What's Worse for Hard Drives: Heat or Vibration?

gottabeme asks: "I turned on my computer the other day and all of a sudden the BIOS said the S.M.A.R.T. status was "Bad: backup and replace." The drive has continued working in PIO mode (instead of DMA) long enough for me to get a new drive and copy everything over. When I finished copying and put the new drive in the cage where the old one was, I realized that the fan at the front of the cage which was keeping the drive cool to the touch was causing a fair amount of vibration to be transferred to the hard drive. The other 7200rpm drive without a fan was pretty warm, but had no vibration at all. The bad drive is only a few years old, and I've never had a drive fail on me in around 10 years of computer use, until now. And until I got this case and drive I'd never had a fan blowing on a drive before. Who knows what caused the problem, but all this has made me wonder: Which is worse for a hard drive? Heat that's fairly warm to the touch, or constant vibration from a case fan right next to it? Any readers care to offer their experiences and knowledge?"

Comments

Vibration is probably the worst

[stevew]

Vibration is probably the worst enemy to the drive since it can send the head crashing into the drive surface. Modern drives have a pretty high shock rating, but this is substantialy reduced if they're operating. Even then they are considerably better than they were even 5 years ago.

That being said - head is more an issue for the drive electronics than it would be for the physical drive.

Summary - drives have moving parts - they wear out for lots of reasons. Vibration and heat should be avoided to prolong their service life.

Vibration

[TheCrimsonUnbeliever]

Vibration is the killer - knocking heads about all the time - Heat is something that can be bad - but only in extremes (70 C+)

Move the fan - Or screw it in better to kill the vibration

Re:Vibration

[Keith_Beef]

nice bright yellow polyurethane Silentblocs?

Silly question

[MerlynEmrys67]

What is stopping you from replacing the fan, replacing it with a low vibration replacement and then you have low heat, AND low vibration.

The other thing is newer drives seem to be quite a bit more prone to failure than drives even 5 years ago, don't know if that is because of cost reduction, or higher speeds

My experience has been heat

[Zeio]

My experience has been that heat causes more problems over the long haul. Also, any time a hard drive is at an angle from level that isn't 0 or 90 degrees that is very bad. As far as vibration goes, I usually make an effort to fasten the drive firmly to the case (use all 4 screws), so like a seatbelt, this would prevent the hard drive from vibrating much unless the whole case it vibrating.

The new Cheetah 15.3 drives are double the density per platter, faster, give off less noise and dissipate less heat then previous generations. Less heat dissipation is the most impressive attribute moving forward. Any time you do see fast server drives implemented by vendors or in storage cabinets you notice the ventilation is superior, and that they suggest operating them in environments under 80 degrees F. (I prefer 72, low humidity).

The asics and electronics on the drive probably like cold temperatures rather than low vibration, and the speed of the platter's rotation created a gyroscopic effect meaning you would have to jar the drive well beyond the specified maximum (hard drive manuals list a maximum G shock while in operation). If you are vibrating the dive out of the specified limits, most likely a conservative figure, you are essentially intentionally trying to damage the disk.

Don't worry about either

[KevinIsOwn]

The problem probably was niether vibration nor heat. Harddrives are made so that they can withstand high g-forces. Some review sites have said you could throw a harddrive against a wall while it is running and it would be fine. How true this is, I don't know, but I know fan vibrations are no where near large enough to cause a problem. While continued fan virbrations theoretically could be bad, fans don't vibrate all that much. I've used fans missing fins (don't ask!) that virbrate like crazy and never had problems (Not for long though, I generally replace those with non-broken fans)

Heat is one of those things computer geeks fear most. We all want to get it as low as possible. Well let me tell you a little something about harddrive heat: Unless you have a drive spinning at 10,000 rpm or higher, you really have nothing to worry about. If HDD's weren't meant to withstand a bit of heat then you would be hearing about a lot of unhappy customers. My hdd's are warm to the touch, but that is fine, they are well within the limits. Now if it burns your finger when you touch it, then you are probably going to be having problems with all the other components in your computer as well.

Harddrives die. And they die often. If you haven't had one die in a long time, then you have been very lucky. I've had 3 drives die on me in the last 2 years. Granted 2 were IBM Deathstars, the third was a different brand. They weren't all that hot and they did not have any fans vibrating near them too much either. They just die, HDDs are not as reliable as many of us would like. (Can't wait for solid state hdd's :) )

If you decide that everything I have just said is crap and want to take the paranoid way out, that's fine! You know what they say: Better safe than sorry!

