What is optimal hard drive temperture?


ohlwiler

Recommended Posts

I have long had the goal to keep my hard drive temperature as low as possible. I have been rethinking this position lately. I now think it is more important to keep temperatures stable and within a reasonable range, not too cool, not too hot, but just right.

 

My unRAID server is located in my basement, and in the winter it can get quite cold. Unfortunately I don't have an extensive log of smart temperatures to analyze (now that would be an interesting package), but as I recall my SMART temperatures would get down to 16 C in the winter (on a spun down drive). Once spun up my temperatures would never get above 25 C. In the summer my temperatures range from 22 to 30 C. So how low should I allow my temps to drop? What are you comfortable with? Should I relocate my server to a location that has more stable temps?

Link to comment

Unfortunately answers to questions like this generally fall into the area of conjecture. None of us have large enough drive samples or labratory conditions to be able to say that increasing or decreasing drive temps by a few degrees, or keeping the temp range in a narrower range will or won't improve longevity. The google study is the closest thing we have - which tells us that drives kept too hot or too cold are not good for long term health.  But common sense is a pretty good tool and lead us to make smart, if not optimal, choices.

 

But even if we had a perfect recipee for optimal health, we may not follow it. Say, for example, that keeping the drive spinning and actively working 24x7 was best. Would we all start running tasks to keep the drives in high utilization, burning the most electricity?  I think we can agree that there are several variables we are trying to optimize and optimizing them for unRaid might be very different than optimizing them for other higher utilization purposes.

 

In general, the forums recommendation is to keep drive temps below 45c and watch for wide temperature variations. If you have had a lot of failures, it probably means that you should look to make some alterations (or maybe you were just unlucky).

 

I can tell you that I personally have been running an array for close to 3 years, and have had 30+ drives in and out of it over that time. I have never had a drive fail (although a couple times cable problems have made it appear I did). It runs in a basement (similar to yours) between 12c (once, normally ~18c) - ~28c (range from coldest spun down temp to hottest spun up temp) in the winter and between ~22c and 42c (once, normally ~36c) during the summer. Even if this is not optimal, it is good enough for me. And good enough may be about the best we can do.

Link to comment

Remember,  the temperature reported by some hard drives via SMART, is WRONG.  Some report it too high, and some report it too low.  Can't trust it unless you calibrate it with an IR thermometer.  I have a Samsung 1TB that says the drive temp is 17, when ambient is 21.

 

Also a temp of 35 when ambient is 30, is a different kettle of fish in terms of concern from a disk temp of 35 when ambient is 15.

Link to comment

I keep my server inside my house, where the AC/heater can help keep the temps stable.  Generally I like to see my drives stay in the 30s, though I have some that consistently run in the high 20s.  I don't think I've ever had drives run in the teens.

 

I agree with above - below 45 C and avoiding wide temperature differentials is the best we can hope to do.  I would consider 15 C to be about my threshold for a temperature differential.

Link to comment

bjp,

 

Thanks for the thoughtful reply. It is good to hear that someone else is seeing the same ranges as I and doesn't feel the need to improve matters. I recently revisited the Google study, and that is what prompted me to be concerned about my low temperatures.

 

I've been using unRAID for a number of years with one drive failure and one drive I pulled because the reallocated blocks count was growing. Right now, I have 18 drives in my array, the most ever and since we won't get another parity drive anytime soon, I'm hoping to minimize my risks without spending lots of money.

 

bubbaQ,

 

All three of my Samsungs seem to report fairly accurate temperatures, at least they seem to be in the range of their neighbors when the drives are first spun up. As I recall you've been chasing a fan control that is driven by SMART temperature data. I'd love to see motherboard maker's enable easy monitoring and adjusting fans for hard drive temps instead of processor temps.

Link to comment

I wrote to the authors suggesting that there could be an unidentified attractor in the data, asking if the "cold" drives were evenly distributed among sites, and if there was uniform distribution of drive failures over all sites.  It is possible that cooler drives come from cooler climates, and some other environmental or infrastructure factors at those sites were contributing to the failures.  (Or even possible hotter climates, where they run the A/C harder and longer).

 

 

I got no response.

Link to comment

I was looking at the google study again, particularly figure 5 on page 6.

 

In the 3 month numbers, cool drives (15-30c) fair the worst with about a 9% failure rate. But this is only for 3 months and these numbers may say more about early drive failures.  

 

Through 6 month, 1 year, and 2 year the failure rates are all under 5% across the board. The coolest drives fair the worst by a % point or two.

