25

My secondary hard-drive (platter) sometimes has a lag of 1-5 seconds when it hasn't been accessed in a while, causing explorer or other software to stutter.

Reading online, I saw the advice that you should change it so the hard drive never turns off using this setting:

enter image description here

Is that relatively safe to do?

JoshuaD
  • 845
  • 2
  • 9
  • 24
  • 6
    I'm more curious if a normal windows computer with multiple programs running ever enters a state when HDD is not used, even if OS is on extra SSD. – Tomáš Zato Jun 06 '19 at 12:47
  • 4
    If you have enough RAM, that should be possible. Programs only read directly from RAM (regardless of OS), so if you're not using giant files, you could theoretically run just with RAM. I'd expect something like web browsing, and even small downloads to be fine without the HDD. – Cullub Jun 06 '19 at 14:44
  • 9
    This is my secondary harddrive, so it's when I go to access data files, not programs. – JoshuaD Jun 06 '19 at 17:35
  • 7
    I've been leaving all my desktop systems running 24/7 for years (decades, actually), and have always disabled power savings on hard drives. I've only had one HDD fail in the last 20 years and that was on a laptop that _did_ power down the drives. – Carey Gregory Jun 07 '19 at 00:42
  • 6
    [Is turning off hard disks harmful?](https://superuser.com/questions/17228/is-turning-off-hard-disks-harmful) and [**How** harmful is a hard disk spin cycle?](https://superuser.com/questions/197862/how-harmful-is-a-hard-disk-spin-cycle) - the only consensus is that there's no consensus. 20 plus years ago this was worth debating; I don't think it is anymore. – Mazura Jun 07 '19 at 00:54
  • I never knew desktop hard drives even went to sleep. – RonJohn Jun 07 '19 at 13:50
  • @RonJohn It depends on the hard drive, e.g. WD Green drives will aggressively spin down to save power. – forest Jun 08 '19 at 07:21
  • Desktop drives? Sure. It's more a question, as Tomás noted, of whether you'll ever get to actually see it happen. If it's a primary system drive, on any OS, forget it. Doesn't matter how much RAM you have, when in use the OS is constantly writing to the disk — updating logs, committing state data, etc. But a secondary drive, yeah. My Linux fileserver boots off an SSD, and has two multi-TB HDDs in it. The one BitTorrent uses is never even _idle_, forget asleep. But my first SSHFS request of the day will often be followed by the faint sound of that third disk spinning up. – FeRD Jun 30 '19 at 20:15
  • Of course, the other reason you don't see desktop drives spinning down much is, idle spindown is disabled by default on nearly all desktop drives, except for a few rare exceptions. (Like the WD Green drives, as forest pointed out.) Ditto the IDE/SATA powersave timeout feature in the machine's BIOS, traditionally. It's _there_ in any AWARD/Phoenix BIOS with a power-management page, but I don't think I've ever seen it enabled out-of-the-box. (I'm pessimistic about the chances that UEFI managed to break with that tradition.) – FeRD Jun 30 '19 at 20:20

4 Answers4

35

Yes, this is perfectly safe.

The advantage of powering mechanical drives down when not in use is simply to save electricity. Mechanical drives typically draw between 5 and 6 watts. You can look at your electric bill and see how much your power company is charging you per watt-hour and get an estimate on how much more it will cost you to leave it on. It isn't much.

