87]

I believe air coolers are nowhere near their end - even in "enthusiast" markets.

There are relevant coolers in the 30-40$ range with performance fair enough to run most cpus. Also, let's not forget the simplicity and reliability of air. No water flowing, no additives in the fluid to avoid bacteria and whatever other crap forming; no water pumps.

What I think are getting closer to their end are direct-heatpipe coolers. With modern cpus having smaller and smaller core dies, and therefore denser heat output, without any buffer besides the cpu's IHS there's way too little space to efficiently spread the heat to all the heatpipes. Most modern designs have a significant cold plate to counter this issue.

Personally I'm running a Thermalright Peerless Assassin 120 SE (around 40€ on amazon EU) on a ryzen 5800x, with temps under 67°C in most stress tests, running it at 110W / 4,6GHz all core. Room at 20°C. I don't feel any need to invest in a water cooling system, as temps are acceptable. With my current setup the only "maintenance" i might have to do is swap fans or replace the thermal paste when necessary.

 

[80251]

You want decent fans? Delta, Sunon, Mechatronics...

Barely known in the prosumer/gamer RGB+tempered glass world but FAR better than Noctua or new Corsair fans. Strap a 120x38 to a large air cooler and that CPU will be ice cold.

Sunon is hard to get here in the US. You left out SanAce and Nidec. I have a really nice 127x50mm Nidec fan (alleged to be capable of 200CFM in free air).

 

[AusWolf]

Sunon is hard to get here in the US. You left out SanAce and Nidec. I have a really nice 127x50mm Nidec fan (alleged to be capable of 200CFM in free air).

Is it noisy, though?

 

[freeagent]

I used to beat the 120x38 drum and no one cared.. those fans are good if you need high pressure, but most don’t.

 

[oxrufiioxo]

I used to beat the 120x38 drum and no one cared.. those fans are good if you need high pressure, but most don’t.

They are also loud, use a ridiculous amount of power for a fan, and are hard to mount in some cases.

The T30-120 are already sorta ridiculous at 3000 rpm as it is.

 

[freeagent]

They are also loud, use a ridiculous amount of power for a fan, and are hard to mount in some cases.

The T30-120 are already sorta ridiculous at 3000 rpm as it is.

Noctua iPPC is just as crazy at 3000rpm. I can hear them from the basement on the main floor when they are running at top speed. Not a huge house, and has an open concept so good bye walls and privacy lol.

 

[80-watt Hamster]

I used to beat the 120x38 drum and no one cared.. those fans are good if you need high pressure, but most don’t.

38mm fans would be great, but it's hard to find them simultaneously a) in PWM, b) with a preinstalled connector, and c) available through retail channels. Dell was using 92x38 fans as exhausts in their mini towers for awhile (maybe they still do), and they worked really well.

 

[AnotherReader]

Air cooling will always be more popular than water cooling due to ease of maintenance and reliability. Even the performance differential isn't as great as it's made out to be. As for the recent space heaters masquerading as CPUs, they can be cooled more easily by just setting a lower power limit. Computerbase's 7950X review shows this:

CPU	max. temperature (Tctl/Tdie)
Ryzen 9 7950X Default (230 Watts)	95.3°C
Ryzen 9 7950X 142 watts	73.9°C
Ryzen 9 7950X 88 watts	65.2°C
Ryzen 9 5950X Default (142W)	68.0°C

The 7950X with a 5950X like 142W power limit is over 20 C cooler and only 6% slower in Blender than its stock settings.

 

[jayjr1105]

You'll have to pry my fuma 2 from my cold dead hands.

 

[Eskimonster]

Perfect storm does this, competitors fighting each other to produce the best of the best hurts/bennefits us all.
When i see a 7950X performing the same at Ecomode (65W), as a 12900k full throttle, i see redflags all over.

edited typos

 

[Mr McC]

Focusing on the enthusiast segment exclusively, it is interesting that a number of members have pointed to undervolting stock settings in order to control higher temperatures, whilst others have indicated that more powerful fans will provide a solution. This does suggest that there is at least some admission that, within the enthusiast segment, current cooling solutions are not entirely up to the job at stock settings. Of course, we could also question the validity of stock settings when small voltage tweaks can improve temperatures so drastically with a relatively minor impact on overall performance.

Moreover, if such solutions work today, will they continue to work as CPU dies shrink even further in future generations? Perhaps case design will have to change to the point where the entire case acts as giant heatsink.

I am not trying to be overly dramatic or alarmist, I am simply playing devil’s advocate to keep an interesting conversation going.

 

[freeagent]

Dell was using 92x38 fans as exhausts in their mini towers for awhile (maybe they still do), and they worked really well.

I have a 92x38 DC Koala III.. it’s a 3900RPM beast that will draw blood and remove flesh if you attack from the right angle, and it will pop you’re fan header if you try to hot plug it.. it’s an awesome fan, moves waaay more air than a TY-143. And it’s pretty loud too lol, min speed is 1800 revs at 5v, at that speed it’s not horrible, but you can hear it for sure.

