Is suspending vs. shutting down bad in the long term?

[jyc]

Suspend & resume from suspend (to RAM) is very fast and convenient on my ThinkPad, but after thinking about durability I started to wonder if anyone here frequently suspended their laptop instead of shutting it down. Has doing so caused any hardware problems? The only thing I could imagine being damaged would be RAM, and I'm not sure how long the real-usage lifetime for it would be.

Thanks for indulging in my esoteric question.

[Medessec]

I really don't think it causes any problems for the hardware as much as it does for the software... I used to use Standby all the time with my old Toshiba laptop, and I'd encounter networking issues and applications would refuse to start due to .dll errors after 2-3 resumes from Standby. Hibernation also takes a similar toll on Windows, but to a lesser extent.

But you shouldn't be worried about the hardware going out due to suspending your laptop.

[ajkula66]

Welcome to the forum!

Suspend & resume from suspend (to RAM) is very fast and convenient on my ThinkPad, but after thinking about durability I started to wonder if anyone here frequently suspended their laptop instead of shutting it down. Has doing so caused any hardware problems? The only thing I could imagine being damaged would be RAM, and I'm not sure how long the real-usage lifetime for it would be.

Thanks for indulging in my esoteric question.

Not an esoteric question at all.

I'd argue that the answer depends on the specific machine.

Older units with early application of lead-free BGAs tend to dislike thermal cycling and are known to eventually fail from it. The last generation affected by this phenomena would be - albeit for somewhat different reasons - *61 series with nVidia GPUs.

Which ThinkPad do you have?

[Cigarguy]

Don't think it matters either way. I find hardware in general is pretty reliable. I suspend to RAM on my desktops because I know there's less chance of a power interruption which can affect software and is a bit faster. I hibernate my laptops because there's no drain on the battery.

[precip9]

There is an estimate of RAM lifetime, ~7 years of power-on time. It is a very old estimate; I have no idea whether it's gotten better or worse. And not all specs are equally valid; the number could have been someone's napkin exercise.

Nevertheless, there are two competing mechanisms of semiconductor demise:
1. Thermal cycling, which tends to break the chip bonds.

2. Junction diffusion. There is a naive notion that because semiconductors are solid, they do not burn out. This is incorrect. The miracle of semiconductors is in the "dopants" small quantities of impurities deliberately introduced into the silicon, which provide the charge carriers. Over time, the dopants diffuse across the junction. At a certain point, when enough diffusion has occurred, the semiconductor is no longer a semiconductor. It has, in fact, "burned out", and this is a direct consequence of power-on-hours.

On my desktops, I use Windows 7/8 hybrid sleep, specified under "advanced settings" of the power scheme. The machine first suspends to RAM, but also writes a backup to disk. If power fails, it turns into hibernation. After an interval specified by the user, in my case, 4 hours, the machine goes into hibernation. This seems a reasonable compromise for my personal habits. The same choices are available with laptops. Since I don't use my Thinkpads as desktops, the intervals for transition from suspend to hibernate are shorter.

[Medessec]

RAM tends to be one of the more rock solid components of a computer, especially when bought from a noteworthy brand (Patriot, Corsair, Kingston). Although what precip9 noted does apply, even if RAM fails, it's one of the more cheaper and much easier components of a laptop to replace or upgrade. And if RAM fails due to the chip contacts coming loose, it probably wouldn't be due to hibernating or putting the machine into standby, which in terms of cycling would be indistinguishable from shutoffs.

[RealBlackStuff]

Corsair RAM is only noteworthy because it is so UNreliable and has so many problems!
Check Google for "corsair ram problems" and come up with About 41,900,000 results

Give me Samsung, Hynix, Axiom, G.SKill, Micron, Elpida, Mushkin to mention but a few...

[precip9]

Here are some dated reliability stats compiled from surveys of manufacturers. Don't ask me for sources; I just picked them up along the way:

Intel CPUs: 1%/anum
RAM: 2%/anum
AMD CPUs: 3%/anum

The RAM stats do not apply across brands. Since they are manufacturer stats, the real junk suppliers have been filtered out. As Realblackstuff points out, some RAM brands are junk.

RAM failures are typically caused by junction diffusion. When the silicon has a crystal defect at some cell, it provides a migration path. The RAM tests good, but is subject to accelerated aging in use.

As a part, not as reliability per cell, RAM is a high failure part. This is because there are so many transistors in a RAM chip, compared to other stuff on the motherboard. It is upsetting to see a trend where the first 4 gigs are soldered in. These machines will have roughly 2% added to their irreparable failure rate. And it deprives the option to try to solve flakiness by swapping the stick.

