On or off?

It’s a reasonable question to be asking: should you leave your computer on or turn it off every night? Recently Simon Hill of Digital Trends tackled this question, bringing in Steven Leslie from Geek Squad to assist.

The biggest argument for leaving your computer on constantly is that it is less damaging to your components in the long run. This is only partly true, and it’s based on what I’d considered a somewhat flawed observation. The time when a computer part is most likely to fail is when it is being powered on, not while it is running — don’t read that to mean a part won’t fail while it is running, because it most certainly can, it’s just not as likely to do so, just as your car is more likely to fail to start than die while running or idling.

All computer components have a rating called MTBF, or mean time between failure. “Mean time” means average. If a part is rated at 50,000 hours MTBF (typical for a computer hard drive), then that means the manufacturer has determined that the average lifespan for a component would be 50,000 hours of constant running — or it should be able to run continuously powered for over 2083 days, or about 5 years and 8 to 9 months. Now given this is the average rating, your part could die in a year run continuously, or it could die in 10 years.

So the idea goes that powering off your computer every day should extend the lifespan of your computer components because the MTBF is rated on constant use. The slight “surge” in powering on your components isn’t going to cause any significant stress to your components unless you’ve got bad wiring in your home or you’re connected to a substandard power grid. But if that were the case, and you’re not using an uninterruptible power supply (UPS, and I’m not talking about the courier), surge suppressor or power conditioner, you’re already causing damage to your computer whenever it’s powered on.

Modern desktop power supplies can more robustly handle substandard power delivery coming from the wall, but there are still limits beyond which that is not true, and that is in part still determined by what your computer requires to operate — i.e. if your computer is trying to draw 500W from the wall, but is having a difficult time getting it due to the power delivery in your home, you’re going to have stability issues. To find out the quality of the power delivery on your grid, you’ll need to talk to your power company. If the grid is fine, you may need to upgrade the wiring in your home — this will have significant benefits beyond your computer, as it could increase the lifespan of everything in your home, including major appliances, and possibly improve energy efficiency while also decreasing the possibility of a fire due to faulty electrical lines in your home.

Now this doesn’t change the fact that a component is still most likely to die when powering on. That’s just the nature of any electrical component.

In his article, Hill writes this: “A traditional hard disk drive, for example, has moving parts, whereas a solid state drive doesn’t and is far more robust as a result.” Okay let’s tackle this idea.

An SSD, or solid state drive, does not have moving parts. As such it consumes much less power and will run cooler and quieter, and is much less prone to shock damage (as in from a striking blow, not electrical shock). But they are significantly more expensive than traditional platter hard drives (HDDs). But whether they are more robust depends on how you use them.

Here’s the caveat: you don’t want to use them in systems where there will be very frequent changes to the data stored on it. Content creators, software developers, and the like really should not use an SSD as primary storage. Using it to store the operating system and software is one thing, to ensure that programs load quickly, and it’s information that is unlikely to change frequently.

But use a traditional HDD for storing work files and the like.

This is because SSD performance degrades over time, and will degrade faster if you don’t keep much free space on it while using it as a primary drive, or are making frequent writes, deletes, or rewrites to the data on the drive. You can keep this from occurring nearly as much by having multiple SSDs configured in a RAID 0 configuration (beyond the scope of this article), thereby spreading writes and rewrites across multiple drives. But you’d likely still be better off having a platter drive, especially since platter drives are much less expensive, and capacity is going to be more important than speed — doesn’t matter how fast your data is saved and loaded if you can’t store it.

The laptops issued by my employer have SSDs in them, and my upcoming device upgrade will also have an SSD. I’m actually considering talking to device support about moving over to a platter HDD, even if I have to pay for the drive out of my own pocket. This is in part due to capacity — see my previous paragraph. The drive I currently have is only 120GB, and the drive in the new laptop is 256GB, and as a software engineer, I prefer capacity over speed, so I’ll be discussing putting a larger platter hard drive in the device in its place. Word has it I may be able to have an HDD alongside the SSD, so I may explore that option as well.

SSDs are faster, but, again, their performance will degrade over time, and the SSD in my laptop is currently not much faster than a traditional platter drive. This is in part because my SSD — like those of my colleagues — has about 75% of its capacity currently used. That will cause an SSD’s performance to degrade faster because the algorithms the SSD’s firmware uses to prevent that performance degradation from occurring cannot work nearly as well.

Several of my colleagues have had to replace their SSDs over the last couple years due to the drive failing. The fact we also use full-drive encryption at work I think plays into that.

Whether you will encounter those limitations depends on what you do with your system, but do not ignore the fact that any drive, HDD or SSD, can fail on you at any time, so keep regular backups of important data. And the power delivery to and inside the system is what is most likely to cause a component to fail.

Hill provided a couple lists of reasons to leave your computer on or turn it off. The first reason to leave it on is if you’re using the system as a server, which most aren’t doing that, so it’s a moot point. The other two reasons are perfectly applicable:

  • There are background updates, virus scans, or other activities you’d like to occur while you’re away.
  • You never want to wait for it to start up.

It’s always best to have scans and updates occur during downtime, especially if the updates require a reboot — which means even if you’re not shutting down every night, at least have the system in a state where a reboot can automatically occur if it must. Speaking of which, even if you don’t shut down every night, a nightly reboot can still be beneficial.

Now let’s talk about the reasons to turn it off:

  • Leaving it on wastes electricity and can slightly increase your power bill.
  • You don’t want to be disturbed by notifications or fan noise.
  • Computer performance generally benefits from an occasional reboot.

Of these, the last is kind of nonsensical as you can manually reboot your computer. You can do a reboot in the morning while making coffee, or at night before going to bed. If you don’t want to be disturbed by fan noise or notifications, then put your system in another room so it’s not going to disturb you. And if the fans in your system are loud, replacing them is always an option, as there are plenty of options available for quiet fans that still push a good amount of air.

Leaving your system on will obviously mean it’s using power. Your system does not draw a continuous level of power from the wall and will only draw what it needs up to the rated power of the power supply. So if  you’re leaving your system idling overnight, it’s increasing your power bill by only a few dollars a month, so shutting it down every night is probably not going to make a huge difference on your power bill, especially if the amount if time it’s off will always be less than the amount of time it’s on.

How little power it draws when idling depends on the processor, mainboard and other components. For example Western Digital has the WD Green drive which is designed to spin down when not in use — this makes the drive undesirable for anything other than backup storage. Modern processors are actually fairly power efficient today, with technologies built in that allow them to sip power when idling. The efficiency of the computer’s power supply will also determine how much power the system uses while idling.

Then there’s the question of sleep or hibernate. Just like powering off, hibernate will only be of significant benefit if the computer is going to be off more than it will be on.

Putting your computer into a sleep state is also a beneficial option, but with one caveat: make sure your system is configured so the mouse cannot wake it up. If you have a wireless mouse, turn it off before putting your computer into sleep. A lot of optical mice on the market have very high resolution tracking on them, meaning the slightest shift on the mouse could wake your computer unexpectedly, negating the benefit of putting it into a sleep state. Hard drives spin down when the system is put to sleep, so unexpected wake-ups caused by a jolt that causes the mouse to report it’s moved will spin it back up sooner, and potentially more frequently.