Does anyone use air cooling on modern CPUs?

[BugFreak]

Just curious on what people use for air cooling on modern CPUs. Can they even be properly cooled on air without having a turbine engine of a fan? It seems I mostly read about AIO and custom loops these days.

 

[EarthDog]

Any mkd-range to high-end air coolers should handle an i9-10900k at least at stock speeds. Who knows when AVX kicks in.

Sadly, it isnt so much the capacity of the coolers but the tiny die trying to get the beat out that is the issue.

 

[freeagent]

Honestly with a CPU like a 10900K I wouldn't even consider a midrange cooler. Le Grand Macho RT, True Spirit 140 Power, NH-D15 and similar.. probably in that order

 

[BugFreak]

Throw the Dark Rock Pro 4 in there and that seems to be about the only options for cooling these modern monsters. It's been interesting reading about air cooling recently since I've been out of that realm for so long. Also seems like stock is a big problem for them as well.

 

[wagex]

the noctua d15 aughta be plenty for any stock cpu... i pump well over 200 watts into mine with the 1200rpm fans or whatever comes with it.

I think EarthDog is more on point it's not the coolers its the tiny surface area of the shrunk dies.

 

[freeagent]

D15 would be like a stock cooler for big power CPU's. If your CPU is dishing out 250w during PL2 and the D15 is rated to 260w, to me that's a bit close.. Guys used to overclock their CPU's to 5GHz now the CPU overclocks itself to that level, just gotta keep it cool to maintain the OC, just like always I reckon..

 

[Blaylock]

I think you might be missing the point Earthdog and Wagex are making though. A D15 would be nothing like a stock cooler. The issue isn't the cooler's Max TDP capabilities, it's because of the higher heat density of the increasingly smaller CPU dies. This creates a physical bottleneck to removing heat rapidly. The best liquid coolers can help with this, but it's really more of a problem with the physics.

Going forward the industry will be tasked with creating more efficient cooling methods. One such method could be microfluid cooling, another could be mainstream TEC coolers.

 

[BugFreak]

Interesting point. With most modern coolers running heat pipes that can't directly sit over the little core I can see what you mean. In that case I would think the older style heatsinks with the fins above the core and the fan blowing down on them would help with that problem. Granted it might become some mix of pipes and direct fins but I wonder if we might see a shift in design coming to help offset that problem.

 

[EarthDog]

The heat pipes sit directly over the core. That hasn't/doesn't change

The problem is the physical size of the silicon in getting the heat out. If all other things remain the same, the more surface area there is the more heat can be removed. To complicate things, the density of the silicon (node shrinks allow for more transistors in the same space also plays a role.

 

[wagex]

Exactly, even the big core I9's with 10 cores and 20 threads turbo to 5.3ghz and are only 125W TDP. They are staying within that TDP running turbo, they aren't hopping up to 350w like my old 7th gen turd to get 5ghz.

Yall gotta remember they are getting more efficient every year.. like my 7th gen hexa is 125w with a turbo of only 4ghz. so they are able to run a 10 core now at 125w with nearly twice the cores and over 25% faster frequency and like 30% faster ipc or something like that.

These tiny dies have way less surface area, less surface area means less heat transfer to the ihs, less heat transfer to the ihs means less heat transfer to the cooler.

linus does a neat video on it, a measly hyper 212 evo was enough to let the i9-10900k turbo up to 4.9GHZ the bigger more expensive coolers offer almost no improvement, all coolers hit 100c.

it wasnt until they unlocked the power limits and whatnot that the d15 and AIO take a small lead. Also you guys need to remember that AIO's are rarely much if at all any better at cooling than a high end air cooler. Most people use the AIO coolers because they look cool and aren't as hard to work around.

View attachment 212924




EDIT: I started typing this up before the previous two replies lol.

 

[freeagent]

I think you might be missing the point Earthdog and Wagex are making though. A D15 would be nothing like a stock cooler. The issue isn't the cooler's Max TDP capabilities, it's because of the higher heat density of the increasingly smaller CPU dies. This creates a physical bottleneck to removing heat rapidly. The best liquid coolers can help with this, but it's really more of a problem with the physics.

Going forward the industry will be tasked with creating more efficient cooling methods. One such method could be microfluid cooling, another could be mainstream TEC coolers.

No, the point that I was making was the TDP rating of that cooler is more in line with the TDP ratings of Intel's upper end CPU's, meaning is a good match and should be considered as a "stock" type cooler because it can take the load. I'm not arguing about thermal density, I have a 7nm chip, its not the easiest to cool when its wound right up. I don't do stock sorry

Before we used to overclock to the numbers Intel uses as boost clocks. And to get there you had to keep it cool. Now they just give you those numbers and you still have to keep it cool if you want those number like the old days. Hard to do that with a 30 buck cooler I would imagine.

