News Zotac's AirJet-equipped mini PC tested – cools N300 CPU in chassis less than an inch thick

[Admin]

AirJet fanless coolers have finally been tested in one of Zotac's mini PCs, and the results are impressive.

Zotac's AirJet-equipped mini PC tested – cools N300 CPU in chassis less than an inch thick : Read more

 

[Bluoper]

I'm really exited to see this used in ultra thin laptops to give them a bit of extra horse power.

 

[gg83]

Is the N300 a good example to use? Can it run hot under load without active cooling? It's about time we used other tech to move air, and eliminate the need for liquid.

 

[usertests]

Is the N300 a good example to use? Can it run hot under load without active cooling? It's about time we used other tech to move air, and eliminate the need for liquid.

I think it's the best they can do right now. One AirJet Mini can handle 5W, two can handle 10W. Source: https://www.froresystems.com/blog/t...r-has-arrived-with-features-youd-never-expect

They claimed in one interview that they intend to scale the capabilities with newer process nodes. I hope it's widespread soon enough to see use in things like a Steam Deck 2.

 

[TechLurker]

It'll be interesting if the performance scales up enough to implement on a larger macro scale; such as some sort of "120mm/140mm replacement" that can suction air through radiators/fins and exhaust it 90% off to a side, top, or rear panel (which most modern cases currently have vents on).

Beyond that, I'd actually like to see this integrated into mobo cooling; both chipsets, VRMs, and NVMe, adding some much-needed airflow across fins and heatsinks in lieu of tiny 40mm fans.

 

[wr3zzz]

Is the N300 a good example to use? Can it run hot under load without active cooling? It's about time we used other tech to move air, and eliminate the need for liquid.

That's what I am thinking too. Intel 7W TDP is specifically for passively cooled notebooks so AirJet is kinda pointless. Can't they put 3 chips to cool the mainstream 15W TDP ultrabook CPU? Or is there not enough surface space to fit 3 chips or 2x2?

 

[bit_user]

Is the N300 a good example to use?

It is, if you don't simply drink the Intel cool aid. Remember, this is the same company claiming the i9-14900K runs at 125 W.

Can it run hot under load without active cooling?

It'll hit 20 W, easily.

Granted, this review is for the N305 (TDP: 15W), but should give you an idea of how flexible these numbers really are.

Beelink EQ12 Pro Review - World's First Mini PC with Core i3-N305 Processor

As the most powerful chip in the Intel Alder Lake-N low-power family, the Core i3-N305 has attracted much attention since its release. It may be because the current cost of chips is relatively high, so only a few computers on the market have chosen this solution. Beelink EQ12 Pro is the world's...

minixpc.com

With a little trouble, I managed to use Google Translate to decode the power chart, at the bottom:

Application Scenario	Average Power (W)​	Noise (dB)​
Standby/router	9.9​	N/A​
Online video playback (4k)	13.3​	36.7​
Game (Genshin Impact)	24.9​	39.9​
CPU fully loaded	40.5​	41.6​

Okay, so that's a "15 W" N305 running at up to 40.5 W. Need I say more?

It's about time we used other tech to move air, and eliminate the need for liquid.

This is at the wrong end of the spectrum, for that.

As an aside, liquid cooling isn't going anywhere. Air-cooling still has limits, assuming you don't want to sit next to something sounding like a jet engine. Yes, there are air-cooled servers which dissipate kW of heat via air, and you'd have to use hearing protection to be around them all day.

 

[bit_user]

It'll be interesting if the performance scales up enough to implement on a larger macro scale; such as some sort of "120mm/140mm replacement"

I think the economics of scaling it up are rather poor.

Beyond that, I'd actually like to see this integrated into mobo cooling; both chipsets, VRMs, and NVMe, adding some much-needed airflow across fins and heatsinks in lieu of tiny 40mm fans.

High-end devices with modest cooling requirements will certainly get this treatment. We've already seen them on M.2 SSDs, for instance.

As for motherboards and fans... it's hard for me to see how you can really beat a fan, in terms of price, performance, or reliability. The only way I'd be sold on these is if they have some sort of property for combating dust build-up, because that's a perennial weakness of fans. However, in all the coverage I've read of these coolers, I don't recall seeing the manufacturer mention dust even once. If they were somehow more resistant to dust-buildup, I'd expect them to be crowing about it. Perhaps it'll turn out to be their Achilles heel, instead!

 

[usertests]

It'll hit 20 W, easily.

Granted, this review is for the N305 (TDP: 15W), but should give you an idea of how flexible these numbers really are.

