News [pcworld] MEMS cooling device for thin devices

[gorobei]

21 db
1750 pascal back pressure
5 and 10 watt rejection models

Better Than Fans? New 'AirJet' Chip Promises To Overhaul Laptop Cooling

The San Jose-based Frore Systems claims the new cooling technology can deliver a 50% to 100% increase in performance for laptops without generating much sound.

www.pcmag.com

microelectromechanical systems actuated membranes create pulses of air at high velocity to break thru boundry layer. something like this would be a godsend to the steam deck.

 

[Stuka87]

I think it was GN or LTT that had a video about a similar way of cooling. Its a neat idea, especially if it results in lighter weight laptops due to them not needing as much copper.

 

[gorobei]

pc world follow up with trip to Frore Systems hq. no real details on engineering of the mems, but a few examples of how they can be implemented in laptops.

4 modules for a 20watt load at far less noise than a fan. supposedly the costs arent outside of what laptop oems are willing to spend.

i think the big improvement is there is no need for intake openings on the bottom of the laptop so you can use it in bed/lap with no worry over blocking the intake. this on a 2-in-1 tablet folder like the lenovo yoga would solve some issues.

 

[Heartbreaker]

i think the big improvement is there is no need for intake openings on the bottom of the laptop so you can use it in bed/lap with no worry over blocking the intake. this on a 2-in-1 tablet folder like the lenovo yoga would solve some issues.

Plenty of active cooled laptops/2-in-1s have no bottom vents. Look at a tear down of an Apple M1/M2 Macbook Pro 13, or a MS Surface Pro 9.

IMO the biggest improvement is silence in an active cooled device.

 

[gorobei]

frore airjet module in actual product: zotac zbox

the thermal imaging showing how far the exhaust projects out is pretty impressive. handhelds like the switch/steamdeck/ally need this asap.

 

[mindless1]

Something fishy about this, like half snake oil.

After watching the video I also saw another, here:
There they are pretending that a chassis 'sunk CPU could only run at 2.5W which is absurd. I know, not really, instead they didn't want to compare that, just one designed for this cooling method but with it unplugged, which is craziness. Take ANY cooling designed for airflow and stop the airflow and yeah, same result.

It also strikes me as strange that this cooling requires a dust-proof membrane, which I'm sure it would to keep dust from clogging the smaller passageways, but then this dust proof membrane itself will clog up all that much faster, so it looks like a tech that works in a dust free lab for a period long enough to do a test only, as long as you're testing but not so much in the real world where there is dust.

Doesn't really make sense in the video I saw,

that you can use a heatsink without those expensive active cooling chips, merely make the heatsink that amount taller with fins to accomplish the same at lower cost.

Obviously anyone could rig up a way it only stays cool enough with the active chip cooling it. Thing is, nobody would design it that way, instead design for normal cooling. In one video he stated it weighed less... ??? What do I care if my microPC weighs a few ounces more? When is this even relevant unless I have it in a solar powered RC/passive-code-instructed scale airplane?

If the race is to make it ever smaller, no matter cost or longevity, then I see it as a fail, much like moving to mUSB, mHDMI, etc. I could see it for portable devices except that ultimately the trend there is going to be using as client with host or intermediary processing.

Regardless, thanks for providing this topic, it is interesting even if I don't buy it and think it's more of a niche tech being misapplied in the examples I saw.

 

[gorobei]

part of the point is performance and longevity.

demonstrating the performance difference against a 100% passive cooling solution is about being roughly equivalent in noise levels. the cost of custom tooling for aluminum extrusion or billet milling for a passive case vs going with a conventional plastic enclosure with a solid state fan could balance out in manufacturing cost.

the usage cases in the 2nd video shows it is meant for laptop/notebook mobile products in the 15 to 40 watt range which will usually have some sort of fan. if it is a choice between conventional fan or solidstate device, solidstate will win on (according to frore)performance, noise, and longevity by reduction of number parts affected by friction. frore says the will be able to get the price down to approximately the cost of a conventional fan cooling solution (so probably the upper end of the price range of a custom 2 fan heatpipe solution in a $1000-1700 thin and light)

i agree that the intake filtering is the real point for skepticism. while they are supposedly targeting high speed low pressure, the air has to come from somewhere and they arent providing any info on the membrane.
solidstate cooling is a thing. there are piezoelectric fans out now with no parts that rub so no friction, it would be interesting to see if the price for these industrial solutions can drop enough to be used in consumer products. frore going with a mems approach is only radical on their projected pricing.

