Hard Drive Temperatures: Be Afraid

December 17, 2006

I recently had a noisy fan failure in my ASUS Vento 3600 case. The particular fan that failed was the 80mm fan in the front panel, which is responsible for circulating air by the hard drives in the front of the case. I disconnected it while I considered my options. There's not a lot of airflow by the hard drives in this case. I've actually had a hard drive failure in this system, which I strongly suspect was due to leaving the front fan disconnected.

The two hard drives are mounted with rubber grommets to reduce conducted vibration noise, a standard feature of many new PC cases.

hard-drive-grommets.jpg

Avoiding direct metal-to-metal contact will always help quiet drives-- they are, after all, giant hunks of metal spinning at 7,200 or 10,000 RPM. But the lack of metal-to-metal contact also means the drives don't benefit from the significant auxiliary cooling effects of metal contact.

Of course, hard drives don't generate nearly as much heat as your CPU and video card do. They only consume around 10 or 12 watts under load, and around 7 watts at idle. But unlike your CPU, they're generating a lot of mechanical movement, which means friction-- and heat disproportionate to the power input. They still need some airflow to stay at a reasonable temperature.

I often read about users obsessing over their CPU or GPU temperatures, while ignoring their hard drive temperatures entirely. That's a shame, because the hard drive is the most temperature sensitive device inside your computer. Most manufacturers rate CPUs up to 70C, and GPUs commonly rate to 90C and beyond.

Manufacturers measure off quite a modest range of operating temperatures for hard drives, from +5 to +55C as a rule, and occasionally to +60C. This operating range is much lower than processors, video cards, or chipsets. Moreover, hard drive reliability depends heavily on their operating temperatures. According to our research, increasing HDD temperature by 5C has the same effect on reliability as switching from 10% to 100% HDD workload. Each one-degree drop of HDD temperature is equivalent to a 10% increase of HDD service life.

Hard drives are only rated to 55C in most cases. Although there's still a lot of ongoing discussion on what exactly a "safe" temperature is for a hard drive, the general consensus is that high temperatures are much more risky for the hard drive than any other component inside your computer.

When your CPU, video card, or motherboard fails, you buy a new one and replace it. Big deal. Life goes on. But when your hard drive fails, unless you have a rigorous backup regime, you just lost all your data. Failure of a hard drive tends to have catastrophic consequences for your data. That's why I'm always very careful with hard drive temperatures. When I disconnected the failing fan, I used the excellent DTemp hard-drive temperature monitoring utility to keep an eye on the temperatures.

dtemp-screenshot.png

Sure enough, with the front fan disconnected, both drives inched up to 46C in 15 minutes. And that was at idle. I can only imagine what the temperatures would look like after internal temperatures increased under load. I've already had one drive failure in this case with sustained temperatures around the same level. Some kind of replacement airflow is essential. I used foam tape to mount an 80mm fan on the front of the drives, blowing across the drives and back towards the case. As I write this, they're down to 33C -- a whopping 13 degree drop.

Hard drive temperature is arguably the most important temperature to monitor in your computer. If you regularly see temperatures of 45C or higher on your drive, consider improving airflow in your case. If you don't, you've substantially increased your risk of hard drive failure or data loss.

Posted by Jeff Atwood
124 Comments

I read couple of case cooling articles where air flow must be from front to back, etc. Well I applied them to my Computer and I experienced significant loss of performance.
I removed all the additional fans and my system is back to its original kick ass speed.

Check your power supply, the additional fans could have put it over its power load limit?

When ever adding additional fans make sure you have them blowing in the same direction, otherwise you may make things worse.
Most power supply fans blow out, so if you place a fan on the front of the case it should be blowing into the case. It is rare but some old power supply fans do blow into the case.

For keeping hard drives cool I have found passive coolers to be effective and of course silent.

alex kid on December 8, 2008 7:45 AM

Chris - In regards to your problem, there is not much you can do, the drive can not be cooled below room temp. Well unless you try setting up a peltier system, but it would give you condensation problems.

To be honest the high humidity may prove more of a problem for your drive :(

If possible try avoiding running you drive for long periods, back up to optical discs and if your lucky find a building with refrigerative a/c.

alex kid on December 8, 2008 8:10 AM

Jeff,
If the input power is 7 watts; there can be no more than 7 watts of heat generated. Conservation of energy still applies even when mechanical parts are involved. Friction does not magically create anything, it converts mechanical energy to heat energy.

Honesuki on December 27, 2008 9:49 AM

According to SMART the two 500G drives in my newer RAID-1 have never been over 35 C -- I do check them occasionally and they're almost always 32-33 C.