Alder Lake-N is unusual in that Intel decided not to publish turbo TDPs, which it had done as a new policy with all the other Alder Lake chips over a year earlier.

N300 has the lower 7W "TDP" (N50, N100, and N200 have a slightly lower 6W "TDP"). N305 also has the "configurable TDP-down" of 9W.

Here's an N300 review showing similarly high power consumption, maybe because PL1 is set at 10W and PL2 at 25W:

iKOOLCORE R2 Review: A 3 inch mini PC with four 2.5 GbE Ethernet ports and up to Intel Core i3-N300 - Liliputing

iKOOLCORE R2 Review: A 3 inch mini PC with four 2.5 GbE Ethernet ports and up to Intel Core i3-N300

liliputing.com

• Powered off (shutdown)* – 2.6 Watts
• UEFI (BIOS) – 13.9 Watts
• GRUB menu – 15.4 Watts
• Idle – 10.0 Watts (Ubuntu)
• CPU stressed** – 29.5 Watts (Ubuntu “stress”)
• Video playback*** – 27.2 Watts (Ubuntu Chrome 4K60fps)
• Proxmox – 12.6 Watts (No VMs)
• Proxmox (firewall) – 16.1 Watts (pfSense VM)
• Proxmox (gaming) – 33.4 Watts (pfSense VM & Batocera VM)

Reviews that compare power consumption and benchmark performance of the Zotac with AirJet against other N300 or N305 systems with fans should settle the matter of cooling superiority.

The only way I'd be sold on these is if they have some sort of property for combating dust build-up, because that's a perennial weakness of fans. However, in all the coverage I've read of these coolers, I don't recall seeing the manufacturer mention dust even once. If they were somehow more resistant to dust-buildup, I'd expect them to be crowing about it. Perhaps it'll turn out to be their Achilles heel, instead!

This was addressed in interviews and articles. They use dust guards over inlets (and outlets?), and the much higher pressure created by AirJet allows it to suck air in fast enough to work. It is supposedly dustproof. If you were to damage the dust guards, perhaps the results would be catastrophic for the vibrating membranes inside AirJet.

This palm-sized PC might contain the future of gadget cooling

This Zotac Zbox is the first product with Frore AirJet.

www.theverge.com

 

[bit_user]

They use dust guards over inlets (and outlets?), and the much higher pressure created by AirJet allows it to suck air in fast enough to work. It is supposedly dustproof. If you were to damage the dust guards, perhaps the results would be catastrophic for the vibrating membranes inside AirJet.

Thanks. I must've missed that. I'll be keenly interested in seeing how well those actually work, in practice.

Even if they successfully manage to keep dust out of the device's active parts, how quickly do they clog up and what then? It would be clever if they could temporarily reverse the direction of airflow, in order to unclog them.

 

[DavidC1]

Okay, so that's a "15 W" N305 running at up to 40.5 W. Need I say more?

This is what the review says: "However, the power consumption value of 40W will only last for about 10 seconds, and then it will decrease and remain stable at 30W. around the level."

Also, 30W is system power, not CPU power. 30W-10W = 20W load, and even that's not entirely CPU power. 15W is very accurate in this case. Lilputing's tests also show the same thing. 33.4W peak and 10W idle = 23.4W. Higher idle power will result in higher load power because there are fixed power users.

Those are still assumptions. Accurate CPU power is only shown by using monitoring applications. My Yoga 710-11 uses 0.6W for CPU while system is at 3.8W at idle. Certainly 3.8W isn't CPU!

The sustained performance of E series are very stable, unlike the Core lineup: https://www.notebookcheck.net/Intel...ink-Mini-S12-Pro-mini-PC-review.758950.0.html

The stress test shots also show the CPU rigidly adheres to the set TDP settings. There's no concern here.

This is another system with PL1/PL2 of 15/25W, but only reaches 15.8W at max: https://www.notebookcheck.net/Geeko...n-a-well-known-Intel-NUC-design.771494.0.html

Load Maximum = 30W. Compare that figure to other laptops. 30W is a very low figure. The M3 Pro macbook uses 73W at max, which is similar to a 13700H using Iris Xe. Unlike most other systems where Load Max and Load Sustained varies substantially, this particular system increases by less than 5%. the M3/13700H goes from 40-ish watt to 70W+.

Desktops are different as you have a practically unlimited cooling budget with heatsinks heavier than an entire laptop, and the mobo manufacturers set PL1=PL2.

 

[bit_user]

This is what the review says: "However, the power consumption value of 40W will only last for about 10 seconds, and then it will decrease and remain stable at 30W. around the level."

Also, 30W is system power, not CPU power. ...