but you only need to look at a steamdeck teardown or fan replacement video to find a usage case. the steamdeck is pulling air across 1 mm narrow x 1 cm wide x 10cm long snaking path to reach certain components that are nowhere near the exhaust vents and the heat buildup is significant enough to affect hand comfort over longer sessions. changing to a baseplate that covers the remote components and the soc and just blasting that plate with the mems cooler can increase the heat rejected and reduce the noise. for handhelds the overall length/width can dramatically affect the whole design and ergonomics, so they are very picky about thickness of the unit.

 

[Heartbreaker]

Something fishy about this, like half snake oil.

After watching the video I also saw another, here:
There they are pretending that a chassis 'sunk CPU could only run at 2.5W which is absurd. I know, not really, instead they didn't want to compare that, just one designed for this cooling method but with it unplugged, which is craziness. Take ANY cooling designed for airflow and stop the airflow and yeah, same result.

Yeah, that was garbage.

It's the equivalent of: "Lets compare a cooler with the fan on and off. Wow, see how much better it is with the fan on."

 

[aigomorla]

lol...
i think someone said this technology isn't any better then having a thin sheet of metal which vibrates on a sin wave.
Then it could act as a fan moving hot air via the sin wave movement of the thin metal.

But efficiency would be really bad... or it couldn't get scaled up even enough to blow out a match stick.

 

[gorobei]

LTT from computex

they have a mini module taken apart. 124mph airspeed. there are more details on the membrane and filtering. there is an option to blast the membrane to clear it. 5x performance claim needs some units to be relevant. future scaling potential seems pretty inviting.

im a little disappointed in Linus for not at least touching on MEMS basics. but at least he covered the boundry air aspect. for those complaining about the zotac passive example, there are 2 of the same laptop with stock and a frore.

 

[aigomorla]

I not buying the 124MPH statement.
Do you know how fast 124MPH winds are?
I do not think my 120mm San Aces server fans at full boat 4500rpm could do 124MPH.

Just so you know, 124MPH winds is upper upper Catigory 3, with the high chance of even being lower Catigory 4 hurrican / typhoon.
Building roofs start flying at 124mph, infact you probably even need to BOLT down the fan if it was to shoot out 124mph winds.
Your telling me you have this in a tiny concentrated little stream.

I do not think so....

Another example.. you need a leaf blower, for 124MPH winds.
Yes the kind you strap to your back when your blowing leaves on your lawn away, pissing off your eco friendly neighbors.

If its that great... forget installing them on heatsinks... put scaled up and enough of them under a solar powered wing for an air plane, and you'd make billions. You'd litterally bankrupt Rolls Royce / GE Jet engines, over night and make Tesla's success look like nothing.

Also Fans are never calculated on wind speed, they are calculated in CFM, as in how much cublic feet / min, the unit displaces.
But again at 124mph winds, your CFM would probably be in the upper 3 digits, which i am LOLing if they are claiming that tiny device or even a pair let alone 1000 of them could pull off.

 

[igor_kavinski]

But man, if this takes off, no more fans in desktops or GPUs or PCIe 5.0 M.2 slots!

Just air guides to help direct the airflow to where it's needed. Gets in, gets out, too fast, too furious. Just the way we want it!

 

[aigomorla]

But man, if this takes off, no more fans in desktops or GPUs or PCIe 5.0 M.2 slots!

Just air guides to help direct the airflow to where it's needed. Gets in, gets out, too fast, too furious. Just the way we want it!

you know you have to take that review with a grain of salt.
Because the regular fan was spitting out cooler air, it isn't because its less efficient.
Its because it had more vol. of air moving, that less air got saturated with heat, and hence increased the air flow.

Also better temps? that can also relate to what TIM they used, and if it was repasted versus left as is from factory.

Really take this Linus with a grain of salt, as most of his videos are hype.

 

[Heartbreaker]

I not buying the 124MPH statement.
Do you know how fast 124MPH winds are?

High winds do damage with speed and volume.

It's possible that a microdevice operating a very high frequency is creating high speed micro-stream of air.