 

But the truly scarey numbers occur in year 3, with drives above the 40c mark spiking to ~11% failure rate and drives over 45c ~14%. Cooler drives fair very well in yr 3 with < 5% failure rate. In year 4 things settle down but drives over 45c showing significantly higher risk.    

 

I think the study is instructive that  even moderately hot drives temps (low 40s) is not a good temperature to allow drives to operate. And I also think it shows that the benefits of keeping drives cooler become more apparent with time.

 

The study is the only one and as bubbaQ notes, there may be other factors at play besides temperature affecting the reported results. But I do think that the spike in year 3 is clearly indicating that drives temps should stay out of the 40s if possible.

Link to comment
  • 3 months later...

"optimal" from a more casual point of view:

my first unraid server ran until yesterday with 42-45°C while accessing movies.

it got up to 50-53°C when parity check was running.

yes that seems deadly high for many and i was afraid myself, but never anything bad have happened.

 

 

now my server got ported to a bigger case with better airflow.

since then my temperature got down to 29°C peak. i even shut down my 2 case fans (top fan and rear fan)

only fans running right now are one small cpu fan, one quiet PSU fan and two cage fans for my 4in3 cages.

those cage fans got throttled down to lowest possible value by my fan controller. making it nearly silent.

 

now, all in all cooler feels better of course. but check your drive specs. my say it holds up to 60 °C. thats what i tested over 1.5 years and everything is still fine.

Link to comment

If you have OCD you will never stop checking your drive temps, playing around with fans, or doing something alike. I'm not into keeping my hard drives as cool as possible since most of the time the temps are just fine. For some reason people think 45c-50c is a real high temp, but is actually a normal operating temp for a WD Black drives. While it is a normal operating temp, mine somehow stay below 40c with no additional fans, rigs, or setups for a cooler case.

 

Here's the deal. Check your hard drive manufactures operating specs and keep it in that range. Your all set.

 

 

Link to comment

If you have OCD you will never stop checking your drive temps, playing around with fans, or doing something alike. I'm not into keeping my hard drives as cool as possible since most of the time the temps are just fine. For some reason people think 45c-50c is a real high temp, but is actually a normal operating temp for a WD Black drives. While it is a normal operating temp, mine somehow stay below 40c with no additional fans, rigs, or setups for a cooler case.

 

Here's the deal. Check your hard drive manufactures operating specs and keep it in that range. Your all set.

 

The Google study (which is all we've got), shows >10% failure rates in year 3 with drives run at over 40C.  My guess is that the failure rate is going up rapidly as those temps climb.  Could be (and this is just conjecture), that drives run at 50C could be failing at 20% or more. 

 

I think the fact your drives are under 40C means you're running in a safe zone, and would caution people to not run drives too hot.  If you have no choice than to run them hot, you might want to consider running them 24x7 so the temp variations stay low.  I'd feel better about running a drive at a constant 48C, than to have that same drive idling at 22C, and going up to 48C several times a day under load.  This is just my opinion.

Link to comment
  • 3 months later...

looking at that google study, wouldn't it make more sense to be aware of and possibly cycle your disks based on age, rather than OCD about temperatures?

 

Assuming a lot of people setting out building a new unRAID system will buy a bunch of drives which will be around the same age (possibly even the same batch).

 

Temperatures don't seem to have much influence in the first couple of years, but in years 3> they look like they do. If your disks are all in the same location and therefore roughly the same temperature, and they're roughly the same age, they are all going to drive up that 'increased risk' ramp at the same time.

 

So would it make sense to stagger your disk purchases where possible - and even vary the store you buy them from to avoid same batches? If you do need to buy them all at once (like me for instance), consider swapping a disk out for a new one every year - use the old one as a backup in an enclosure, or as a tepid swap drive. So eventually you get to a situation where you have a good mix of new and old drives, which in theory might help protect against mulitple simultaneous failure.

Link to comment

I haven't read the google study, but was it carried out over a short timespan?  Perhaps it says more about drives that were manufactured within a particular time frame (three years before the study), than it does about the age of the drive?

 

Perhaps I'm just clutching at straws because the ambient temperature here tends to be between 28C and 35C.  Of course, my drive temps do exceed 40C fairly frequently.

Link to comment

My Drives as I type this are performing a parity check and typically run around 35C. I under volted my fans on my CoolerMaster 590 because often my drives would run around 20C-22C all the time and honestly I'd rather them ramp up some in temp, but not from chilly to warm. Under volted I'm seeing my drives running around 25-27C.

 

Of course I'll keep an eye on my drives this summer when the house warms up some, however we are getting central air installed so it might be a moot issue. LOL

 

My CPU normally appears to run around 20-23C of course I have no idea how accurate it is, but with (3) 120MM pulling air in and (1) 120MM sucking it out + my CPU fan I'm guessing its running pretty cool.

