News Noctua NH-D15 G2 update has quieter fans, more heatpipes, new offset design

[Admin]

Noctua showed off its highly anticipated next-generation NH-D15 CPU air cooler at Computex 2024. Key changes with the new NH-D15 G2 design include quieter fans, more heatpipes, a denser fin stack, and a new offset design. PC DIYers should be able to grab a new Noctua NH-D15 G2 later this month.

Noctua NH-D15 G2 update has quieter fans, more heatpipes, new offset design : Read more

 

[bit_user]

I can't believe how much air they're still allowing to leak out the top and bottom of the fan frames! If I got this, the first thing I'd do is tape something over the exhaust side of that part of the fan frames, at the top, thus forcing more air through the fin stack.

What to do at the bottom is a trickier question. I wouldn't want to block the entire gap, but maybe just a strip right below the bottom of the fin stack.

 

[helper800]

I cannot wait until Albert's review of this new cooler!

I can't believe how much air they're still allowing to leak out the top and bottom of the fan frames! If I got this, the first thing I'd do is tape something over the exhaust side of that part of the fan frames, at the top, thus forcing more air through the fin stack.

What to do at the bottom is a trickier question. I wouldn't want to block the entire gap, but maybe just a strip right below the bottom of the fin stack.

I have seen people do this mod to prior coolers and the differences were always margin of error. Though I can understand trying to eek out every last drop of water from the rock.

 

[Eximo]

https://imgur.com/fLQGi1I

View: https://i.imgur.com/fLQGi1I.gif

 

[bit_user]

I have seen people do this mod to prior coolers and the differences were always margin of error. Though I can understand trying to eek out every last drop of water from the rock.

I'm sure it would work better if you had actual ducts and not just walls.

 

[TheHerald]

More interested in the fans rather than the cooler. These are the first 140 LCP fans

 

[Notton]

If you use ducts, it'll work better if you use a thick single tower and do a 180mm push and 140mm pull (but only because 180mm doesn't fit on the back of a case)

 

[35below0]

I was looking at replacing my NH-D15 fans... what fortuitous happenstance!

 

[cyrusfox]

I can't believe how much air they're still allowing to leak out the top and bottom of the fan frames! If I got this, the first thing I'd do is tape something over the exhaust side of that part of the fan frames, at the top, thus forcing more air through the fin stack.

What to do at the bottom is a trickier question. I wouldn't want to block the entire gap, but maybe just a strip right below the bottom of the fin stack.

Air flow across the top of the fins is still doing useful heat transfer, blocking out that small portion will not appreciably improve the fans static pressure to make up for the reduced airflow over the external fins. As others have suggested, doing a duct design enables overcoming the impedance issues open air coolers deal with, blocky + ugly, really only make sense for the OEMs(many Dell and HP SFF use to utilize chambers for this very reason).

If anyone is doing CFDs on their design choices it would be noctua, I only wish they would up their lowprofile game, as other choices are superior than their meager offers at sub 55mm height, but their fans are world class, and their coolers hold resale value. I resold mine for the same price I bought it at 3 years later (got to love inflation...).

 

[eX_Arkangel]

on the bottom part i keep wondering why all manufacturers refuse to try and somehow attach more fins as close to the actual base/90ish degree bend of the pipes, even if they are smaller fins in every metric and spread wider, half clipped from one side etc, also why they dont enlarge the socket part above the heatpipes with an chipset like heatsink (old chipsets) around (or being part of) the mounting mechanism, with the middle fan having a bottom cavity in the frame to allow some airflow to go down and the outer fan already pushing air in that direction, i mean those are the closest parts to the actual heat source and somehow are the ones with less dissipation capacity equipped.
Another doubt/curiosity i have... do heatpipes work better if seated on a horizontally laid heat source (cpu-gpu) with the heatpipes bending vertically up (not down like in current GPU's), just like in aircooled server racks, if so why isn't there a more aggressive alignment offset to one particular side for lateral seated CPU's.
Also why they keep just aligning the heatpipes (in most DT coolers) straight parallel instead of sig zagging them for better fin area heat dissipation distribution (and maybe if done with lab testing, some airflow enhancement due to using pipes as air directional elements (something like they did with the NH-U12A etc).
Also why straight fins and no slight wave shape to them like a soft horizontal "~" shape (maybe done in both axis), forcing some air turbulence within the fins and making the air bounce more over the surface of the fin stack.