My older drives (which I think are above the newer ones, which would make sense) have been up to about 40C. But at the end of the day I never worry about any of that. Why? Because they all have 5 year warranties, and they're on two RAID-1 arrays with the important stuff backed up to the other array and to a RAID-1 box in a different room, and the *really* critical to rsync.net in Switzerland :) If a disk fails, just take it out, send it to the manufacturer, who cares.

If I were really paranoid I'd get a bigger tower and have some hot spares. But I haven't lost any data since the mid 90s (when I was 10 and still used Windows).

What RAID are you people using that doesn't let you see SMART data? Just use mdadm :) Maybe a little slower but then you don't have to worry about your RAID controller failing either.

lydgate on March 5, 2009 8:25 AM

I've had 4 hard drives fail all at once on me. When I got my system going again I realized that these hard drives all failed due to overheating all at once. This makes no sense to me but for some reason I can't keep these hard drives cool. Apparently I need a new case that will support several hard drive fans. I don't quite understand how a enclosed with a hd fan keeps these hds cooler though. The case will eventually get very hot with all the components in them running. I guess I'm a power user but these hd's these days aren't made to last.

Branin on March 11, 2009 2:04 AM

In my Fujitsu Siemens notebook, the hard drive and CPU are mounted ridiculously close to each other, causing drive running at 50C, +-3C (~23C room), and the CPU normal temp is between 57-77C.

The machine is placed on a table, on a device making an extra inch of free space underneath it, yet the drive temp is way higher than other laptops I've used in similar conditions!

But hey, what else to expect from drives having 2GHz dual core special coolers from AMD!?

Speaking of temperatures... Only 3 countries left now in the world - refusing to give up the use of Fahrenheit, and other completely illogical units of measurements: Burma, Liberia and USA!

RoyMccoy on April 3, 2009 11:02 AM

While a comment above states there's a study [3] showing no correlation, it's difficult to understand how--when the A/C failed at two buildings--we lost several drives; first those in external enclosures, then those in machines where the fans were failing...in more than one event and at different times of the year.

Another comment states an energy conservation rule--what goes in must come out--and makes a sweeping generalization concerning heat transfer. Unfortunately, this is entirely incorrect; one must take into account the energy required to create magnetic moments, lateral head translation, and excess energy that travels through the system (relatively) unhindered. To wit, energy loss through components that generate it either through resistance, friction, or by virtue of simply failing to be a superconductor is a small part of what's happening. If energy out (as heat) was the same as electrical energy in, the device wouldn't function.

Finally, friction. In a hard drive, heads FLOAT a small distance from the platters. Someone else said it best, in 1974:

The engineering requirement for a computer magnetic read-write head is similar to flying a Boeing 747 jet airliner at full throttle a few inches above the ground. [1] -- Donald Zipperian, CTO, Pace Technologies, Tuscon, AZ

In 2007, the 747 has to fly at 0.01--with 100,000 passengers [2]--but it's still not touching the platters! If they do, your heads crash and the drive--platters spinning 5-10K/second--will shear off metal particles, flinging them all over a landscape that's supposed to be clean and flat. As for the study, IT DID find negative effects over 104F (40C). When metal gets hot it EXPANDS, and there is NO guarantee that it will expand evenly, within the tolerance of the counter-force exerted by mechanical constructs such as compression springs or leaves, OR that the heads will remain in the same position in 3D space, resulting in data errors.

Finally, if you heat up a (ferro) magnet, it dies [4]. These temperatures are high [5], but it's useful info in case of fire.

Sources:
[1] http://www.nsf.gov/news/news_summ.jsp?cntn_id=100367
[2] http://www.magneticsmagazine.com/images/Presentations/2007/Goglia.pdf
[3] http://www.pcworld.com/article/129420/high_heat_may_not_harm_hard_drives.html
[4] http://www.physlink.com/Education/AskExperts/ae472.cfm
[5] http://www.tomshardware.com/reviews/raid-recovery,1542-3.html
and http://www.ocforums.com/archive/index.php/t-454159.html
and http://www.priorartdatabase.com/IPCOM/000111895

Kirk M. Schafer on April 23, 2009 1:28 PM

My HDD runs at 36-38 celsius when idle. Not sure how much is the load temperature, but when I use HDTune to scan for bad sectors, the temp hits 44-46 max celsius, so I assume 46 would be max load temp. Is it safe then?

