Hardware Insights Forum

Here's my take on it:

• For a fan to be considered decent it must last at full speed constantly for at least 30,000 hours at 40°C. Of course I'd prefer quite a bit more - upwards of 50,000 hours. The lifespan of fans, however, is specified not as the average time to failure, but the time after which only 10% of fans will have failed, so most quality fans will go further than the spec says.
• If there's no failure detector and it's somewhere failure could lead to serious devastation (e.g. PSUs), better safe than sorry.
• I do not consider temperature control an excuse to risk it with a low quality fan.
• Fans with one ball bearing and a sleeve bearing are just silly, providing no accountable benefit to reliability compared to having just a sleeve bearing, or for noise or shock resistance compared to dual ball bearing, and by design, they don't load the ball bearing properly. The highest-quality brands don't even make them.
• A fan in horizontal orientation should either use dual ball bearings or an advanced bearing design with axial support (ARX CeraDyna for example does not provide that).
• An "enhanced" (as opposed to conventional) bearing design may provide a longevity benefit, but only if it's from a good manufacturer.
• Tiny high-speed fans seem most prone to dropping like flies, so be very careful with them.

Of course, reliability is a huge deal to me, and should take precedence when desirable. Some other people may not be so demanding.

As for my manufacturer preferences, for conventional sleeve bearings, it would probably be something like this:

1. Nidec
2. Delta
3. Power Logic
4. AVC

But it's not easy to rate them in an exact order. For the rest, I'm looking at something like Globe Fan, Yate Loon, Dynatron/Dynaeon, Superred, ADDA, and finally the bottom-of-the-barrel brands like Ruilian Science and T&T (which I would not use under any circumstances). As for ball bearings:

1. Sanyo Denki
2. Papst
3. Nidec
4. NMB
5. Delta
6. Power Logic
7. AVC

But again, it's not easy. For advanced bearings, it's too hard to say.

For what it's worth, I've had a few NMB 3110SB (they also have a 2410SB and 3610SB; for other sizes you have to use different series) ball bearing fans for maybe a year. First impressions are good - well-made plastic, sculpted blades with shaped outer edges and even balanced rotors. They are specified as being somewhat quieter (about 2dB less) than comparable fans of the same speed, and that seems real enough - though I can't back up their claim that they move more air than their peers (but then again, I don't have actual test equipment). They are also stated to save power, by using a single motor winding driven bidirectionally rather than two opposing windings being alternately unidirectionally driven. The bearings, however, are seemingly noisier than some of the used ball bearing fans I have from lesser manufacturers. (This is for the B40 model - the two B50s I have seem to have better bearings. Seems they're saving the best bearings for the fastest fans, so if you want the least bearing noise, you'll have to take the fastest model in the series and slow it down. The B50 is too loud for constant full-speed operation, so those went into PSUs with temperature controllers.) The one B40 currently in my PC (I had two installed for a while, since that was what my case supports, but eventually decided one was enough and this is my current primary use of the second) hasn't become noticeably louder.

For a fan to be considered decent it must last at full speed constantly for at least 30,000 hours at 40°C. Of course I'd prefer quite a bit more - upwards of 50,000 hours. The lifespan of fans, however, is specified not as the average time to failure, but the time after which only 10% of fans will have failed, so most quality fans will go further than the spec says.

Many sleeve bearing fans won't achieve anything like this because they're not oiled well enough. The biggest problem with most sleeve bearing fans is that they aren't oiled well from the factory so their lubricant doesn't last long. Once well oiled I've found they can last as long as dual ball bearing fans, but with better shock resistance. Plain old sleeve bearing fans do have problems being mounted horizontally, though... and unfortunately, like Taiwanese and Chinese manufactures of capacitors, a big problem with fan brands (especially sleeve bearing) is that they are not consistent with their quality... only the best brands seem to be.

Tiny high-speed fans seem most prone to dropping like flies, so be very careful with them.

Especially the 10mm thin 40mm ones that may qualify for "high RPM". I don't recommend running them at 2000RPM or above.

The only issue I've had with ADDA fans is that they aren't lubricated well. Otherwise, they are very quiet for the amount of air they push (except at full speed) and are easy to maintenance.

If there's no failure detector and it's somewhere failure could lead to serious devastation (e.g. PSUs), better safe than sorry.

Usually a sleeve bearing fan will give audible indication before it fails, or will start spinning slower, so that's indication enough for me.

I'll hapilly use a sleeve bearing fan if it has proper lubrication in the bearings

Wester547 wrote:Many sleeve bearing fans won't achieve anything like this because they're not oiled well enough. The biggest problem with most sleeve bearing fans is that they aren't oiled well from the factory so their lubricant doesn't last long.

Yes, that's exactly it.

Plain old sleeve bearing fans do have problems being mounted horizontally, though...