Thanks for the info.

I have a passively-cooled mini-ITX board with a J4205 (Apollo Lake; 10W-rated) and it easily runs at 25 W, sustained. That's with only one DDR-2133 SODIMM and a USB stick plugged in! No fans and no other storage devices. There might've been a gigabit ethernet cable plugged in and a HDMI monitor @ 1600x1200 (static desktop), but that's it.

Now, that was measured at the wall (powered via an 80+ Titanium-certified PSU), but I just bought a Corsair PSU with iCue, so I can further isolate what just the CPU is using.

So, I can't (currently) point to concrete evidence of PL1 violations in the Alder Lake-N series, but Intel and I have some history, here.

Accurate CPU power is only shown by using monitoring applications. My Yoga 710-11 uses 0.6W for CPU while system is at 3.8W at idle. Certainly 3.8W isn't CPU!

Laptops aren't what we're talking about, here. In this article, they're measuring a mini-PC. Laptops will tend to be configured for more efficient operation - especially something super low-profile, like a Yoga.

The sustained performance of E series are very stable, unlike the Core lineup: https://www.notebookcheck.net/Intel...ink-Mini-S12-Pro-mini-PC-review.758950.0.html

The stress test shots also show the CPU rigidly adheres to the set TDP settings. There's no concern here.

This is another system with PL1/PL2 of 15/25W, but only reaches 15.8W at max: https://www.notebookcheck.net/Geeko...n-a-well-known-Intel-NUC-design.771494.0.html

Those are N100's, but at least they're mini-PCs.

Load Maximum = 30W. Compare that figure to other laptops. 30W is a very low figure. The M3 Pro macbook uses 73W at max, which is similar to a 13700H using Iris Xe.

Talk about comparing Apples & oranges! Such high-performance laptops are a completely different subject!

 

[DavidC1]

Laptops aren't what we're talking about, here. In this article, they're measuring a mini-PC. Laptops will tend to be configured for more efficient operation - especially something super low-profile, like a Yoga.

Talk about comparing Apples & oranges! Such high-performance laptops are a completely different subject!

No man. Those are examples so you get the idea. Go scroll down to the bottom of the review to see what I am talking about. There are various Alderlake-N examples. They are mostly at the same level. Read about the TDP settings of the actual device and they loosely follow the system power consumption. Example, a similarly sized system with 10W higher TDP CPU uses 10W more under load.

System power = Idle + CPU TDP + misc

My point is that you are making incorrect assumptions that it's exceeding TDP because the system is using 40W, when it's not. And clearly the performance shows that all Atom-based systems show an extremely consistent performance that doesn't drop radically as with other CPUs. For example, the Core CPUs go from 1.3x performance and 1.5x TDP but after 30 seconds it drops to the 1x performance and 1x TDP as rated. Actually on the Atom CPUs the variation between peak and sustained performance is less than 5%!

All previous Atom-based systems show this. Laptop, desktop, AIO, it doesn't matter.

Same with my laptop. You can't say the CPU uses 4W. No, it uses 0.6W. You post often here, so I'm giving you the benefit of the doubt you get that it's an example.

Laptops have to follow TDP settings, because it's a constrained form factor. Desktops can "violate" it because it has infinite headroom, but no matter who wishes, thermal laws cannot be violated. Sure, Intel gives manufacturers some flexibility in setting TDP, but that's an aside.

Hence, why for Core it can use 40W for 30 seconds, but has to come down to a thermally manageable level, say 25W on sustained loads. On one of the reviews for ADL-N the author points out due to thermals, it doesn't even use the said TDP.

I have a passively-cooled mini-ITX board with a J4205 (Apollo Lake; 10W-rated) and it easily runs at 25 W, sustained.

PSUs are under efficiency ratings at such low load. For an mini-ITX board like yours, maximum you can't exceed more than a 100W PicoPSU with a Class V or higher adapter to get the idle power well under 10W. Ideally you'd get a 60W one. You did not say what your windows desktop idle power is, but I bet it's close to that 10W level.

I know this partially because I wanted a similar system.

Here a J5005-ITX uses 12W on idle and similar-to-yours 26.5W on full load: https://androidpctv.com/review-asrock-j5005-itx-j4105-itx/

The gap between idle and load is 14.5W. That's with a PicoPSU 160W. I believe with a PicoPSU 60W and efficient adapter you can get idle down to 6-8W. Mind you, at 10W, even a 160W is only under 6% load. Corsair Titanium means 400W at least, so 10W is 2.5% which is nothing. Titanium label with 90% plus efficiency rating would only apply after 15% load or so.