Toby Speight
  • 4,866
  • 1
  • 26
  • 36
Keltari
  • 71,875
  • 26
  • 179
  • 229
  • 7
    Also, just because it is on, doesn't mean it is spinning or at least spinning at full RPM. It will still have a low power mode that will consume less wattage when not in actual use. – Jammin4CO Jun 06 '19 at 14:00
  • 14
    @Jammin4CO: Low-power / non-spinning but still "on" is what this setting is all about. It doesn't literally cut power to the HD's power connector; the controller stays on and it leaves this state by simply sending a SATA command. – Peter Cordes Jun 06 '19 at 14:48
  • 8
    Do they really have hard drives that spin at a lower RPM to save power? – JPhi1618 Jun 06 '19 at 20:01
  • 1
    @JPhi1618 I've only heard of different actuator "aggressiveness" settings to lower noise. – Nick Jun 06 '19 at 20:22
  • 3
    @JPhi1618 just found [this PDF](https://www.seagate.com/files/docs/pdf/en-GB/whitepaper/tp608-powerchoice-tech-provides-gb.pdf) from Seagate describing such a power saving measure. Seems to be more of an enterprise feature. – Nick Jun 06 '19 at 20:26
  • 5
    @JPhi1618 WD Green drives were initially marketed as variable-RPM drives, but they ended up only including that feature on a few specific models. The rest were just crappy 5400 RPM drives in a desktop form factor, when everyone else was building 7200 RPM or 10000 RPM drives. – FeRD Jun 07 '19 at 07:30
  • 2
    @FeRD IIRC I read somewhere that they were marketed as "between 5400 rpm and 7200 rpm", but they were actually fixed at 5405 rpm. Not sure if the source was credible, though. – gronostaj Jun 07 '19 at 13:03
  • @FeRD I thought Green drives were low heat, making them good for low stress systems like home media servers. – RonJohn Jun 07 '19 at 13:56
  • 3
    @RonJohn, Green drives have very aggressive power-saving settings. They'll spin down at the slightest excuse. – Mark Jun 08 '19 at 07:18
  • @PeterCordes The PCH SATA Controller supports a D3 state that actually cuts power from the device (probably for use during hibernation/suspend). But for the SLEEP state, the ACS-4 spec defines the PM states in terms of latency (as usual), the non-spinning state of the HD is likely mapped to the formal IDLE state, while the more power saving state (entered by the OS) is likely mapped to STANDBY (which has a timer feature too). While that setting won't cut the power it probably controls STANDBY, not IDLE (which is what Jammin4CO was referring about, I think). – Margaret Bloom Jun 08 '19 at 16:23
  • @MargaretBloom: I was talking about the electronics onboard the hard drive, which control the mechanical parts. Not the "SATA controller" on the motherboard. In all cases there is power on the supply rails on the power connector of the HDD, unless the whole computer is asleep with the power supply only providing power on the standby supply rail. – Peter Cordes Jun 08 '19 at 17:22
8

Yes, it’s safe but it has its disadvantages.

1.- Shorter life span

Due to constant spinning mechanical parts will start to fail earlier. This is just how stuff works, things wear, tolerances broaden, things break. Most popular brands of hard drives will last a long time, if you take care of them. My thoughts on this, is have one small fast drive for the work and large drives for storage, which will be powered down most of the time. This way you use them efficiently.

2.- Damage to platters

If bumped when when mechanical parts are moving then you will risk irreparable, way too expensive, damage. Some laptops have an accelerometer that deactivates the HDDs, or move the heads away, when it's in free fall. I don’t know if this feature is available for desktops, or servers.

3.- Electric bill/ Laptop battery

It will draw more power, plain and simple. The power consumption will vary according to physical size, storage technology, etc; but in the end it is going to take its toll on your bill, or battery.

Do it if you need to, otherwise leave it be: it's just a couple of seconds.