But the consumer has voted. The consumer is not like us

[RealBlackStuff]

That's one more reason you should also stay away from so-called "Value-RAM" and its ilk, it's 2nd choice RAM or worse!

[precip9]

"Valueram" is not a brand. It is an umbrella, and it would not be right to tar it all small-run SODIMMs with the same brush. Over the past year, I bought six sets of 8G DDR2 SODIMMs from the same eBay seller. Located in San Jose, this seller assembles the RAM in his own shop, or somewhere in the Valley. One set was obviously defective upon arrival. A problem with another set (" My X61s is sick" thread) turned out to be a defect on some of the gold fingers. Swapping the pair in the sockets fixed the problem.

I don't advocate no-name RAM, but in this case, it was justifiable. The seller sourced chips manufactured for router upgrades, providing a product with too little demand for the "names" to bother. These are the only chips I've seen with timings identical to Lenovo spec.

For DDR3, I go with with the "names", Kingston and Micron.

[rkawakami]

Speaking from a memory test engineer's viewpoint, I'd offer the following:

- Yes, semiconductors do burn out, but when compared to the spinning platters and moving mechanical parts inside a traditional hard drive, then I believe that DRAM still has a superior reliability factor. Drop a laptop to the floor while the system is on and I'd bet that you'd see more hard drive failures than memory . Generally speaking, a semiconductor can be forced into failing early in its life (i.e., infant mortality) by subjecting it to a burn-in cycle. This is typically done by baking the parts in an oven at a higher than normal operating temperature for a specified amount of time (24 to 72 hours). In most cases, manufacturers will also electrically operate the memories ("dynamic burn-in") at this high temp. This is all done to accelerate the junction breakdown process that precip9 mentioned earlier.
- Even if you consider using an SSD in place of an HD, because of the inherent nature of flash memory having a limited number of erase/write cycles (and even taking into account the wear-leveling algorithms in place), I'd still argue that DRAM has a reliability edge over flash.
- As long as the proper burn-in and test has been performed on DRAM, it should last the lifetime of the system it's plugged into. Can't say the same about HDs as I've encountered more drive failures than memory failures. It's why most memory manufacturers will offer a "lifetime" guarantee on their parts. Let me know when you find a hard drive manufacturer that will do the same, and if they don't cost an arm and a leg, I buy them.
- Corsair (and other module builders) will generally cut costs by not doing enough exhaustive testing and/or burn-in. They tend to buy DRAM (or wafers) on the open market at the lowest prices they can negotiate and then assemble them into modules. To keep costs low they may only confirm that the completed module passes some simple tests to make sure it "works". This could be done by plugging in the module into a motherboard and if it POSTs, then that's a good one. They could also use a bench-top module tester that performs a series of functional patterns similar to what memtest86+ can do. However, this will take more time. The old adage "time is money" applies here.
- I've not heard of the ~7 year power-on time for DRAM (or SRAM for that matter) before. I've been testing various types of memories since about 1979 but don't recall ever hearing anyone mention a specific lifetime. Within the last week, I've tested several old Seagate 40GB and 60GB IDE drives and the one with the most power-on hours (POH) was 35,276, which is more than halfway to that 7 year figure. Other POH figures ranged from 1,100 to around 11,000 but I did get one with only 62 hours (data codes from the 10 drives indicated they were all built in 2006).

All that said, I never use standby mode . I always hibernate my systems if I intend on using them again within a day or two. This does place them under more thermal cycles than if I used standby but it saves power (energy) as the laptops are powered off.

[Cigarguy]

From my experience, RAM is very reliable and long lasting compared to other components.

On my desktops, HDD, MB caps and power supplies tend to fail first before anything else. All of which is more concerning and expensive than RAM.

On my laptops, HDD and abuse are the biggest culprit. Very rarely does memory fail on me.

HDD failure is more worrisome in that it affects data. Memory failure usually don't take any data with it.

Regardless when it comes to suspend, hibernate or turn off, the only concern is what is most convenient for me. Life is to short to worry about such things. In my experience, obsolescent rather than hardware failure is more the doom of my computer. How often is HDD, RAM and CPU are upgraded? Upgrades are usually not due to failure but mainly is due to larger capacity and more speed.

[Summilux]

Corsair RAM is only noteworthy because it is so UNreliable and has so many problems!
Check Google for "corsair ram problems" and come up with About 41,900,000 results

Give me Samsung, Hynix, Axiom, G.SKill, Micron, Elpida, Mushkin to mention but a few...

Corsair doesn't produce its own chips, AFAIK. It just buys and tweaks them.
Never had any issue with their products, either with DDR DIMMs or the the ValueSelect DDR3 SO-DIMMs that I currently run in my X220. And the brand isn't exactly known as being cheap or unreliable where I am. So... I continue to count them amongst my preferred manufacturers.