 

[bit_user]

System power = Idle + CPU TDP + misc

My point is that you are making incorrect assumptions that it's exceeding TDP because the system is using 40W, when it's not.

Thanks - I get that. I only replied to the parts of your post I took issue with.

Actually on the Atom CPUs the variation between peak and sustained performance is less than 5%!

Depends quite a bit on how those peak and sustained values are defined, but it seems very unlikely. ChipsAndCheese ran some extensive analysis to look at performance scaling on Gracemont and found it's rather more linear vs. power.

Source: https://chipsandcheese.com/2022/01/28/alder-lakes-power-efficiency-a-complicated-picture/
​

The first thing to point out is that article is comparing 4-core clusters. So, the E-core measurements would be roughly equivalent to an entire 4-core Alder Lake-N SoC, rather than just one of the cores.

Given your example of a SoC with PL1/PL2 of 15/25 W, we should expect to see a 10% performance difference on x264 transcoding and a 15% performance difference on 7zip compression. If it were only 5%, that would be pretty terrible, in fact. As it is, it's rather poor that 66.7% more power nets you such meager gains.

PSUs are under efficiency ratings at such low load.

That's true that the lesser certifications only go down to 20% load, but 80+ Titanium actually specifies at least 90% efficiency at 10% load.

80 Plus - Wikipedia

en.wikipedia.org

What it means is that if my system is using 20 W, the most the PSU can be drawing from the wall is 22.2 W. Assuming it's using at least 10% of the PSU's rated capacity, that is. I'll grant you the PSU is bigger than that, but it still gives an idea that the PSU isn't throwing the measurements
way off.

You did not say what your windows desktop idle power is, but I bet it's close to that 10W level.

Linux.

I could dig up my old data, but it's pointless since I'm about to do a new set of measurements with my new Corsair iCue PSU. I'll post a thread about it and @ you when I do. Probably closer to the end of the year, when I get some more time off.

 

[thestryker]

Seems like the configuration for these Zotac boxes are 7W/11W so they should stick pretty close to that in operation.

And clearly the performance shows that all Atom-based systems show an extremely consistent performance that doesn't drop radically as with other CPUs. For example, the Core CPUs go from 1.3x performance and 1.5x TDP but after 30 seconds it drops to the 1x performance and 1x TDP as rated. Actually on the Atom CPUs the variation between peak and sustained performance is less than 5%!

All previous Atom-based systems show this. Laptop, desktop, AIO, it doesn't matter.

This is an extremely inaccurate take as it has nothing to do with the CPUs and everything to do with how the OS and BIOS are setup. Most Atom based machines are tightly controlled, but that has nothing to do with the chips themselves. The N5105 box I'm currently using as a router has a 20W boost profile out of the box which can also be user modified. It also performs notably faster at ~2.9ghz than it does ~2ghz which I assume is accurate for all Tremont based CPUs that have boost capability.

You can also see the power/performance disparity in a couple of ServeTheHome's reviews of N200 boxes. The one from Asus has lower performance than the one on a board by CWWK.

 

[gg83]

It is, if you don't simply drink the Intel cool aid. Remember, this is the same company claiming the i9-14900K runs at 125 W.


It'll hit 20 W, easily.

Granted, this review is for the N305 (TDP: 15W), but should give you an idea of how flexible these numbers really are.

​

Beelink EQ12 Pro Review - World's First Mini PC with Core i3-N305 Processor

As the most powerful chip in the Intel Alder Lake-N low-power family, the Core i3-N305 has attracted much attention since its release. It may be because the current cost of chips is relatively high, so only a few computers on the market have chosen this solution. Beelink EQ12 Pro is the world's...

minixpc.com

​

With a little trouble, I managed to use Google Translate to decode the power chart, at the bottom:

​

​

​

Application Scenario	Average Power (W)​	Noise (dB)​
Standby/router	9.9​	N/A​
Online video playback (4k)	13.3​	36.7​
Game (Genshin Impact)	24.9​	39.9​
CPU fully loaded	40.5​	41.6​

Okay, so that's a "15 W" N305 running at up to 40.5 W. Need I say more?


This is at the wrong end of the spectrum, for that.

As an aside, liquid cooling isn't going anywhere. Air-cooling still has limits, assuming you don't want to sit next to something sounding like a jet engine. Yes, there are air-cooled servers which dissipate kW of heat via air, and you'd have to use hearing protection to be around them all day.

Thanks for the great response!