PD: just my personal opinion, maybe I'm just getting wrong how heatpipes really work, and that bare gap is needed for proper cooling. And the rest is just doubts/crazy thoughts i've had for years now regarding air coolers in particular, the only ones close to my imagination were old Zalman Coolers from 2000s, and some old Thermalright coolers.

 

[bit_user]

Air flow across the top of the fins is still doing useful heat transfer,

Not much.

blocking out that small portion will not appreciably improve the fans static pressure to make up for the reduced airflow over the external fins.

Think about it like this: if it increases airflow by like 5%, that's equivalent to another 5% higher RPM but without the associated noise. When you get into the loud range of a fan's RPMs, 5% more can make a noticeable difference. Depends on the fan, of course.

If anyone is doing CFDs on their design choices it would be noctua,

I'm not saying it's not good enough, as is, but maybe just not as good as it could possibly be.

Also, you've got to look at what Thermalright has been able to accomplish, often at significantly lower price points, to see that Noctua isn't the be-all and end-all of CPU cooling!

https://www.tomshardware.com/pc-com...-is-the-best-cpu-air-cooler-now-only-dollar33
​

their fans are world class, and their coolers hold resale value.

I used to think this until I tried both their NF-A9x14 PWM and NF-A9x14 HS-PWM fans. They're only good below about 1000 RPM. Above that, they whine as badly as any other fan I've heard. I saw some reviewers claim this, but I decided to try the HS version just to make sure I didn't get a dud with the first fan.

Beyond that, perhaps you didn't notice but there are other players in the fan game and they're not just sitting on the sidelines.

https://www.anandtech.com/show/21198/the-alphacool-apex-stealth-metal-120-mm-fans-capsule-review

https://www.anandtech.com/show/21376/capsule-review-sunon-maglev-120mm-fan
​

I resold mine for the same price I bought it at 3 years later (got to love inflation...).

Must've been due to some pandema-era supply chain craziness. I doubt that would happen today.

 

[TheHerald]

Air flow across the top of the fins is still doing useful heat transfer, blocking out that small portion will not appreciably improve the fans static pressure to make up for the reduced airflow over the external fins. As others have suggested, doing a duct design enables overcoming the impedance issues open air coolers deal with, blocky + ugly, really only make sense for the OEMs(many Dell and HP SFF use to utilize chambers for this very reason).

If anyone is doing CFDs on their design choices it would be noctua, I only wish they would up their lowprofile game, as other choices are superior than their meager offers at sub 55mm height, but their fans are world class, and their coolers hold resale value. I resold mine for the same price I bought it at 3 years later (got to love inflation...).

Their fans are definitely not world class, only a12x25 is a top contender, every other fan they have is a very old design that isn't competitive anymore. Noctua is a great company but they move waaaay too slow - between each release they are just getting leapfrogged by the competition.

 

[bit_user]

on the bottom part i keep wondering why all manufacturers refuse to try and somehow attach more fins as close to the actual base/90ish degree bend of the pipes, even if they are smaller fins in every metric and spread wider, half clipped from one side etc, also why they dont enlarge the socket part above the heatpipes with an chipset like heatsink (old chipsets) around (or being part of) the mounting mechanism, with the middle fan having a bottom cavity in the frame to allow some airflow to go down and the outer fan already pushing air in that direction, i mean those are the closest parts to the actual heat source and somehow are the ones with less dissipation capacity equipped.

My guess about why they didn't is factors like cost and weight. Otherwise, I'd have to agree that you might as well take advantage of that airflow that's mostly just going to waste.

Another doubt/curiosity i have... do heatpipes work better if seated on a horizontally laid heat source (cpu-gpu) with the heatpipes bending vertically up (not down like in current GPU's), just like in aircooled server racks,

From what I've read, it really doesn't make a significant difference. The fluid returns by capillary action and I suppose the mass is low enough that the force of gravity isn't a significant factor. I assume you've seen the way droplets of water can stick to the surface of a glass or some other object and defy gravity? Now consider the way water clings to something with lots of micro-grooves, like unfinished wood!