The motor magnets, at least in smaller fans, fight the gravity acting on the impeller and most sleeve bearing fans have small rubber rings to keep the rotor/retainer from making hard contact with the ends of the sleeve so some of them can hold up for quite a while even when mounted that way. I should know - remember my bunch of MPT-401s??? They had 120mm transparent sleeve bearing fans from Power Logic. The computers with them were used for a fair while, the case fans (also Power Logic sleeve bearing types, but black) ran at ludicrously low fixed speeds, the secondary capacitors were devastated (not entirely the fault of the bad case cooling, though - Macron kept using GP caps where they should have used low ESR), but the fans are fine to date. I still swapped them - and the resonance of the transparent plastic plays a significant part in that. I have more of those units waiting - I've considered trying out a few different grades of secondary capacitors (e.g. Chemi-con LXZ/Nichicon PW/Panasonic FC, Chemi-con KY/Nichicon HE (with Rubycon YXJ for the small 220µF 16V caps on the negative rails), Nichicon PA, Chemi-con KZE/Rubycon ZL, Panasonic FR/Rubycon ZLH, Panasonic FM (in descending order of ESR) (but the difference between the latter three groups is minimal) on one, meanwhile testing it under load (building a load tester - that's gonna be fun - and I intend on making it as quiet as practical, given 120mm cooling fans - the good news is that the aluminium-bodied power resistors used can withstand a ton of heat when de-rated. And yes, I know about temperature coefficient - but with the specs given (for Stackpole KAL) and the values used, it will result in a variation of just 1.25% between 25°C and the limit (275°C). Carbon resistors do have a substantial TCR but it's negative, not positive, so with them you'd draw more power when hot) and measuring the ripple results, to see which does best (and that I'd recommend from there on).

Tiny high-speed fans seem most prone to dropping like flies, so be very careful with them.

Especially the 10mm thin 40mm ones that may qualify for "high RPM". I don't recommend running them at 2000RPM or above.

Well, if you run such a fan at that speed you're going to get sod-all airflow, and no real static pressure capacity at all. To me, the sensible thing in that case would be to swap it for a good 2BB fan.

Usually a sleeve bearing fan will give audible indication before it fails, or will start spinning slower, so that's indication enough for me.

Maybe not if it runs really slowly. Hipro HP-D3057F3H does that when lightly loaded (it does have massive heatsinks and excellent ventilation, though). Considering how crappy new HDDs are...

(WD800JD-00LSA0s, the drives used in the mentioned PCs, do seem pretty bulletproof. They seem just fine (no bad sectors and maybe one CRC error) despite the capacitor destruction in the PSUs, although to be fair, the motherboards (Intel D945GTP - which I otherwise dislike, though) had high quality capacitors and HDDs also usually have ceramic capacitors, which have excellent high frequency response and, being multilayered, are actually quite compact for their capacitance (current X7R types can make electrolytics look bulky, especially as much of the volume of tiny electrolytics is used by the bung. MLCCs have just one minor drawback in comparison - cost). Two of them run my system, in fact, and have done so for over a year. I take it that it was the older ball bearing WDs that were dodgy.)

As an aside, dust accumulation (if, as in the mentioned PCs, you don't have a filter) will ruin the "good looks" of transparent fans. Not that there's a point in them, anyway.

The motor magnets, at least in smaller fans, fight the gravity acting on the impeller and most sleeve bearing fans have small rubber rings to keep the rotor/retainer from making hard contact with the ends of the sleeve so some of them can hold up for quite a while even when mounted that way.

The oil in regular sleeve bearing fans dissipates much faster when they're mounted that way. However, a very well lubricated sleeve bearing fan might handle it just fine.

Maybe not if it runs really slowly. Hipro HP-D3057F3H does that when lightly loaded (it does have massive heatsinks and excellent ventilation, though). Considering how crappy new HDDs are...

How slow do you reckon the fan is spinning at the lowest load? The Hipros I've seen probably don't have their fan spinning slower than 1370RPM at the lowest load (assuming the fans they use are capable of up to 3000RPM and are 80x80x25 in dimensions with 7 blades). I agree about the ventilation, though I think it could stand to have a better fan grille (c_hegge doesn't hesitate to swap it for a wired grille, to my knowledge, which is a good thing). I don't like the fact that they use 16A ultrafasts in parallel for the +12V rail, especially in a half-wave rectified output, but them's the breaks.

(WD800JD-00LSA0s, the drives used in the mentioned PCs, do seem pretty bulletproof. They seem just fine (no bad sectors and maybe one CRC error)

Sometimes Western Digital drives last, sometimes they don't. That can be said of all "decent" manufactures. Truly crappy manufactures (I.E Toshiba) are guaranteed to fail, though.

As an aside, dust accumulation (if, as in the mentioned PCs, you don't have a filter) will ruin the "good looks" of transparent fans. Not that there's a point in them, anyway.

Sure there is, you get a better look into the power supply (transparent blades aren't blocking your view the way the usual blades do) without having to open it up.

Wester547 wrote:How slow do you reckon the fan is spinning at the lowest load?

