True hardware health.

[TrueWisdom]

I'm getting ready to deploy a home media server, which I configured to be as low-power as possible, so that it could feasibly be run 24/7 and use as much power as a light bulb. However, I've been doing some reading on Wake-On LAN, and it seems like this would be a much better option for energy conservation in my particular scenario. But that begged the question--what is truly better for the computer?

How much does starting-up and shutting-down (or entering/resuming from standby) reduce the lifespan of the average computer? I know that hard drives are hit the hardest by this, since they have the most moving parts, but to what extent are they harmed? In the end, the difference between letting the computer enter standby regularly, and leaving it on and idle 24/7, is probably only a few dollars a month in electricity bills. Is there any advantage to leaving the computer on all the time? Or am I better served with WOL, with the server in standby at any time it's not being used?

 

[Matthias99]

Being in 'standby' mode just puts the CPU in a low-power state and stops running the OS. Usually the hard drives are also spun down. Full startup/shutdown may be harder on the power supply's internal components, although if you're targeting 60W total power it's not like your PSU will be stressed. Modern CPUs support dynamic clock rate and/or voltage changes, so you can run in a low-power mode without fully shutting down when the system is idle. In my experience 'wake-on-X' is not always 100% reliable.

Spinup is tougher on hard drive motors than steady-state operation -- but the bearings or motor will eventually wear out if you run it long enough. There should be a crossover point beyond which it's better to save the steady-state wear despite having to go through a spinup when you come back online. My general take on it is if you expect the drives to be idle for more than a couple hours, it's probably better to spin down and save wear on the motor/bearings. However, I am not aware of any large-scale studies on this kind of thing.

 

[CTho9305]

How long does the computer really need to last?

 

[BrownTown]

yeah, my computers have lasted as long as I've ever needed. The only things that have broken are several CD rom drives and a battery on my laptop died.

 

[alimoalem]

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

 

[Special K]

Originally posted by: alimoalem

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

On a related note, all of the academic literature I have read on CPU reliability says that 30 years is the target lifetime for modern processors. Why would manufacturers need their chips to last even half of that time? Who is still going to be using the same chip in 30 or even 15 years?

 

[Fallen Kell]

Originally posted by: Special K

Originally posted by: alimoalem

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

On a related note, all of the academic literature I have read on CPU reliability says that 30 years is the target lifetime for modern processors. Why would manufacturers need their chips to last even half of that time? Who is still going to be using the same chip in 30 or even 15 years?

You would be surprised to the answer of this. Many business may purchase infrastructure systems that more then meet their needs and use them until they finally die. Case in point, why upgrade a DNS server with faster CPU's when the 15 year old SPARC 5 barely uses 15% of its CPU power to serve your site? DNS is not the only thing like this, many authentication services like NIS, NIS+, LDAP etc, are not very CPU demanding as they depend directly to how many client systems that handle on an as needed system (i.e. rarely will an entire site's worth of computers need to access these services at the exact same time). So old hardware is more then capable for these roles.

 

[CTho9305]

Originally posted by: Special K

Originally posted by: alimoalem

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

On a related note, all of the academic literature I have read on CPU reliability says that 30 years is the target lifetime for modern processors. Why would manufacturers need their chips to last even half of that time? Who is still going to be using the same chip in 30 or even 15 years?

If your product might be used in embedded systems, I would imagine you want it to have a very long lifetime.

 

[jagec]

The best way to increase your computer's lifespan is to add better cooling. The hard drive is pretty much the only thing that really hates power on/power off cycles (well, and the fans I suppose), but EVERYTHING in a computer hates heat.

 

[DrPizza]

Originally posted by: Special K

Originally posted by: alimoalem

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

On a related note, all of the academic literature I have read on CPU reliability says that 30 years is the target lifetime for modern processors. Why would manufacturers need their chips to last even half of that time? Who is still going to be using the same chip in 30 or even 15 years?

Uhhh, have they ever upgraded the computers in the space shuttle?

 

[Special K]

Originally posted by: DrPizza

Originally posted by: Special K

Originally posted by: alimoalem

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

On a related note, all of the academic literature I have read on CPU reliability says that 30 years is the target lifetime for modern processors. Why would manufacturers need their chips to last even half of that time? Who is still going to be using the same chip in 30 or even 15 years?