John on April 30, 2009 4:38 AM

HDData is a small program that is always alert in the background of your Window Computer.
HDData measures all kind of Hard Drive parameters. If one of these parameters is exceeding a limit or sensing an error, HDData will warn you.
You can control what actions HDData will execute for you when the Hard Drive Temperature reaches a Warning and/or an SMART Alarm setting.
You have still time to save your work before the Hard Drive gets overheated and gets irretrievable broken down. You can act before losing everything!

Frits Molenkamp on May 4, 2009 2:29 AM

my laptop has a fujitsu hdd which works at 54 degrees, yet no problems

Zack on June 1, 2009 11:03 AM

Wow. You guys are temperatures that surprise me! My HDD have no active cooling and very little passive cooling besides natural convection and these are my average temps, I SWEAR!!!

Idle = 0% - 15% load
Active = 16% - 60% load
Load = 60% - 100% load

Seagate 7200.12 500GB Barracuda S-ATA II:
Idle: 30C
Active: 33
Load: 35

Hitachi 7200 RPM 750GB S-ATA II:
Idle: 38C
Active: 39C
Load: 40C

Other Temperatures:
CPU/GPU/Room/Motherboard/NBridge.

Idle PC: 45C/78C/26C/47C/67C
Active PC: 52C/85C/27C/52C/72C
Load PC: 61C/96C/27C/54C/78C

Basic Specs:
"All Stock Cooling"

"Part" (cooling type)
AMD Phenom x4 9850 @ 2.5GHz (stock)
6GB Samsung DDR2-800 PC-6400 (passive)
Nvidia GeForce 9800GT Overclocked (stock)
HD1: Seagate Barracuda 7200.12 500GB S-ATA II
HD2: Hitachi 7200RPM 750GB S-ATA II
NBridge Chipset: Nvidia nForce 720a MCP/ 8200 mGPU (offline)

So, ya are these good temps? Need your opinion.

Sincerely,
Narayan.

Narayan on June 16, 2009 3:38 AM

Wow. You guys are temperatures that surprise me! My HDD have no active cooling and very little passive cooling besides natural convection and these are my average temps, I SWEAR!!!

Idle = 0% - 15% load
Active = 16% - 60% load
Load = 60% - 100% load

Seagate 7200.12 500GB Barracuda S-ATA II:
Idle: 30C
Active: 33
Load: 35

Hitachi 7200 RPM 750GB S-ATA II:
Idle: 38C
Active: 39C
Load: 40C

Other Temperatures:
CPU/GPU/Room/Motherboard/NBridge.

Idle PC: 45C/78C/26C/47C/67C
Active PC: 52C/85C/27C/52C/72C
Load PC: 61C/96C/27C/54C/78C

Basic Specs:
"All Stock Cooling"

"Part" (cooling type)
AMD Phenom x4 9850 @ 2.5GHz (stock)
6GB Samsung DDR2-800 PC-6400 (passive)
Nvidia GeForce 9800GT Overclocked (stock)
HD1: Seagate Barracuda 7200.12 500GB S-ATA II
HD2: Hitachi 7200RPM 750GB S-ATA II
NBridge Chipset: Nvidia nForce 720a MCP/ 8200 mGPU (offline)

So, ya are these good temps? Need your opinion.

Sincerely,
Narayan.

Narayan on June 16, 2009 3:39 AM

Wow. You guys are temperatures that surprise me! My HDD have no active cooling and very little passive cooling besides natural convection and these are my average temps, I SWEAR!!!

Idle = 0% - 15% load
Active = 16% - 60% load
Load = 60% - 100% load

Seagate 7200.12 500GB Barracuda S-ATA II:
Idle: 30C
Active: 33
Load: 35

Hitachi 7200 RPM 750GB S-ATA II:
Idle: 38C
Active: 39C
Load: 40C

Other Temperatures:
CPU/GPU/Room/Motherboard/NBridge.

Idle PC: 45C/78C/26C/47C/67C
Active PC: 52C/85C/27C/52C/72C
Load PC: 61C/96C/27C/54C/78C

Basic Specs:
"All Stock Cooling"

"Part" (cooling type)
AMD Phenom x4 9850 @ 2.5GHz (stock)
6GB Samsung DDR2-800 PC-6400 (passive)
Nvidia GeForce 9800GT Overclocked (stock)
HD1: Seagate Barracuda 7200.12 500GB S-ATA II
HD2: Hitachi 7200RPM 750GB S-ATA II
NBridge Chipset: Nvidia nForce 720a MCP/ 8200 mGPU (offline)

So, ya are these good temps? Need your opinion.

Sincerely,
Narayan.