 

[caroline!]

Sunon is hard to get here in the US. You left out SanAce and Nidec. I have a really nice 127x50mm Nidec fan (alleged to be capable of 200CFM in free air).

Nidec are great, the ones made in Japan at least. Haven't seen any SanAce but it has its rep too.

I used to beat the 120x38 drum and no one cared.. those fans are good if you need high pressure, but most don’t.

For thicc heatsinks and rads.

I have a 92x38 DC Koala III.. it’s a 3900RPM beast that will draw blood and remove flesh if you attack from the right angle, and it will pop you’re fan header if you try to hot plug it.. it’s an awesome fan, moves waaay more air than a TY-143. And it’s pretty loud too lol, min speed is 1800 revs at 5v, at that speed it’s not horrible, but you can hear it for sure.

puny mobo headers can't handle those, they must be wired directly to the PSU, if you want speed control add a pot. I went the easy way and wired a few 3-position switches at the back of the case so I can swap between off, 5 and 12V

Focusing on the enthusiast segment exclusively, it is interesting that a number of members have pointed to undervolting stock settings in order to control higher temperatures, whilst others have indicated that more powerful fans will provide a solution.

Perhaps case design will have to change to the point where the entire case acts as giant heatsink.

It should, in part. Hence why I like cases with little plastic and absolutely no glass, if the case gets a bit hot when the parts are under load it means it's letting some of the heat dissipate passively to the ambient, while the fans do their part.

My case gets hot yet the parts are running at acceptable temps.

Then you have reddit gamers who say nooo the case should be cold and boom put a whole glass pane on top to act as an insulator lol.

 

[AusWolf]

Perfect storm does this, competitors fighting each other to produce the best of the best hurts/bennefits us all.
When i see a 7950X performing the same at Ecomode (65W), as a 12900k full throttle, i see redflags all over.

You see red flags. I see a CPU that can either go full boost with some high-end watercooling, or work as a quiet little monster by just turning on Eco mode. No manual tinkering required.

I love my 11700 and my Asus TUF motherboard with its "Asus Optimizer" setting for the same reason.

 

[Eskimonster]

You see red flags. I see a CPU that can either go full boost with some high-end watercooling, or work as a quiet little monster by just turning on Eco mode. No manual tinkering required.

I love my 11700 and my Asus TUF motherboard with its "Asus Optimizer" setting for the same reason.

i must be bad at this, im exited that we gotten so far, i was adressing that CPU´s are getting so far they need a bigger cooler then so far.
my red flag was pointed at people dont like this and want to compare them at older CPU´s which uses less power.
I´m happy we have the option to run it at we see fit, even more happy that this 7950X can slap 12900k at Eco mode. But some dont get that, and thats my red flag.

 

[AusWolf]

i must be bad at this, im exited that we gotten so far, i was adressing that CPU´s are getting so far they need a bigger cooler then so far.
my red flag was pointed at people dont like this and want to compare them at older CPU´s which uses less power.
I´m happy we have the option to run it at we see fit, even more happy that this 7950X can slap 12900k at Eco mode. But some dont get that, and thats my red flag.

Ah OK, fair enough.

 

[Vario]

I used to beat the 120x38 drum and no one cared.. those fans are good if you need high pressure, but most don’t.

You should go for 76mm instead

9CR1212P0G03 | Counter Rotating Fan | San Ace | Product Site | SANYO DENKI

 

[cvaldes]

Moreover, if such solutions work today, will they continue to work as CPU dies shrink even further in future generations? Perhaps case design will have to change to the point where the entire case acts as giant heatsink.

There is probably an upper limit to how much power a consumer CPU can draw mostly for practical real world reasons. The first main thing I can think of is the limitation of a standard household circuit: 120 V x 15 A = 1800 W. I'll get to the second one later.

With a monitor and typical accessories (like a set of powered computer speakers + subwoofer, maybe an external drive or USB hub), the typical desktop PC -- even the enthusiast grade builds -- are probably have a soft limit about 1500 or 1600 watts peak draw.

Proper airflow appears to be more effective than case material choice. Don't forget Lian Li originally focused on all aluminum computer cases (back in the Nineties) which many believed would help dissipate heat. A flat panel doesn't scale very well which is why typical desktop PC heatsinks are designed with spikes or ridges to increase surface area. And if you were using the exterior case panels, you'd still need a way to transfer the heat from the source to those panels.

The radiator is really just an extension of this large surface area for cooling. You're just using water (or other liquid coolant) to move the heat from the source (silicon chip) to a place where the heat can be exhausted fairly easily because water has great thermal capacity.

Remember that all of these cooling solutions are compromises. Thermal performance, ease of installation, acoustics, size, ease of construction, durability, materials, maintenance requirements, cost, etc.