Also why they keep just aligning the heatpipes (in most DT coolers) straight parallel instead of sig zagging them for better fin area heat dissipation

Maybe it would interfere with return of the working fluid? Perhaps it's also the case that heatpipes are difficult or time-consuming to work with, and adding more bends increases costs or the likelihood of defects?

Also, I'm not sure you're really optimizing the bottleneck. I suspect the limiting factor could be in the base, at the point of heat transfer into the heat pipes. Maybe the reason why more capacious heatsinks use more heatpipes is all about the base and drawing more heat out of the CPU, rather than bottlenecks in the fin stack.

Also why straight fins and no slight wave shape to them like a soft horizontal "~" shape (maybe done in both axis), forcing some air turbulence within the fins and making the air bounce more over the surface of the fin stack.

Again, I'm not sure where the bottleneck actually is. Maybe the surface area of the fins is adequate for the amount of heat they can each draw off of the heat pipes, and the better way to optimize them would be just to have more of them?

Also, any time you're proposing to increase the amount of material, you have to consider cost and weight. Regarding weight, just imagine shipping a PC with a > 1 kg mass attached to the motherboard, torquing it this way and that, as it's tossed around or dropped.

 

[Amdlova]

Got here a CORSAIR ML 120 First GEN no RGB CRAP! and The spinnig boi do 2450 RPM noise intensifies after 50%^of rated PWM at 850 RPM is inaudible.
The best fan noise rpm ever I have found is the akasa piranha =)

https://ibb.co/wYdDzT6 ML120 (original one) AT Thermalright SL-100 =)

 

[eX_Arkangel]

From what I've read, it really doesn't make a significant difference. The fluid returns by capillary action and I suppose the mass is low enough that the force of gravity isn't a significant factor. I assume you've seen the way droplets of water can stick to the surface of a glass or some other object and defy gravity? Now consider the way water clings to something with lots of micro-grooves, like unfinished wood!

In all events gravity always helps specially with easy to stick surfaces for water, as the droplets build up gravity forces them to fall sooner than if just sitting there,, but as you said maybe the overall improvement is negligible, and not really worth the R&D plus assembly extra expenses.

Maybe it would interfere with return of the working fluid? Perhaps it's also the case that heatpipes are difficult or time-consuming to work with, and adding more bends increases costs or the likelihood of defects?

Also, I'm not sure you're really optimizing the bottleneck. I suspect the limiting factor could be in the base, at the point of heat transfer into the heat pipes. Maybe the reason why more capacious heatsinks use more heatpipes is all about the base and drawing more heat out of the CPU, rather than bottlenecks in the fin stack.

On this one i probably failed to explain what i meant by zig zagging the heatpipes, its not bending them, but aligning them with the fin stack in an irregular pattern maximizing fin area usage (imagine a heat circle around each heatpipe in the fin stack, now imagine rearranging the position of the tubes so the circle of each heatpipe crosses as little as possible with the circles of the other tubes), hence dissipating more heat without using any extra material, bend in the pipes, etc.
Im not after the bottleneck as i really cant tell were it truly is, but after every improvement left (possibly) to be made or looked at.
I agree with you in the more heatpipes use, also making thinner contact bases to get the actual heatpipes as close as posible to the heat surface (not direct die contact that almost always has worse contact that with a proper soldered base, and adding the mentioned extra mini- heatsink above the base in the unused space between the tubes (well used only for the mounting kit as of now).
here a top fin view with pipes.
[-_-_-_-_-] zig zagged pipes, [- - - - - - -] straight aligned pipes

Again, I'm not sure where the bottleneck actually is. Maybe the surface area of the fins is adequate for the amount of heat they can each draw off of the heat pipes, and the better way to optimize them would be just to have more of them?

Also, any time you're proposing to increase the amount of material, you have to consider cost and weight. Regarding weight, just imagine shipping a PC with a > 1 kg mass attached to the motherboard, torquing it this way and that, as it's tossed around or dropped.