So here's what you should do:
1. Get some grommets for that fan. They will reduce fan vibrations to practically nothing. They'll also make the fan quieter too! You can pick these up from PCMods: http://www.pcmods.com/details.asp?ProdID=20
2. Get a HDD cooler. They will cool your hdd a lot more than a fan that's blowing air over it will. While I'm at the pcmods site, I might as well link there. If you shop around you will probably find better prices. Lower end cooling solution: http://www.pcmods.com/details.asp?ProdID=46 Higher end: http://www.pcmods.com/details.asp?ProdID=452 (Even has an LCD!)

I just want to stress this again: You don't NEED these two products unless you have an ultra-fast SCSI hdd. Your hdd should be well within its limits with some small vibrations and a bit of heat. But if you want to spend some money, I'm not about to stop you!

Re:Don't worry about either

[twilightzero]

*Insert standard disclaimer: I work for Western Digital tech support*

Gee what's your name? I'll be sure to put a note in your case file in the WD call tracking system that you intentionally damage your drives and should never receive an RMA again ;) That point aside, an error code is NOT at all required to RMA a drive. Quite often we get people calling with drives that won't run diags at all, in which case it's ridiculous to require a code. We WILL encourage ppl to run diags if we think it may be a problem not with the drive itself. A classic scenario is someone receiving bad sector warnings from Scandisk. This definitely does not mean the drive has bad sectors, only that scandisk can't read the data at those points (most likely the result of corrupt data). Besides, you can always do an RMA through the web site, which doesn't require any type of code - there's a field for it but you can just leave it blank.

Now on to the main topic, the problem you had here was definitely heat. Only an electronics failure would cause the drive to run in PIO instead of DMA like that - if it was a physical failure with the platters or somesuch caused by vibration, the drive would run at DMA but you'd get no data. Possibly the asic that controls data transfer got a bit fried and was no longer able to signal at top speed. Instead of quitting outright, the drives are all backward compatible all the way down to PIO1 and such and are programmed to drop their transfer speeds down if they encounter a problem at the higher speeds. So logically, (if...she...weighs the same as a duck...wait wrong explanation...) this particular one was probably a heat problem.

Now, let me touch on a few more I saw floating around in this thread:

1) No, do NOT throw a HD against the wall while running, your data will be gone faster than you can hear the clunk.

2) It doesn't matter what orientation you have the drive mounted at, just so long as it's mounted SECURELY (read: 4 screws and to a metal enclosure that's grounded).

3) Vibration is BADBADBAD for a drive. While it may not cause outright failure immediately, it will cause a huge number of misreads and retries on the drive, thereby slowing down the overall performance of the drive. This is one of the things that's driving the overall industry move to fluid dynamic bearings (FDB) - they cause much less vibration and therefore contribute to the logevity of the drive. And if the vibration is getting bad, it WILL cause the heads to touch the platters momentarily. Now the drives are designed with this in mind and have an extremely thin (several atoms) layer of lubrication on the platters. But don't encourage it if you value your data.

Any other questions you want answered, drop me an email :)

Re:Don't worry about either

[The Pim]

but I see several a month where scandisk reports bad sectors, but if you back up the data and low level format it (read: write zeroes, I'm not going to go into an explanation of why you can't do a real low level format on an IDE drive) no errors are found and when the drive is repartitioned/formatted, scandisk is happy as a clam.

What happens here is actually really simple. When the drive notices a dead or dying block, it can use one of its "reserved" blocks instead, and enter the substitution into a table somewhere. Except in one case: When asked to read from a dead block, it can't just return bogus data: it has to admit that there was an error. When asked to write to a dead block, on the other hand, it can do the remapping, write to the reserved block, and nobody will ever know the difference.

It's rather counterintuitive until you think it through (or better yet, get a bad block and try it!). My intuition was that writing is "harder" on a flaky disk than reading, when in fact it's much easier for the disk to cope with a problem writing. Similarly, a problem reading is much less likely to signal a dying disk than a problem writing. A problem reading just means that there is one bad block somewhere on the disk; a problem writing probably means that all of the reserved blocks are used up (having been mapped to other bad blocks), so there's nowhere left to write.

Re:RAID.

[Hex4def6]

Raid 0 is not real Raid. That doubles the speed of data transfer, by using 2 HDD, and pretending its one. You probably need Raid 0+1, which has speed benefits, and redundancy. It means of course 4 HDD, all identical...

Re:My take

[unitron]

There is no inert gas inside a hard drive, it's just plain ol' air, albeit extremely clean air. Hard drives have a vent (which is a filter with extremely small holes) which allows the air pressure inside and outside of the drive to remain equalized.

