News Fanless AirJet Mini G2 cooler promises 42% higher performance at the same form-factor

[Admin]

Frore Systems launched its second-generation AirJet Mini solid-state cooling chip at Computex, delivering 50% greater heat dissipation in the same compact form factor.

Fanless AirJet Mini G2 cooler promises 42% higher performance at the same form-factor : Read more

 

[thestryker]

I'm not sure their efficiency is quite there for battery powered devices yet, but they're getting close. Since we're stuck with M.2 for desktop systems I could see one of these being about perfect now though. While it wouldn't be enough for gen1 PCIe 5.0 controllers it's plenty for the new node shrunk ones. If a cooler could be designed so the airjet exhaust went over fins for cooling NAND the airjet wouldn't have to be connected to it.

Looking forward to seeing them keep advancing the technology because it seems so good for low noise/industrial type application.

 

[bit_user]

I'm not sure their efficiency is quite there for battery powered devices yet, but they're getting close.

I disagree. First, you should consider that existing solutions already involve some kind of active cooling (i.e. which use power). Second, it only needs to run at max speed when the SSD is under high load, which should be a minority of the time.

Since we're stuck with M.2 for desktop systems I could see one of these being about perfect now though.

I think it would add too much cost for most SSDs. Also, you'd only need it in cases like where a SSD was stuffed under a dGPU. In most M.2 slots, there's enough headroom for a bulky heatsink that's adequate when the case has some reasonable airflow.

If a cooler could be designed so the airjet exhaust went over fins for cooling NAND the airjet wouldn't have to be connected to it.

But you might lose efficiency, that way. Also, the NAND is what's most temperature-sensitive. IMO, it would make more sense to attach the cooler to the NAND and have its exhaust flow through a heatsink attached to the controller.

 

[thestryker]

But you might lose efficiency, that way. Also, the NAND is what's most temperature-sensitive. IMO, it would make more sense to attach the cooler to the NAND and have its exhaust flow through a heatsink attached to the controller.

When the controller makes up the vast majority of the power consumption it really doesn't make sense to attach it to the NAND. If you were only worried about the NAND there would be no point in using one of these at all.

I think it would add too much cost for most SSDs. Also, you'd only need it in cases like where a SSD was stuffed under a dGPU. In most M.2 slots, there's enough headroom for a bulky heatsink that's adequate when the case has some reasonable airflow.

While I certainly agree about the cost the drives most likely to need cooling (PCIe 5.0) are also the ones that are usually going to be interfered with by modern video cards. My video card is small at ~2.5 slots (comparatively speaking) and it still covers up two M.2 slots and the top one is close enough a large heatsink might not fit due to the back plate.

So again while it wouldn't make financial sense it would be perfect for the job. I wouldn't be surprised if the only place we saw them with SSDs is expensive external enclosures.

I disagree. First, you should consider that existing solutions already involve some kind of active cooling (i.e. which use power). Second, it only needs to run at max speed when the SSD is under high load, which should be a minority of the time.

Nobody actively cools SSDs in battery powered devices. I was referring to replacing fans with these. Though that does bring up a good point as fans do have a small amount of ancillary airflow which these do not which means they don't help cool anything not attached to the heatsink.

Anyways back to the point: handhelds tend to use fans that max out at 2.5W (typically I'd expect less than 100% fan speed, but that depends entirely on the cooling system). Steam Deck is up to 15W (1 fan) and ROG Ally up to 30W (2 fans) and keep in mind this doesn't include memory/platform power. To replace the fans and keep the same maximum TDP it would take 3 (3W) or 5 (5W) of the new airjets (could maybe get away with 2/4, but I'm not certain if the airjet ratings are hard limits as the only retail compute product using first gen was TDP locked below maximum airjet rating). While I think the heatsink design could be less complex I doubt there would be any material savings since they all have to be connected.

I would be surprised if a single airjet was anywhere near as cheap as a fan, twice as many are needed minimum and wouldn't gain anything power efficiency or performance wise. The gains by using them would be purely customer focused as the hardware itself would probably run less hot and be less noisy. I would certainly pay more for that, but I'm not sure how many people would for a mass market type product.

 

[bit_user]

When the controller makes up the vast majority of the power consumption it really doesn't make sense to attach it to the NAND. If you were only worried about the NAND there would be no point in using one of these at all.

The NAND is the most temperature-sensitive component on a SSD. A controller can probably tolerate close to 100 C, while NAND (IIRC) shouldn't get above ~70 C.

It's like why we don't tend to have active cooling of MOSFETs on motherboards, even though they're easily the hottest thing on the board. You don't focus on cooling the hottest thing, you focus on cooling the thing that's most temperature-sensitive.

the drives most likely to need cooling (PCIe 5.0) are also the ones that are usually going to be interfered with by modern video cards. My video card is small at ~2.5 slots (comparatively speaking) and it still covers up two M.2 slots and the top one is close enough a large heatsink might not fit due to the back plate.

But how many of the occluded slots are PCIe 5.0-capable?

 

[thestryker]

The NAND is the most temperature-sensitive component on a SSD. A controller can probably tolerate close to 100 C, while NAND (IIRC) shouldn't get above ~70 C.

Definitely above that for NAND, but I have no idea what the threshold really is since every SSD manufacturer seems to have arbitrary limits. A lot of SSD manufacturers don't expose multiple temperature sensors, but based on those that do a maximum in the low-mid 80s seems likely.

It's like why we don't tend to have active cooling of MOSFETs on motherboards, even though they're easily the hottest thing on the board. You don't focus on cooling the hottest thing, you focus on cooling the thing that's most temperature-sensitive.

It's not about what runs hottest it's about what outputs the most heat. If the controller is making up the majority of the power consumption that's the important part to isolate. The NAND isn't what's soaking SSD heatsinks it's the controller (on high performance drives). I would love to see the split on the newest drives, but Samsung's sensors apparently both read the same and I haven't seen numbers on the new SanDisk (working on the assumption that if the temperatures are close to the same with a heatsink that the power consumption is as well).

But how many of the occluded slots are PCIe 5.0-capable?

It's never going to be more than 2 on client systems, and on the workstation side varies sometimes significantly by board design.