 

[TechLurker]

As for motherboards and fans... it's hard for me to see how you can really beat a fan, in terms of price, performance, or reliability. The only way I'd be sold on these is if they have some sort of property for combating dust build-up, because that's a perennial weakness of fans. However, in all the coverage I've read of these coolers, I don't recall seeing the manufacturer mention dust even once. If they were somehow more resistant to dust-buildup, I'd expect them to be crowing about it. Perhaps it'll turn out to be their Achilles heel, instead!

In addition to the response by usertests, the company mentioned in one of their videos that there is a built-in dust-buster mode where it temporarily reverse air flow to punch away any fine build-up on the thin mesh covering the intakes. It only takes a few seconds to reverse the flow for the intake cleaning, then shift back to normal flow and can be done every few minutes without affecting temperature. I forget if it's the LTT video where they get to play around with the AirJets, or one of their promo videos, but it's mentioned as a built-in feature given how fine their intakes are.

 

[bit_user]

the company mentioned in one of their videos that there is a built-in dust-buster mode where it temporarily reverse air flow to punch away any fine build-up on the thin mesh covering the intakes. It only takes a few seconds to reverse the flow for the intake cleaning, then shift back to normal flow

That's exactly what I had in mind. I could've even read about it and simply forgotten that I did.

It will be interesting to see how well it works, over the course of long-term usage.

 

[DavidC1]

This is an extremely inaccurate take as it has nothing to do with the CPUs and everything to do with how the OS and BIOS are setup. Most Atom based machines are tightly controlled, but that has nothing to do with the chips themselves.

I'm talking about specific machines tested, as NBC shows.
https://www.notebookcheck.net/Geeko...n-a-well-known-Intel-NUC-design.771494.0.html

Section titled "Geekom MiniAir 12 during the Cinebench R15 Multi Loop"

The variation in sustained Cinebench R15 loop shows a ~5% gap: 462-474 points.

The CPU has to stay stable otherwise such figures can't be shown. This is unlike Core-based machines where you go from say a 650 point for the first 20 seconds and drops to 500 sustained after that.

 

[thestryker]

I'm talking about specific machines tested, as NBC shows.
https://www.notebookcheck.net/Geeko...n-a-well-known-Intel-NUC-design.771494.0.html

Section titled "Geekom MiniAir 12 during the Cinebench R15 Multi Loop"

The variation in sustained Cinebench R15 loop shows a ~5% gap: 462-474 points.

The CPU has to stay stable otherwise such figures can't be shown. This is unlike Core-based machines where you go from say a 650 point for the first 20 seconds and drops to 500 sustained after that.

When you look at the graph you can see clearly it's hitting the set CPU power limit and just staying there. That's the only thing being shown in that benchmark.

 

[bit_user]

I'm talking about specific machines tested, as NBC shows.
https://www.notebookcheck.net/Geeko...n-a-well-known-Intel-NUC-design.771494.0.html

Section titled "Geekom MiniAir 12 during the Cinebench R15 Multi Loop"

The variation in sustained Cinebench R15 loop shows a ~5% gap: 462-474 points.

That doesn't show any boosting-type behavior, for whatever reason.

If you're talking about boost vs. non-boost performance (which is the only reason I can see why you'd be talking about peak vs. sustained performance), that data doesn't show it.

The CPU has to stay stable otherwise such figures can't be shown. This is unlike Core-based machines where you go from say a 650 point for the first 20 seconds and drops to 500 sustained after that.

PL1 vs. PL2 is determined by a 3rd variable, Tau. In a situation like you mention, Tau would be 20 seconds.

 

[MartenKL]

I am sceptical, couldn't this have been passively cooled? I am not impressed with this tech as it still shows very small air movements. The promise of a bigger one has been around for ages, feels like the promise of fusion by now. This CPU is on par with a raspberry Pi 4 which most people cool passively, some even without a heatsink.

 

[bit_user]

I am sceptical, couldn't this have been passively cooled?

Yes, with an airy case and a big enough heatsink. However, the value of these air jets is they're super low-profile and clearly aimed at laptops and similarly space-constrained devices.

This CPU is on par with a raspberry Pi 4 which most people cool passively, some even without a heatsink.

No, on both counts.

First, the Pi 4 performs like a Core 2 quad, while the N300 is more like a Skylake i7 (non-K). Even if you meant to say Pi 5, that only gets you to Sandybridge i5-class performance.

The Pi 4 also uses way less power than the N300's peak (PL2) of 25 W.

Source: https://edc.intel.com/content/www/u...core-i3-n-series-datasheet-volume-1-of-2/001/
​

The Pi 4 has a peak power of just 8 W, but you're unlikely to see more than 6.25 W in normal usage conditions. And yet, if you try to run the Pi 4 without a heatsink, it throttles like crazy!