For now my only extra materials would be the extra heatsink right above the base of the heatpipes mixed with the mounting system, which would have to be small by any means, as it has either the middle Fan on top or the fin stack in ST coolers as the height limit, the extra fins as close to the bottom of the heatpipes as possible would be like the ones currently used for the Dram offset, smaller overall and widely spread (lets say 2 to 2.5mm between each for easier airflow as close to the base the separation between the heatpipes gradually disappears), assuming all of this is just lightweight aluminum i dont see any extra problem with the extra weight or production cost.
Most High End Air coolers already weight 1kg, so that problem already exist in shipping, adding and extra 50-100g wont really change it.

 

[bit_user]

In all events gravity always helps specially with easy to stick surfaces for water, as the droplets build up gravity forces them to fall sooner than if just sitting there,, but as you said maybe the overall improvement is negligible, and not really worth the R&D plus assembly extra expenses.

If gravity made a significant difference, I'm sure there would be products on the market exploiting it and telling you which orientation they're designed for.

On this one i probably failed to explain what i meant by zig zagging the heatpipes, its not bending them, but aligning them with the fin stack in an irregular pattern maximizing fin area usage (imagine a heat circle around each heatpipe in the fin stack, now imagine rearranging the position of the tubes so the circle of each heatpipe crosses as little as possible with the circles of the other tubes), hence dissipating more heat without using any extra material, bend in the pipes, etc.

Okay, I've seen them staggered like you describe. In this case, it looks like Noctua decided it was more important for each heat pipe to be centered within its tower.

also making thinner contact bases to get the actual heatpipes as close as posible to the heat surface (not direct die contact that almost always has worse contact that with a proper soldered base,

Here's where I wish cooler and CPU makers could agree on where the heat-spreading should happen, so we have it only once. If we have an IHS, you'd hope that spreads the heat well enough that the heatsink wouldn't also need its base to do the same thing, because now you're just introducing another surface. Of course, the best case scenario is probably a vapor chamber. So, it would go die -> vapor chamber -> heat pipes.

and adding the mentioned extra mini- heatsink above the base in the unused space between the tubes (well used only for the mounting kit as of now).

I've seen some heatsinks do this. Not sure how much additional dissipation you can get from that. It probably comes down to a price/performance term that Noctua is trying to optimize. If something adds cost but not at least equivalent performance, don't do it.

I wish some expert heatsink maker would be willing to share all of their secrets, but I'm sure the only way that would happen is if they were exiting the business. I have lots of questions.

 

[Eximo]

I think the issue with a direct die vapor chamber on high end enthusiast chips is that they would always be overwhelmed because they would be too small. They work on GPUs because there is enough area to actually get a vapor cycle going.

So for that to work, your vapor chamber or heat pipe cooler should be in direct contact with the die. But that is too risky for the typical setup.

Direct die AMD days, so many people cracked their chips. I want to say my 1800+ was missing a corner though not deep enough to cause issues.

IBM did an experiment running water cooling through silicon vias on stacked chips. It may come down to something like that, but at some point I think CPU sockets might go away, which would solve a lot of problems with building something like a large vapor chamber on something like a big CPU card if we want to keep modularity. I think Intel was going down the right path with their compute board, just a little too early to be worthwhile.

 

[bit_user]

I think the issue with a direct die vapor chamber on high end enthusiast chips is that they would always be overwhelmed because they would be too small. They work on GPUs because there is enough area to actually get a vapor cycle going.

AMD had gotten one to work as an IHS for AM5, but it offered too little advantage probably due to the incredibly tight physical margins they had to work within.

https://www.techpowerup.com/311047/amd-explored-vapor-chamber-cooling-design-for-zen-4-cpus
​

I think performance Laptops frequently use vapor chambers, but they're working with a (nearly) bare die, so it's pretty much the same thing.

Server CPUs aren't a good analogy, because they always come with an IHS, which already reduces any hotspot problem that could occur. If the IHS is playing an essential role for vapor chambers used in servers, then you could get the equivalent effect for direct-die vapor chambers by simply making the bottom of the chamber thicker. However, its thickness can be precisely tuned to the vapor chamber, whereas the IHS might be overkill for what's needed.