Community
  • 1
dmb
  • 1,284
  • 9
  • 14
  • 6
    @Navin Seconded: This is why servers which have been powered on for months / years, will often lose a HDD after a power-outage. – SiHa Jun 06 '19 at 10:58
  • 8
    #2 also seems irrelevant, since [active hard drive protection](https://en.wikipedia.org/wiki/Active_hard-drive_protection) (if present) is unlikely to be affected by the turn-off-when-idle setting the OP is asking about. (Well, except that if the drive happens to be already off — and the head parked — when it's dropped, then it _may_ survive even if the active protection fails. But that's not something one should rely on.) – Ilmari Karonen Jun 06 '19 at 12:09
  • 8
    For #2, no - active protection is a laptop thing only. If your server or desktop finds itself in free-fall while powered on and operating then you probably have bigger problems on your hands. #1 is just wrong. – J... Jun 06 '19 at 14:13
  • 6
    @J... No it's not wrong. Drives have two lifetimes, one is a lifetime on how many hours it can spin from e.g. bearing wear and aerodynamic wear of the heads floating on the platters. The second lifetime is the wear from unloading the heads, either as they touch down on the platter or they pull off on a ramp. Exceeding either lifetime can cause a drive to fail, this is why SMART monitors start/stop cycles and spinning hours. – user71659 Jun 06 '19 at 17:44
  • @user71659 For a desktop system I'll partly agree - the extra start/stop cycles will be harder on the drive than just letting it run, though. It's not just the starting and stopping, it's also the temperature cycling that contributes to wear. A desktop system will probably shut down or sleep/hibernate every day anyway, so whereas a data centre drive might continue spinning and working (if just left on) long past when it would fail if stopped and restarted a consumer device will be stopping daily anyway so it will just fail when it fails. – J... Jun 06 '19 at 18:50
  • 26
    In my experience, your point #1 is flat wrong: a hard drive is most likely to fail during spinup. An always-on hard drive will last longer than one that's constantly being spun up and down. – Mark Jun 06 '19 at 20:20
  • 3
    @user71659 In a modern HDD having the heads touch down on the disk is generally entirely fatal. They are designed to move onto the ramp immediately in case of sudden power loss. But yes, ramp load and unload is bad for the heads and will wear them out if they are parked often. It is usually better to avoid lots of ramp loads and unloads by not parking at all if the disk is accessed regularly. Much like stopping the engine of a car when it stops for long periods, but not for a moment in traffic. – Vality Jun 06 '19 at 22:17
  • @Vality No that is incorrect. Many enterprise drives and a number of [3.5 inch desktop drives](https://www.seagate.com/www-content/product-content/barracuda-fam/barracuda-new/en-us/docs/100804187a.pdf) still utilize contact start/stop technology which means the heads touch down on a laser microtextured region of the disk. Ramp load is not universal. – user71659 Jun 06 '19 at 22:54
  • 4
    @user71659 Im aware this is true for older drivers, and the drives linked there are of very old design (even if still available, some old drives have very long production runs). I don't recall having worked on a drive design that low capacity for many years. – Vality Jun 06 '19 at 23:03
  • 1
    @J... that depends on the person, I guess. My desktop stays on 24x7, only powering down for (now vanishingly) rare h/w upgrades. – RonJohn Jun 07 '19 at 13:53
  • 3
    #1 is incorrect and in fact a hard drive with a constantly spinning disk will survive potentially a lot longer than one that has to stop and restart frequently. This type of failure is common in all sorts of devices, from airplanes to light bulbs. There's a reason why the oldest bulb in the world is kept on 24/7. – forest Jun 08 '19 at 07:19
  • 1
    @forest - Incandescent light bulbs are a bit of an extreme case: the electrical resistance of the filament increases a lot with temperature (a factor of 5 maybe? I forget what I got last time I measured the cold resistance, then current at known voltage when on). But anyway, a thin spot on the filament will heat faster because it already has less resistance, amplifying the effect until the rest of the filament heats up. So it's very easy for a part of a filament to turn into a fuse and "blow". Lights on dimmer switches last *much* longer for this reason. – Peter Cordes Jun 08 '19 at 17:28
7

Yes, this option is only designed to save electricity. In fact, spin-up/spin-down count is a much stronger indication of wear and tear than actual time spent spinning. Hard drives are designed to spin up and stay spinning; constantly spinning them up and down actually damages the drive more than keeping it spinning. See Is turning off hard disks harmful?

user464014
  • 114
  • 4
0

This is an old post an a lot of opinions, but wanted to add an experience that I had

I used to work on ISP long long time ago, and we had 2 identical servers. They were configured as fail over, and some times we switched to one server while we were updating (patching) the other one and so on. It wasnt so critical so we had couple of drive and using raid at software level, both running 2016 (top OS at that time).

One server had the option to put the disk to sleep after 20 mins of inactivity, and the other not (A mistake during the set up). 2 years latter one server start having issues with 2 of the 6 disks. And ended up being the server with the "sleep" disk configuration, and the smart errors were mostly on the spin up.

From there, I understood that the drives suffer more stress spinning down and up, than kept running all the time.

At the end of the day, the energy difference (not a subject of interest on this ISP), wont make difference if you need to replace a disk early. In fact, will be more expensive by far.

DefToneR
  • 103
  • 4