Narayan on June 16, 2009 3:40 AM

Sorry for the Uber posting but the first time I did it, my comment, it didn't post so I thought that I messed up the verification code sorry y'all.

Narayan on June 16, 2009 3:44 AM

Tha hard drive could fail for unknown reason

Computer Repair Service on June 18, 2009 11:47 AM

I have an old Seagate ST3160021A hard drive and an older Western Digital WDC-800BB in a system I put together 4 years ago.

The Western Digital is not SMART and has no temperature sensor. A kitchen meat thermometer told me it was around 52ºC. I could not find any temp specs on it from Western Digital.

Seagate says the ST3160021A's maximum temperature is 69ºC. It heats up to about 46ºC within an hour and then varies about 3 degrees. That same meat thermometer says it is at 45ºC while SpeedFan says it is 46ºC
SpeedFan reports it has gotten up to 63ºC, still well below its rated maximum.

My prime use is for photos and I back up regularly to external drives and disks. I have no idea how hard I am working the hard drives but the system is typically running 6 hours or more a day.

Both hard drives are quiet and seem to be doing fine. For what it is worth, all drive health programs I've used say they both are in good condition.

My other temps: Case 24ºC Radon 9250 Video Card 53ºC
CPU 50ºC Power Supply 34ºC

I agree that generally cooler is better but if you stay within the manufacture's specs and guidelines, devices should work throughout a significant portion of their rated MTBFs.
And most do.
Those that do not aren't on the market for very long.
At least that has been my experience as a lab QA (now retired) for many years.

BTW: At 24 hours a day, 10000 hrs MTBF is almost 14 months, 50000 hrs is over 5 years, 8 months and 100000 hrs is double that at 11+ years.
I do not expect many run 24 hours a day. Few devices can do so without some sort of scheduled maintenance or down time.

Ted

Ted on June 29, 2009 4:29 AM

Running at a cool 86F. Sidenote, do you know of any good free CPU/Case temp monitors? I've been using motherboard monitor for a while now, but there's gotta be something better out there.

jayson knight on February 6, 2010 9:52 PM

I'm also curious as to this so-called "research". The source here is a sort of e-zine and doesn't seem to state what the experiment(s) was/were. It sounds about as scientifically sound to me as something from GRC.

...which doesn't make it untrue, of course. I'm simply not convinced.

How do you even test for such a thing? Not all drives are created equal - for obvious reasons, there's no way to test the same drive's lifetime at different temperatures. So you can only do this in huge batches expecting a random distribution, and you can only look at the MTBF. And if the MTBF is 30 years, then do you *really* care if you get another 5 years (or even another 50 years) out of it? Nobody keeps a drive for that long!

Aaron G on February 6, 2010 9:52 PM

dc, I use motherboard monitor, but it's a bit dated (however it gets the job done) and is lacking support for some of the newer CPUs.

jayson knight on February 6, 2010 9:52 PM

My test utility says my drive is at 253?

You got to believe the utility is wrong here since the drive
is working great..

(Grin)

ron ernie on February 6, 2010 9:52 PM

well now wasnt that educational,...... shame MY HD seems to run 131F (58C)idle and arround 147F (64C) under load. been that way about a year now.

Time to get the gateway cleaned.

Castiron on February 6, 2010 9:52 PM

Malarky.

Typical operating specs allow for AMBIENT temperatures in the 5 - 55C range, which you've misread as DEVICE temperatures.

Example:
http://tinyurl.com/397s7pp

Vis Croceus on May 22, 2010 2:11 PM

In reply to Eric,
I was using a program called Speccy and it showed heaps of information about everything in your computer, inc hard drives. It showed S.M.A.R.T. info.
you can download it here:
http://www.filehippo.com/download_speccy/
:)

Kazz The Honourable Cheese on May 7, 2011 12:22 AM

Jeff Atwood is quite right you shouldn't play chicken with your data. That's what BACKUPS are for. It's far smarter to have your HDDs at 45 degrees C and regularly backup then it is to have your HDD at 20 degrees C and never or rarely backup. Particularly if your data is your professional work. In other words, the most important thing you should do is to keep backups of important data NOT to monitor HDDs temperatures.

Incidentally, anecdotal claims like 'My HDD died at it was 50 degrees C' or even 'I've usually kept my HDDs at 30 degrees C but this one case I let it reach 50 degrees C and it died' are basically useless in determing quantitative failure statistics like Google provided.

P.S. And no, RAID isn't a backup. If you think it is, you probably shouldn't be working professionally in the computing field.

Tet Yoon Lee on July 17, 2011 9:03 PM

«Back

The comments to this entry are closed.