 

I found this to be very interesting.

http://en.wikibooks.org/wiki/Minimizing_Hard_Disk_Drive_Failure_and_Data_Loss

 

Link to comment

A good quote from that very informative link

 

"A common misconception is that a colder hard drive will last longer than a hotter hard drive. A study by Google showed the reverse to be true.[1] Hard drives with average temperatures below 27 °C had a failure rate worse than hard drives with the highest reported average temperature of 50 °C, and a failure rate at least twice as high as the optimum temperature range of 37 °C to 46 °C.[1]"

 

 

Link to comment

 

"A common misconception is that a colder hard drive will last longer than a hotter hard drive. A study by Google showed the reverse to be true.[1] Hard drives with average temperatures below 27 °C had a failure rate worse than hard drives with the highest reported average temperature of 50 °C, and a failure rate at least twice as high as the optimum temperature range of 37 °C to 46 °C.[1]"

 

Except that the study was done on HDs between 80 GB and 400 GB in size and nowadays we use HDs with double of that on a single platter. The wide temperature variations will play bigger role.

A more recent study (done by a Russian data recovery lab) concluded that the heads of the newer high density WD HDs are very heat sensitive and the prolonged use above 45 °C will result in a very short lifetime. It is up to you to believe or not.

Couple that with the Unraid community obsession with the 5x3 with their rather limited air-flow, and the fact that WD drives are the only ones that do not keep the temperature range in their SMART status and you may understand the reported by many high "fatality" rate of the newer WD20EARS.

 

Link to comment
  • 2 weeks later...

All my drives have a 5 year warranty so I really doubt running them at 40C constant will kill them in 3 years. With proper cooling and air flow all my WD Black drives run 24/7 at 37/42C. According to Western Digital the operating temps for a Black drive is 0-60C. I really doubt running these in the 40's will have any negative affect. To top it off, I've ran similar drives (high speed / performance) for many years without failure. WD has a pretty good RMA service so if I was really bad I would RMA my drives before the 5 year mark. Now, the issue with that is by that time the speed and size of the drive will be obsolete so it wouldn't matter anyway, HENCE don't worry about it people!

 

 

Link to comment

I am sure that there are many factors at work when attempting to ascertain the 'ideal' or 'optimal' temperature for a hard drive, which no one seems to have really considered:

 

Thermal gradient - difference between internal and external surface temperature

Rate of change - how quickly the temperature goes up or down

Frequency of thermal cycling - how many times the drive goes from min - max - min

etc. etc.

 

It seems to be very common to judge quality of cooling by an absolute reported drive temperature.  However, the realistically achievable temperature will be very dependent on ambient air temperature ie, the temperature of the air entering the computer case.

 

I would much rather my drive runs at 40C, with an ambient of 30C, than that it runs at 35C with an ambient of 20C.  Also,  I would rather it runs between 30C and 40C than that it runs between 20C and 35C.

 

Ambient temps tend to vary between 25C and 35C here (no air-con, no heating).  Currently, at 1am, the air temp is 27C and my new WD20EARS is preclearing with a reported temp of 39C.  This would suggest that at an ambient of 35C, the drive would be running at 47C.

 

On the subject of cooling, good cooling is achieved by design - controlling the flow of air, where it comes from, and where it goes.  Simply making lots of holes in a case and hoping that, by some minor miracle, the air will flow to where it is most needed is a forlorn hope (except that I do acknowledge that convection may help to achieve this).  The incoming air should be forced to pass the most temperature-sensitive areas first, and then on to those areas which are less sensitive and can safely run at higher temperatures.

 

I would suggest that the best way to achieve this is to arrange fans to exhaust air from the hottest areas and supply the incoming air to the coolest areas.  Because, typically, the processing load in an unRAID server is fairly minimal, while disk activity tends to be high, we should arrange for the incoming air to pass the disk drives first and then progress to the main cavity in the case, where the main electronics (processor, memory etc) reside.

 

Ideally, I would arrange for the only fans to exhaust from the main case cavity, and for the only path for air to enter the case to pass through the drive housings (possibly using additional fans, if really necessary).  To do this, block all holes in the case except the fan apertures, and any apertures in front of the drive mountings.

 

I have already proven that, with only the standard fans in my Thermaltake V5, the drive temperatures are reduced by around three - four degrees celsius simply by blocking the open grills in the side panels.  When I can, eventually, obtain drive backplanes, I intend to experiment with removing any fans from the backplanes, and adding the optional 120mm fan in the case side panel.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.