Though the speed is kind of irrelevant without the volume, because it's moving so little air.

It's fairly obvious that the volumes they are moving are miniscule, since they can only remove 5 watts worth of heating.

This really seems like it's being oversold and it's mainly an alternative to passive cooling portable devices.

 

[gorobei]

I not buying the 124MPH statement.
Do you know how fast 124MPH winds are?
I do not think my 120mm San Aces server fans at full boat 4500rpm could do 124MPH.

the 124 mph quote needed some elaboration by Linus, it should be referring to the speed of the jet at the nozzle impinging on the copper interface surface. the speed is needed to break the boundary layer air at the surface. it is high speed and low pressure, but by creating a low pressure area by the surface radiative cooling comes into effect. boundary layer effect functions as an insulator so radiation is barely in single digit percentages and convection is the biggest effect.

these are MEMS, this is 40 year old tech. the idea started in the 1960s, but in the 80's it became manufacturable. they started with with inkjet nozzles and went thru to medical blood testing machine microfluidic pumps, car airbag actuator accelerometers, smartphone gyroscopes, hard disc read heads, and various sensors and resonators. this is simply a novel application of old tech to a new use case.

the ink droplet speed at the nozzle on inkjets is 20m/s or 44mph. that is a liquid, so air should be able to be moved even faster.

jet impingement cooling is used in industrial equipment, military systems, turbine engines. microjet is simply the application on consumer electronics. the dyson cyclonic vacuum cleaner was simply the application of vortex separation chambers from the mining/agricultural industry that had been around for a good part of the 20th century. dyson was just the first to apply/patent it to consumer products in the 80's.

High winds do damage with speed and volume.

It's possible that a microdevice operating a very high frequency is creating high speed micro-stream of air.

Though the speed is kind of irrelevant without the volume, because it's moving so little air.

It's fairly obvious that the volumes they are moving are miniscule, since they can only remove 5 watts worth of heating.

This really seems like it's being oversold and it's mainly an alternative to passive cooling portable devices.

at the current process node they are using(not stated), this is the size of the module they can have manufactured. the airjet mini has 4 modules, the airjet pro has 8. the air impingement cooling effect increases the heat transfer from the turbulent flow at the surface. i imagine by sequencing the pulses you can move the overall flow towards the outlet opening (DLP projector MEMS mirror arrays have a 9000cycles/sec switching speed so cascading a line of pulses should be pretty easy). gordon at pcworld says that frore is claiming that the extra area around the chip in the chamber creates entrainment effect (bladeless fan tech in dyson floor fans) to increase volume of airflow and reduce max temp. the flat chamber is supposed to create laminar flow out. the video in post#5 has some thermal imaging of the exhaust hitting some paper. the hot spot extends out a good 5 or 6 inches.

at 5 watts for a mini and 10 watts for the pro, the main thing is the linear scaling by using multiple modules. they had a development unit with 8 modules, so 40w.

where the airjet beats conventional fans is in the energy used, volume used, and the longevity. mems power usage is a fraction of a fan, and with no bearing surfaces to wear down the longevity is measured in decades. Linus' excitement comes from the fact that mems are on older process nodes (think 100nm era) which could be on 200mm wafers. with migration to 300mm wafers the airjet modules could be scaled up and potentially reach the 65w range of desktop cpu. for any use case where thinness is critical the frore could eventually obviate fans assuming they can tune out the noise issues.

gordon with teardown

 

[Heartbreaker]

where the airjet beats conventional fans is in the energy used, volume used, and the longevity.

Energy used? Doesn't it use 1 Watt, to cool 5 watts. That using 20% of the energy you are cooling.

My cheep 120mm tower cooler easily cools 100 Watt CPU load with a 5 watt fan, so 5% of the energy.

IMO, the advantage remains thinner for cooling previously passive mobile devices.

 

[WelshBloke]

Its the volume of air moved that defines how much heat is going to be transfered from the heat sink that defines the cooling power.

These things absolutly arent going to be replacing your CPU or GPU coolers in a desktop. They might work in a low powered laptop but even then I cant see them being of much use.

 

[gorobei]

Energy used? Doesn't it use 1 Watt, to cool 5 watts. That using 20% of the energy you are cooling.