So slow that you can barely feel the airflow.

Sometimes Western Digital drives last, sometimes they don't. That can be said of all "decent" manufactures though.

I never said that any HDD was completely fail-safe and indestructible, and as for the analogy, a literal "bulletproof" vest itself isn't completely bulletproof - a powerful enough bullet won't break a sweat.

For curiosity, StorageReview's database gives these percentile ratings for two similar(-ish) models:
WD2500KS - 64th
WD3200Jx - 58th
And that's it. The other ≤250GB models are ignored as while there were FDB models down to 40GB, the database doesn't distinguish between the different revisions of each "model". Those are better ratings than Seagate 7200.8 and 7200.9 have (49th and 43rd respectively), though.

But it's still a user survey.

momaka said that these and related WDs are even more reliable than the Seagate 7200.7. It would seem tough to beat the earlier Seagates but I don't think it's an unreasonable proposition - especially if cooling is limited, as if this article is anything to go by, the WDs don't drain monstrous +5V currents during sequential reading like most drives of the era did (7200.7 actually used more +5V power doing that than it takes from +12V to run the motor, which would explain why that microcontroller/read channel chip gets hot so easily (without the metal base I'm sure they would have self-destructed), and those damn thermal fluctuations when I had my ST380011A in that USB case - while it was at least an aluminium case, I have to admit, I did use it horizontally...).

Sure there is, you get a better look into the power supply (transparent blades aren't blocking your view the way the usual blades do) without having to open it up.

Well, I just took a look at one of my unrepaired MPT-401s and could see some of the busted caps, but not because of the transparent fan - it was too dusty to see through. I was looking through the gaps between the blades. (That's with a wire grille - with punched grilles, you'll get a poorer view.)

LongRunner wrote:So slow that you can barely feel the airflow.

At what load does the fan start kicking up speed?

(7200.7 actually used more +5V power doing that than it takes from +12V to run the motor, which would explain why that microcontroller/read channel chip gets hot so easily (without the metal base I'm sure they would have self-destructed), and those damn thermal fluctuations when I had my ST380011A in that USB case - while it was at least an aluminium case, I have to admit, I did use it horizontally...).

That's strange. My ST3120026A and ST3160023A's PCB never runs hot, even in hot weather, even the microcontroller never runs hot (one is an Agere, the other is a STMicroelectronics one from Malaysia). It would explain why you found some STMicroelectronics microcontrollers go bad, though. At least those units have QFP microcontrollers and not BGA.

Wester547 wrote:At what load does the fan start kicking up speed?

No idea. I don't run many intensive tasks on my PC.

That's strange. My ST3120026A and ST3160023A's PCB never runs hot, even in warm weather, even the microcontroller never runs hot (one is an Agere, the other is a STMicroelectronics one from Malaysia).

It gets the entire ground plane to spread heat through so with enough airflow, it may stay only moderately warm.

It would explain why you found some STMicroelectronics microcontrollers go bad, though.

So to sum it up...

• Six watts dissipation in a single small(-ish) chip...
• ...which only has its metal base soldered to the ground plane of the PCB for cooling...
• ...and the base in question is smaller than that of the motor driver, when it should have been larger...
• ...and some weak batches of the chip were produced. Sounds risky, doesn't it???

Or maybe ST pushes their components to the limit while other manufacturers are more conservative with their designs...?

LATE EDIT: It's not all dissipated in the microcontroller, as non-NCQ 7200.7 and earlier used linear regulators; Vcore is +1.5V so the controller dissipated 1.8W tops.

And that document you linked to before (on Badcaps.net) in reference to the platter coating thing says this:

In [Alpine/APLUS], the following malfunction is most frequent: failure of microcontroller portion responsible for reading and writing operations (read-write channel).

Even if it doesn't say anything about heat or the ST ones being worse, that's very telling. The codename isn't an accurate description, is it .

I'm curious: If they made that chip with a metal pad on top as well as the base on the underside, and with a thicker aluminium version of the SeaShield used as a heatsink, would the reliability of the electronics be improved???

At least those units have QFP microcontrollers and not BGA.

Which I find interesting now because in combination with the high heat, they used SnCu (I think) solder which is the worst type of solder available, and they still don't have widespread troubles with fracturing (that I'm aware of) even over a decade later.

I had to retype the whole sentence myself because the PDF is copy-protected.

I've found a failing Globe Fan. It's an S01138812M, which I think is supposed to run at 2500RPM, but I can only get it to full speed (well, a little higher) with 24V through it (which results in a ridiculous vibration). The bearing is a bit "sticky". It came from an HEC-145FB, which admittedly has no fan controller.

So it runs at ~1300RPM @ +12V? That's lame. Does it start spinning automatically when you apply +12V or lower to it? I've had some sleeve bearing fans fail so epicly that they won't actually spin (for the lack of lubricant) until you rotate the fan manually before applying a voltage to it...