Anyways.. didn't mean to rock the boat fellas, Ill see myself out.

 

[EarthDog]

Anyways.. didn't mean to rock the boat fellas, Ill see myself out.

No need to go anywhere... this is a forum and a discussion!

because modern cpu's don't have any higher TDP than cpu's a decade ago.

On paper, modern CPUs don't have any higher TDP. Because TDP has gone to mean something different. TDP, on Intel, is the base clocks. Intel's boost clocks, (PL2) are a lot higher than the TDP. This chart is with a 95W CPU, for example.

View attachment 212925

In essence, both things are affecting it. And when you overclock that all goes out the window.

 

[BugFreak]

The heat pipes sit directly over the core. That hasn't/doesn't change

The problem is the physical size of the silicon in getting the heat out. If all other things remain the same, the more surface area there is the more heat can be removed. To complicate things, the density of the silicon (node shrinks allow for more transistors in the same space also plays a role.

I think you missed what I was saying or I failed to say it correctly. With the small dies less of those heat pipes are actually near the core so not directly cooling it. I would think a nice big bank of fins with a fan blowing directly on them would be better compared to the side draft most coolers have these days which rely on the heat pipes to lift the heat up to be cooled.

 

[JrClocker]

I'm running the Noctua NH-U14S TR4-SP3 air cooler on my Threadripper 3790x (32 core, 64 thread).

This CPU can easily hit 280 W or higher for power dissipation.

This Noctua air cooler was designed to cool the multi-die Threadripper design...and it does it very well. With a push/pull fan setup, at 280 W, the CPU gets to 70 C (with 28 C ambient). As they are Noctua 140 mm fans, you can barely hear them...the motherboard VRM fans are louder.

I recently purchased (and received) my Ice Giant Thermosiphon air cooler...will be installing this weekend...supposed to cool the Threadripper better than an EKWB water cooling setup with 360 mm worth of radiator...(according to Linus)

 

[Blaylock]

I think you missed what I was saying or I failed to say it correctly. With the small dies less of those heat pipes are actually near the core so not directly cooling it. I would think a nice big bank of fins with a fan blowing directly on them would be better compared to the side draft most coolers have these days which rely on the heat pipes to lift the heat up to be cooled.

MMMmmmmmm, I don't think so. The introduction of heat pipes revolutionized air cooling.

 

[EarthDog]

I think you missed what I was saying or I failed to say it correctly. With the small dies less of those heat pipes are actually near the core so not directly cooling it. I would think a nice big bank of fins with a fan blowing directly on them would be better compared to the side draft most coolers have these days which rely on the heat pipes to lift the heat up to be cooled.

The same amount of heat is coming out of the CPU regardless. So, IMO, it won't really matter. That said, I'm not a thermodynamist(lol)... so no clue.

I'd imagine, since we've had several years of this behavior already, that if there was a significantly better idea, we'd see it... but who knows.

 

[JrClocker]

I think you missed what I was saying or I failed to say it correctly. With the small dies less of those heat pipes are actually near the core so not directly cooling it. I would think a nice big bank of fins with a fan blowing directly on them would be better compared to the side draft most coolers have these days which rely on the heat pipes to lift the heat up to be cooled.

It's about maximizing the surface area for the heat to be dissipated on. The cooler needs to be so tall (with the corresponding heat pipes to get the heat there) to create enough surface area to dissipate the heat.

Here's a pretty good example explaining the physics: https://www.heatsinkcalculator.com/blog/sizing-heat-sinks-with-a-few-simple-equations/

The reason for the heat pipe is that is has a higher thermal conductivity than straight diamond, copper, or aluminum:

- Diamond has a thermal conductivity of 2,000 W/m-k
- Copper has a thermal conductivity of 390 W/m-K
- Aluminum has a thermal conductivity of 205 W/m-k (higher is better)
- Heat pipes have thermal conductivity ranging from 4,000 to 100,000 W/m-k

 

[BugFreak]

Oh I'm not saying they don't work good but it seems a mix of the two would allow the fins to spread the heat away from the small core then the pipes could spread from there away to be cooled.

Also keep in mind this is me totally bsing because I know as much about how these things work as I do launching rockets (which is none).

 

[EarthDog]

I launched model rockets... though, that isn't relevant.

 

[wagex]

I launched model rockets... though, that isn't relevant.

i had to re-read that a time or 3, i think he means that he knows little about launching rockets and little about coolers.