So for that to work, your vapor chamber or heat pipe cooler should be in direct contact with the die. But that is too risky for the typical setup.

Direct die AMD days, so many people cracked their chips. I want to say my 1800+ was missing a corner though not deep enough to cause issues.

Laptop CPUs don't have an IHS, although there is a blob of protective epoxy. If hundreds of millions of laptops can ship that way without a high failure rate, then why not also desktops? Perhaps because running such a CPU without a heatsink would be instant death, whereas having an IHS can provide enough thermal mass that the CPU has a chance to throttle-back enough to prevent damage.

What I hope someone does (though I doubt they will), is to have a different cooler spec for the premium performance CPUs, which has the additional spacing needed to accommodate CPUs that come with a decent vapor chamber permanently attached (i.e. instead of IHS). The best coolers for these CPUs will have direct-contact heat pipes.

 

[Eximo]

Laptop CPUs don't have an IHS, although there is a blob of protective epoxy. If hundreds of millions of laptops can ship that way without a high failure rate, then why not also desktops? Perhaps because running such a CPU without a heatsink would be instant death, whereas having an IHS can provide enough thermal mass that the CPU has a chance to throttle-back enough to prevent damage.

What I hope someone does (though I doubt they will), is to have a different cooler spec for the premium performance CPUs, which has the additional spacing needed to accommodate CPUs that come with a decent vapor chamber permanently attached (i.e. instead of IHS). The best coolers for these CPUs will have direct-contact heat pipes.

Not a matter of shipping with laptops. They are already assembled when you get them and they don't expect the typical end user to be going in there and changing the thermal paste. With shipping desktops it is all about the torque a large tower CPU cooler can have. If they didn't have an IHS, that load could indeed crack the die. Now some improved mounting method that allows for bare die and a hard mount through the motherboard to the chassis could work.

OEM desktops aside, and the boutique solution has been to switch to AIO coolers for custom rigs anyway, I am talking about people building PCs. If bare dies increases the amounts of RMA, even if they get rejected for customer induced damage, that still costs companies money.

That brings us back to CAMM2, where I can see big vapor chamber placed on the general area. Though that would still be an enthusiast level product, and you would still be delidding your CPU yourself.

 

[bit_user]

With shipping desktops it is all about the torque a large tower CPU cooler can have. If they didn't have an IHS, that load could indeed crack the die. Now some improved mounting method that allows for bare die and a hard mount through the motherboard to the chassis could work.

Good point.

That brings us back to CAMM2, where I can see big vapor chamber placed on the general area. Though that would still be an enthusiast level product, and you would still be delidding your CPU yourself.

If you mean having a vapor chamber covering both the CPU and RAM, I think that doesn't make a lot of sense. Much of the time, the RAM would probably be sitting at below the boiling point of the working fluid, so it would be a point of condensation and your CPU would effectively just be heating it. Then, when you want to stress it, it's starting from a higher baseline and perhaps only deriving somewhat minor benefits from the vapor chamber as opposed to just having its own heatsink.

A vapor chamber is also probably overkill for RAM cooling, anyhow. Just a regular aluminum heatsink should be more than enough. Sadly, I think the physical design margins aren't sufficient to support a wraparound section for backside cooling, like some double-sided M.2 coolers have. However, if a CAMM2 module is designed to be used with a heatsink, then perhaps the manufacturer can find a way to improve thermal conduction between the heatsink and the module's PCB.

BTW, does anyone know what the max power spec is on CAMM2 and LPCAMM2? If we really want to talk about cooling them, we should start by knowing the maximum amount of heat that might need to be dissipated!

 

[TheHerald]

Not a matter of shipping with laptops. They are already assembled when you get them and they don't expect the typical end user to be going in there and changing the thermal paste. With shipping desktops it is all about the torque a large tower CPU cooler can have. If they didn't have an IHS, that load could indeed crack the die. Now some improved mounting method that allows for bare die and a hard mount through the motherboard to the chassis could work.