Could there be 400-450W PPT CPUs one day? Possibly. Could they be cooled? Sure, why not? GPU chips already put out that PPT and most of them are air cooled at this point. Can it be done quietly? Maybe. GPU cooling systems are constrained to certain dimensions. We're seeing larger and larger AIB coolers to handle increased GPU heat production. In some of the high-end graphics card products, there are hybrid AIO liquid/fan cooling systems.

There's an upper limit to what Joe Consumer will find as acceptable fan noise for a consumer electronics product. Hell, some people gripe about fan noise from their video game consoles, it's not just about enthusiast PCs.

And now the second reason: from a total cost of ownership perspective, I'm not sure how many people really want a 1500 W PC running for a long time. It's not just the component cost or how much power a residential circuit can handle: there's also the electricity charges to consider. For sure, there will always be enthusiasts who would want such a machine but I would expect them to be in a very small minority of even enthusiasts.

There aren't many CPU workloads that will max out a typical gaming enthusiast's CPU. Most games are still heavily reliant on IPC/single threaded performance. I look at the performance analyses at DSOGaming and usually peak gaming CPU performance is achieved with 6 cores/12 threads or occasionally 8 cores/16 threads.

There are always unusual edge cases and I'm sure someone here will say "I use Linux on a 64-core CPU to do blah blah blah" which is probably something that 0.001% of people on this planet do but they think is common enough to mention.

 

[80251]

Is it noisy, though?

At full tilt...affirmative. They're PWM though so they can be controlled.

 

[Zach_01]

There is probably an upper limit to how much power a consumer CPU can draw mostly for practical real world reasons. The first main thing I can think of is the limitation of a standard household circuit: 120 V x 15 A = 1800 W. I'll get to the second one later.

With a monitor and typical accessories (like a set of powered computer speakers + subwoofer, maybe an external drive or USB hub), the typical desktop PC -- even the enthusiast grade builds -- are probably have a soft limit about 1500 or 1600 watts peak draw.

Proper airflow appears to be more effective than case material choice. Don't forget Lian Li originally focused on all aluminum computer cases (back in the Nineties) which many believed would help dissipate heat. A flat panel doesn't scale very well which is why typical desktop PC heatsinks are designed with spikes or ridges to increase surface area. And if you were using the exterior case panels, you'd still need a way to transfer the heat from the source to those panels.

The radiator is really just an extension of this large surface area for cooling. You're just using water (or other liquid coolant) to move the heat from the source (silicon chip) to a place where the heat can be exhausted fairly easily because water has great thermal capacity.

Remember that all of these cooling solutions are compromises. Thermal performance, ease of installation, acoustics, size, ease of construction, durability, materials, maintenance requirements, cost, etc.

Could there be 400-450W PPT CPUs one day? Possibly. Could they be cooled? Sure, why not? GPU chips already put out that PPT and most of them are air cooled at this point. Can it be done quietly? Maybe. GPU cooling systems are constrained to certain dimensions. We're seeing larger and larger AIB coolers to handle increased GPU heat production. In some of the high-end graphics card products, there are hybrid AIO liquid/fan cooling systems.

There's an upper limit to what Joe Consumer will find as acceptable fan noise for a consumer electronics product. Hell, some people gripe about fan noise from their video game consoles, it's not just about enthusiast PCs.

And now the second reason: from a total cost of ownership perspective, I'm not sure how many people really want a 1500 W PC running for a long time. It's not just the component cost or how much power a residential circuit can handle: there's also the electricity charges to consider. For sure, there will always be enthusiasts who would want such a machine but I would expect them to be in a very small minority of even enthusiasts.

There aren't many CPU workloads that will max out a typical gaming enthusiast's CPU. Most games are still heavily reliant on IPC/single threaded performance. I look at the performance analyses at DSOGaming and usually peak gaming CPU performance is achieved with 6 cores/12 threads or occasionally 8 cores/16 threads.

There are always unusual edge cases and I'm sure someone here will say "I use Linux on a 64-core CPU to do blah blah blah" which is probably something that 0.001% of people on this planet do but they think is common enough to mention.

I agree with everything but how about this:

There is probably an upper limit to how much power a consumer CPU can draw mostly for practical real world reasons. The first main thing I can think of is the limitation of a standard household circuit: 120 V x 15 A = 1800 W. I'll get to the second one later.

With a monitor and typical accessories (like a set of powered computer speakers + subwoofer, maybe an external drive or USB hub), the typical desktop PC -- even the enthusiast grade builds -- are probably have a soft limit about 1500 or 1600 watts peak draw.

To be honest I dont understand what you mean that a household circuit has a limitation of 1800W.

I run myself (a couple of years ago) on 1 wall plug, 3 PCs, 2 monitors, 1 laptop and a bunch of other smaller stuff (router, bridges etc) with total power of around 2000W.
And the max power allowance of plug/cable/safety was 3600W (230v x 16A). Those (almost all) was running for 6 months straight 24/7. And I'm not even mentioning other house devices.
My house is a standard apartment 45y old and its max power capacity is 230v x 35A. Always has been.

Do I run this now? No. Do I want a single PC of 1500~2000W power draw? No

I sincerely dont get your point there.