SpinRite Only Does FAT(32)

[gottabeme]

Great, if you use FAT or FAT32. If you use NTFS, Linux FS's, etc, SpinRite won't work. You'd have to move all your data off the drive or partition in order to use SpinRite. $90 for a program that won't do anything but FAT seems a bit much nowadays.

Re:Both are bad but not really.

[chunkwhite86]

Turning the system on and off all the time is bad as well.

Well, your half right. IDE drives are designed for MANY power on/off cycles (as the typical home or office PC is turned on and off at least once a day) but not for continuous operation. SCSI drives on the other hand are designed for continuous operation, and NOT for many power on/off cycles.

I had a Seagate 15K rpm Cheetah that recently took a dump. No special filtering, no real cooling scheme to speak off. It ran for 4 years almost to the day. Then during a load of windows, it said bye-bye. Found out the motor crapped out.

I'm willing to bet it's because you turned your PC on and off each day, and you didn't have adequate cooling. The original 15k cheetahs were DAMN hot and without active cooling, they ran above the manufacturer's operating spec. Remember, SCSI drives are designed to be in Servers where noise is not an issue, and they expect to be cooled with fans! And before some AC jumps in with a "your full of crap!" post, I'm a data storage engineer for a large storage vendor so I know what I'm talking about.

Your 15k cheetah BTW should have a 5 year warranty. Not sure if that was voided because you probably overheated it or exceded it's rated number of power on/off cycles.

Personal Experience

[JWSmythe]

Well, here's what I've learned from personal experience. I've bought no less than say 300 drives over the last few years, for servers and workstations.

We'll start with what lives longest.. Machines that we have in our colo's, kept under 75 degrees F, and they are very rarely moved. Some of the machines don't have physical interaction for over a year at a time. These live virtually forever.. We've had less than 1% failure rate over 5 years. We've retired more, simply because they're no longer big enough for our purposes, rather than because they've failed.

We recently shipped 20 of those old drives from New York to Los Angeles, via FedEx. They were all working at the time they were shut off, and packed in a shipping crate. I've only tested 4 so far. 2 were completely dead.. One wouldn't spin up. The other spun, but "knock"ed, and was completely worthless. The other two worked fine. So, the physical abuse of just being shipped was enough to kill them.

Now, consider the drives that haven't been in nice colo conditions. Some have been in offices where the staff seems to think 80 degrees is cold. At 80 degrees F, we have something like a 25% failure rate over 1 year. 25% of the machines will have a drive failure in a year. I can only name off two machines in that environment that haven't had a drive failure in 5 years, one of them being an extraordernarly cooled case (6 case fans, plus 2 small fans on each drive).

In one environment, the staff insisted on keeping the temp at 90 F.. This was mostly because they knew the machines would fail at about 90 F, and they didn't have to work if their workstations crashed. Funny, that business went bankrupt.. Besides over a 30% drive failure rate, they also managed to cook the rest of the parts rather randomly. Motherboards would simply stop working, power supplies would get toasted, and CPU's with good CPU fans would just drop dead.

In a computer store I worked in, when Quantum had first released their "BigFoot" drive series (5.25" wide, and maybe .5" thick), they were shipped to us as OEM parts in large boxes. There was formed sponge foam around them to keep them seperated. The shipping department would receive them, and pull them out two at a time. If we heard them "clack" into each other (it's a distinctive sound), we could pretty much guarantee the drives were both dead. Poor Quantum, they got so many RMA's from that store on those. I think we only had a 50% survival rate. They were nice drives though, in the fact that you could stack 4 of them in two 5.25" bays.

There are always the rare exceptions that are always quoted as fact. One guy would tell everyone about how he has a machine with a SCSI drive running for 10 years, with no fans at all.. Ok, but it's not very good statistical sampling. A sample group of one doesn't show much.. Over hundreds, we get a better picture.

So, yes, keep your drives cool.. If you don't, it will have a shorter life span.

Don't shake your drives.. Hard impacts (less than 2" drop is enough) can destroy it, either damaging the controller board, or bumping the heads into the platters. The space is rather small (see the discussions a few months ago about removing the tops of hard drives. Smaller than a piece of dust). Constant vibration can have the same effect as a good impact. Harmonics can be evil.. Just ask any aircraft engineer.

Re:My take

[plsander]

It is the lower of density in the atmosphere.

Remember, HD heads "fly" above the platter -- if the air pressure (density) is too low, the head does not produce enough lift and will crash.

Most HD specs will list an altitude or pressure range where operation of the drive is supported.