My cheep 120mm tower cooler easily cools 100 Watt CPU load with a 5 watt fan, so 5% of the energy.

IMO, the advantage remains thinner for cooling previously passive mobile devices.

mems devices can go below 1 watt, i suspect they are listing 1w to put themselves in direct comparison to 1w ultrabook fans.

and your 120mm fan isnt going to ever be used in a portable device. the radial (squirrel cage) fans used in notebooks will cool less at a higher noise level while using far more space for the heatpipe assembly. the blades on those things is ~3mm, and need a containment chamber to develop the pressure to drive the air. with dust accumulation and oil loss the longevity of those fans is maybe 3 years before they start making annoying noise and 5-7 where dust reduces the heat rejected leading to throttling.

these things arent going to replace conventional desktop cpu coolers anytime in the next 5 years, but in 10 who knows?

the big thing is frore acutally got a contract from zotac. they also have seeming near to production ready chips. at the very least they will be able to do the next run of wafers.

 

[Heartbreaker]

the big thing is frore acutally got a contract from zotac. they also have seeming near to production ready chips. at the very least they will be able to do the next run of wafers.

To replace a passive cooling solution. Which is kind of what I see as their forte.

I'm happy to see the tech come to market and get some real world results, not terrible misleading stuff like a demo comparing running with the device on, vs the device off. That's like comparing a traditional HSF cooler Fan Off vs Fan On, and remarking how much better it is with the Fan On... IOW Garbage.

When a company resorts to this kind of junk demo, it makes me very skeptical of anything they claim, and I was already skeptical of dust performance.

While they make a big deal of how you can filter the intake air better with their device, the truth is probably more along the lines, that you MUST completely filter the intake air, or the device would quickly clog.

Anyone who has dealt with negative pressure, PC case, knows how that pulls dust in through every crack and connector, even when it's only partial negative pressure.

This cooler is going to need to basically have the computer hermitically sealed to keep that from happening, which is going to need significant reworking of design at significant cost. So it won't just be the cost of the device, but of the case sealing that will drive costs.

 

[igor_kavinski]

the big thing is frore acutally got a contract from zotac.

That gives me hope that it MAY come to a GPU, even if just the low powered ones first.

I they could design an AIO too, with an array of these AeroJets covering the radiator.

 

[soresu]

I they could design an AIO too, with an array of these AeroJets covering the radiator.

Even if they scaled them to twice the length for 20 W per device it wouldn't be ideal at all.

Better to use a fancy flat heat pipe/vapor chamer configuration coming off the card itself to fix the copper base plates of the AirJet's to, and then scale it up to as many as you can fit without cutting off air flow or making the card impossibly heavy from all the heat pipe copper metal.

Say you can manage 4 heat pipe/vapor chambers on each side of the card - with each heat pipe/vapor chamber having an Airjet on both sides.

Even assuming that hypothetical 20W double length device works you still only have 160W worth of total cooling.

I think if GCD/SED chiplets work for >RDNA4 then at 160W you could cram more logic in at a much saner clockspeed for power efficiency - though I wouldn't care to speculate on the astronomical cost of such a system 😅🤑

 

[aigomorla]

Nah.. scale them large... spam them under solar panel wings.
Now we got super efficient means of flight.

Principles of Lift should be achieved if it does anything it claims.

 

[FlameTail]

I am more excited that one day we may see this in smartphones!

Smartphone chips are becoming more powerful and more power hungry, making them harder to cool.

I imagine gaming phones would be the first to make use of this.

 

[igor_kavinski]

I imagine gaming phones would be the first to make use of this.

And portable VR headsets. Handheld gaming consoles. Anything that generates a lot of heat in a small package.

 

[soresu]

Principles of Lift should be achieved if it does anything it claims.

The problem with lift is that speed of airflow to break the boundary layer was the aim of the design rather than volume of airflow - the latter being as important to lift as air speed 🤔

I'm not sure if making them deeper to add more MEMS membranes vertically would be possible to increase volume.

That being said it could still be an interesting if they shifted the intakes to the sides instead, allowing for stacked multiple layers of MEMS + copper vapor chambers per device.

Unless they have purposefully engineered the 1st gen MEMS to have a lot lower perf than it could be given the techs potential it seems like stacking would be the only viable solution for scaling AirJet to higher thermal density use cases like desktop CPU and GPU.