OEM desktops aside, and the boutique solution has been to switch to AIO coolers for custom rigs anyway, I am talking about people building PCs. If bare dies increases the amounts of RMA, even if they get rejected for customer induced damage, that still costs companies money.

That brings us back to CAMM2, where I can see big vapor chamber placed on the general area. Though that would still be an enthusiast level product, and you would still be delidding your CPU yourself.

Yeah, it's about the torque and actually moving the case. If you try to move it with a bare die and a cooler attached, kiss your die goodbye, lol.

 

[Silas Sanchez]

I got the NH-U12A chroma black and I love its rather medium size and how well it cools my 7950x. Doing some tests show limiting its fans at only ~80% speed at full hard bench marking makes no practical difference in process times, maybe a few seconds off while doing the same work. I cant see this making much of a difference, its clear if you want a noticeable difference you need water cooling. I'm thinking of going water cooled to keep my cpu temp down in the name of prolonging the life of my system, less heat means less potential for cpu degradation in longterm and mobo damage from high heat.
We have never seen cpus running this hot before, its just happened, no longterm data or anything, best to play it safe if you want max longevity.

Air coolers can only do so much, they cant compare to the constant bombardment of fresh water. Water has truly terrible thermal conductivity and diffusivity and yet it doesn't really matter because constant flow of cool water in the block just does wonders.

And what is with the silvery look with brown fans? yuk!

​

 

[35below0]

I'm thinking of going water cooled to keep my cpu temp down in the name of prolonging the life of my system, less heat means less potential for cpu degradation in longterm and mobo damage from high heat.
We have never seen cpus running this hot before, its just happened, no longterm data or anything, best to play it safe if you want max longevity.

This may backfire if the extra temperature headroom causes the CPU to push even harder than before.
Pay attention to your motherboard power preset. More often than not it will be set to "as hard as possible until limited by temperature".

Air coolers can only do so much, they cant compare to the constant bombardment of fresh water. Water has truly terrible thermal conductivity and diffusivity and yet it doesn't really matter because constant flow of cool water in the block just does wonders.

The largest and best liquid coolers can only do so much too. Top tier air coolers only lose a few degrees C in cooling compared to best liquid AIOs.
About half of AIO on the market do no better at all. The real benefit is removing the towerblock with 50 families living in it that is an air cooler today. Also, you can physically move your PC case around without fear of ripping the CPU out of it's socket.

There are pros and cons to both. I prefer air because it doesn't have issues with pumps or radiator positioning, but the space taken up is a massive drawback. The Noctua NH-D15 dwarfs an ATX motherboard. And it's a pain in the nether regions to reconfigure the fans, the RAM, the NVMe drive or anything in it's vicinity. Which is tricky when everything IS in it's vicinity because of it's size.

Great cooler though.

 

[cAllen]

Running the first gen D15 in my current/last build listed below...which is a 24/7/365 machine. If for some reason I ever build another one, I will use this Gen 2.

 

[bit_user]

I got the NH-U12A chroma black and I love its rather medium size and how well it cools my 7950x. Doing some tests show limiting its fans at only ~80% speed at full hard bench marking makes no practical difference in process times, maybe a few seconds off while doing the same work. I cant see this making much of a difference, its clear if you want a noticeable difference you need water cooling. I'm thinking of going water cooled to keep my cpu temp down in the name of prolonging the life of my system, less heat means less potential for cpu degradation in longterm and mobo damage from high heat.
We have never seen cpus running this hot before, its just happened, no longterm data or anything, best to play it safe if you want max longevity.

If you want max longevity, you might be better off just limiting the max temperature of your CPU. That will also reduce strain on your motherboard, since you're worried about that.

As @35below0 said, water cooling will increase potential for straining your VRM, but it's worse than that. It will also potentially reduce air movement in the vicinity. For that reason, I prefer downdraft coolers, since they also cool RAM, VRMs, and my M.2 drive.

Water has truly terrible thermal conductivity and diffusivity and yet it doesn't really matter because constant flow of cool water in the block just does wonders.

But it has a high specific heat, which is useful in cooling systems.