How many hours can a hdd last?

[Small2134]

Hello everyone:

 so my friends laptop has one year of disk usage, not years but like one year that the hdd has been on. I know hdd' last for 3-5 years but is that in like usage? My friends hdd is almost if not always at 100% usage but he never takes it off his desk. He is wondering if the hdd will die soon or not? Loading times for Windows from boot is like 20 minutes.

[Enderman]

Depends on the HDD, go look at the manufacturer website, it should say the MTBF there, or how many GB/day it is rated at.

[pwn_intended]

If the drive is 100% (or almost) full, it will be slow AF. Your buddy needs to do some digital spring cleaning. Hard drives generally outlast the useable life of the computer that they come with.

[Curious Pineapple]

It depends on many factors. I've had brand new drives with random faults, and I also have a few IDE laptop drives from Pentium 2 era machines that still work. Operating conditions, shock, spin-up count etc all factor into how long a drive will last.

[Electronics Wizardy]

I have drives with over 50000 hours with no errors. 

 

I also have a batch of 2tb hitachi's at around 30000h hours working great. Normally more than 20000 is normal.

[Cla55ifi3xd]

I built a pc with a 2tb hdd 5 years ago and it just failed. Very light usage maybe once a week for maybe an hour. Stuff happens. 

[Jovidah]

Honestly I don't think there's any telling...it's just luck of the draw. The PC I'm currently on has 2 ridiculously old Seagate drives on it... one is 7,5 years old and the other is 10,5 years old. And they're both still doing fine. They might both die tomorrow, they might last for another decade. There really isn't any telling.

Does that mean your drive will survive that long? Nope. They could. Or they might die tomorrow. The fact that it's so unpredictable is why there's always so much focus on redundancy (in RAID setups) and backups.

Even SMART statistics will only predict about a third of all possible failures. MTBF specs are really only useful in larger numbers of drives and to determine averages, but aren't very useful in predicting the lifespan of an individual drive.

[mariushm]

MTBF doesn't say anything about how long a hard drive will last.

In a very simplified way it's something like this... a MTBF of 1.500.000 hours means that if you install 1.500 hard drives somewhere, they estimate that on average you'll have one hard drive dying every 1000 hours  (1500 x 1000 hours = 1 500 000 hours).  AVERAGE, meaning some may day in an hour , others may die in 10000 hours, others may die in 50000 hours.

Manufacturers don't use this anymore, the go with AFR (annualized failure rate) and for WD and Seagate the value is under 0.8% meaning that they estimate if you have 1000 hard drives running 24/7, less than 8 should fail every year, or something like that, for the first 5 years or the warranty period.

 

See WD's statement : http://support.wdc.com/knowledgebase/answer.aspx?ID=665

See seagate's statement (more detailed) : http://knowledge.seagate.com/articles/en_US/FAQ/174791en?language=en_US

 

Google had a study about hard drives failures which you can read (or at least browse through and view the graphs) here:  http://static.googleusercontent.com/media/research.google.com/en//archive/disk_failures.pdf

 

Basically, most hard drives die in the first three months, and if they manage to last this long, the failure rate goes up after around 4 years, but not by a huge margin. See Figure 3 in pdf, note the high percentage under High usage in the first three months... if you write and read from it constantly in the first three months and it still works you know it's right So if you keep your system running 24/7 for a couple of months keeping a new hard drive spinning and working, statistically there's a high chance that hard drive will last a long time

Also, temperature is quite important ... may seem wrong, but they don't like very cold temperatures, it's best to keep them around 35c.. 45c with the mention that failure rates increase after 3 years if they're constantly above 40c (see figures 4 and 5)

 

backblaze also has some very useful posts about their hard drive population, which brands fail more and so on, but you have to know how to read their values!

Link: https://www.backblaze.com/blog/hard-drive-reliability-stats-q1-2016/

They shove 40+ hard drives in cases with not that great anti-vibration things, and they run 24/7 at higher than usual temperatures and they also use hard drives not rated for NAS usage (more than 4-6 drives in a case, firmware not tuned for vibrations and heavy usage) so their numbers don't often correlate to how hard drives would die in regular computers.  Google's numbers also don't correlate that well, but they're much closer to how regular users would experience failures because they logged hdd data from lots of servers in datacenters, from computers with small number of hard drives in them, from computers with lots of drives in them... not just storage pods like backblaze has.

 

hardware.fr  also periodically publishes  lists of how many parts of specific models were returned to warranty in specific period: http://www.hardware.fr/tag/757/taux-retour.html

 

For example in the latest article here : http://www.hardware.fr/articles/947-6/disques-durs.html  they count all the products returned to the store for warranty reasons before 1st of April 2016, for products sold between 1st of April 2015 and 1st of October 2015.  But you can compare results for APR-NOV with previous period to see if specific brands fail more often as time goes by or the other way around.

 

As you can see in that period - at least in France and other countries served by that store -  WD Black 4TB failed more often, 4.32% of them (over 100 pcs but less than 500) were returned between moment of sale (any time between apr 2015 and nov 2015) and returned before apr 2016. Toshiba followed with 3.59% and 2.88% (both with more than 500 pcs sold in that period).

 

i personally have a WD 1TB drive with over 51902 hours of operation , a Seagate 2 TB with 48434 hours and a WD 2 TB with 23281  hours and a couple of other drives with less than 20k hours.

Only the 2TB Seagate drive has a few sectors with problems but they're marked away and don't affect the functionality of the drive (i store tv shows i may want to rewatch, so i don't care if the drive will eventually fail)

The hours count is so high because I keep my computer running 24/7, the current uptime of this system is almost 74 days.

 

[Jovidah]

AFR is again utterly useful in predicting outcomes for an individual drive though. 

Infant mortality is up for discussion though... This article: https://www.usenix.org/legacy/events/fast07/tech/schroeder/schroeder.pdf actually doesn't replicate it in a 100 000 disk population. In her findings the disks die pretty evenly through time.

 

The backblaze articles...meh... their methodology is utterly bonkers and there's so many problems with their statistics that it's really quite impossible to draw conclusions from them. In my opinion it's so shit Seagate should have seriously considered sueing them for slander. I guess they didn't in fear of a PR-backfire or the Barbara Streisand-effect.

 

I really prefer the harware.fr statistics as they're a good measurement of how drives behave 'in the wild'. However there's also a major problem with those statistics: different products will see different treatment. For example RMA rates on high-end motherboards are relatively high not because the motherboards are worse, but because they are fried to death by overenthusiastic overclockers. The lower RMA rates on some 'boring budget boards' are not a reflection of their higher reliability but of their usage.

The same is likely with other products; high-end and low-end products are likely to see different usage, thus their RMA rates aren't directly comparable. IMO you should really only use their stats to compare 'within category', and then mostly to compare models of the same characteristics.

For example, comparing a 2 GB Seagate disk to a 4 GB WD disk is a bit dishonest, since all 4 GB disks are less reliable. Although you could try to work out the RMA/GB rate to make it fair.

 

Long story short... it's really an uncertain shitshow and no one can tell you anything about the lifespan of an individual drive. We can make estimations, predictions or whatsoever, but all of that is just calculting chances at either of the 2 binary outcomes. If the data is important enough to care, just back up the data. If not, don't worry about it until it dies. Which might be 10 years in the future for all you know.