[Medessec]

Regardless when it comes to suspend, hibernate or turn off, the only concern is what is most convenient for me. Life is to short to worry about such things. In my experience, obsolescent rather than hardware failure is more the doom of my computer. How often is HDD, RAM and CPU are upgraded? Upgrades are usually not due to failure but mainly is due to larger capacity and more speed.

Well... erg. In my case, upgrading any of them is more or less *not needed*, because installing a bigger HDD would require re-installing or cloning your windows installation, usually I buy a lot of RAM to begin with, so years pass before I need more. CPU and graphics I think would be the most critical for upgrades after amounts of time, because I do 3D Animation/Modeling, Video Editing, and Gaming. With most laptops, it doesn't really get done anyways due to lack of options. I think this factor is highly dependent on what you use your laptop for.

HDDs by far fail the most in my experience and end up being the thing I have to replace most of the time. My Sager currently has a 160GB SSD, a 750GB Seagate 7200RPM, and a WDC 1TB 5400RPM. It used to have a Seagate Momentus XT 500GB(the Hybrid drive. Not worth buying, btw) as the boot drive, and a normal 500GB Seagate for extra data storage. Both those have since failed. Most HDDs fail from fall damage, or bad sectors from being shaken around.

[jyc]

Thank you all for the information! I am really amazed by the amount of expertise in this forum.

Responding to ajkula66, the laptop in question is a T430. Unfortunately the older ThinkPad seems to have broken for some reason I have not (yet) had the time to diagnose.

Will definitely be checking back often for whatever other interesting threads there might be.

[ajkula66]

Responding to ajkula66, the laptop in question is a T430.

I'd venture a guess that these are safe to suspend, although I've never owned one.

Good luck.

[RealBlackStuff]

Corsair doesn't produce its own chips, AFAIK. It just buys and tweaks them.
.....
And the brand isn't exactly known as being cheap or unreliable where I am.
So... I continue to count them amongst my preferred manufacturers.

If Corsair doesn't produce chips, how could they be a manufacturer?
They are just a cheap assembly outfit with a pretty nameshield.
Their power supplies, RAM and whatever else they make are junk!

BTW, France has never been known for cheap prices for electronics parts!

[Medessec]

I thought Corsair was a pretty revered name in PSUs... my friend has a CX600 in his rig. I personally use Antecs when I build desktops for my friends, but... eh. I don't know anything anymore.

[Summilux]

If Corsair doesn't produce chips, how could they be a manufacturer?
They are just a cheap assembly outfit with a pretty nameshield.
Their power supplies, RAM and whatever else they make are junk!

BTW, France has never been known for cheap prices for electronics parts!

Corsair may not produce RAM chips but they make RAM modules, so they count as a manufacturer.

Can't comment about other products such as PSUs, waterblocks, etc. as I've never had them. But I'd buy them without prejudices regarding their reliability. Corsair was already a trusted brand when G.Skill was barely known outside ROC, and I am not aware of a massive drop in quality.

[rkawakami]

Corsair may not produce RAM chips but they make RAM modules, so they count as a manufacturer.

In the DRAM module market, a manufacturer is traditionally a company which has the silicon fabs that actually make the individual RAM components AND sells modules, while a builder or assembly house takes parts from those memory manufacturers and puts them onto a module (or in some cases, subcontracts the work to other companies). Note that those assembly houses purchase packaged parts, tested wafers or known good die (KGD) from any number of sources. They can either design their own printed circuit boards or buy them from other sources.

Examples of manufacturers are Micron and Samsung (full disclosure: I'm currently a Samsung employee). Assemblers are Kingston, Crucial, Viking and Corsair. Crucial is kind of a unique company as they are a subsidiary of Micron but also use other memory manufacturers in some of their modules.

[Summilux]

In the DRAM module market, a manufacturer is traditionally a company which has the silicon fabs that actually make the individual RAM components AND sells modules, while a builder or assembly house takes parts from those memory manufacturers and puts them onto a module (or in some cases, subcontracts the work to other companies). Note that those assembly houses purchase packaged parts, tested wafers or known good die (KGD) from any number of sources. They can either design their own printed circuit boards or buy them from other sources.

Examples of manufacturers are Micron and Samsung (full disclosure: I'm currently a Samsung employee). Assemblers are Kingston, Crucial, Viking and Corsair. Crucial is kind of a unique company as they are a subsidiary of Micron but also use other memory manufacturers in some of their modules.

Thank you for making the lexical distinction and briefing us over the DRAM industry, Ray.