Uhhh, have they ever upgraded the computers in the space shuttle?

Good point, although they benefit from using old chips in the space shuttle because larger transistors are less vulnerable to soft errors caused by alpha and neutron particle strikes, which can be a frequent occurence in space.

 

[imported_Seer]

While I do not know precisely what the effect of on/off cycles is, I will say that my dad has servers in his office that have been running for 10 years basically non stop, and most work fine.

 

[aj654987]

You would have to consider the 30 year target goal to be an average as well. Some chips will fail before 30 years some will last beyond. Also its just a design goal and a prediction, they dont really KNOW for a fact that they will last 30 years. I think they are just trying to error on the side of caution and have a high factor of safety. Finally the economics would have to be considered. It may be that a 30 year design would be nearly as economic as a 15 year design.

There are a lot of cases of chips nearly 30 years old such as traffic light controllers, the shuttle as mentioned, even the guys with their old hp calculators from the 70's.

 

[Nathelion]

one consideration is the difference between ambient temperature and the runtime temperature of your system. Every time the components (particularly the chips) heat up/cool down there is thermal expansion and retraction (or whatever the opposite of expansion is) going on. That will lower the lifetime. So if you have a cold room, running continuously might be better. In a warm room, not such a big deal.

This is mostly hearsay though. I've never actually seen a rigorous study concerning the effect of temperature changes on chip lifetime.

 

[Zorba]

Originally posted by: DrPizza

Originally posted by: Special K

Originally posted by: alimoalem

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

On a related note, all of the academic literature I have read on CPU reliability says that 30 years is the target lifetime for modern processors. Why would manufacturers need their chips to last even half of that time? Who is still going to be using the same chip in 30 or even 15 years?

Uhhh, have they ever upgraded the computers in the space shuttle?

Yeah, they updated them a few years back. But they still upgraded them with old a$$ stuff .

 

[mindless1]

The effect of staying on or going into standby is negligible.

Fans, capacitors and hard drives will "usually" have the shortest lifespans.

Fans - low RPM, quality brand, dual ball bearing if near heat or not vertically oriented.

Capacitors - too much to post at once. Quality brand parts, ran conservatively (as you're doing when underclocking and undervolting but not always when building low power PC, because some would just have (motherboard for example) parts engineeded for low power CPU, engineered less robustly with the expectation of the lower currents used (saves them money).

Hard drive - keep it cool enough, nothing extreme is required. Stable, clean power. Beyond that it's a lottery, nothing you do will make as much difference as whether the particular specimen you have, had random variations in any one part that led to an early demise. In short, don't think long term with a hard drive, make periodic backups and replace it every 3-4 years.

 

[IEC]

Originally posted by: jagec
The best way to increase your computer's lifespan is to add better cooling. The hard drive is pretty much the only thing that really hates power on/power off cycles (well, and the fans I suppose), but EVERYTHING in a computer hates heat.

Add better cooling, run it 100% of the time. While you're at it, FOLD! (http://folding.stanford.edu)

 

[nobb]

I am curious too. I have an old machine that I have setup to run 24/7 for surveillance purposes, but I am mostly concerned for the hard drive and how that is holding up to being run 24/7. I dont think there really is a definitive answer on how long your particular hardware will last as everyone's results will vary. But in general, more cooling is good. Chances are, the harddrive will probably be the first to fail, then maybe your fans. All the solid state stuff like your motherboard will probably last longer than you will need it to. However, the electrolytic capacitors may fail. They evaporate faster as temperature increases.

 

[mindless1]

If you want max life out of the system do not FOLD or run any other continuous high load app. There is a clear correlation between this load and generated heat which we know substantially reduces parts' lifespans, especially the capacitors.

 

[Abram730]

The biggest degradations as I understand are from capacitors so getting components with high grade capacitors will extend the life of a computer. Also fans running according to temperatures helps. A system that can run fans after the computer is shut down will last longer. I have seen 20% longer life claimed before.
I thought of a possible issue with PSU's. PSU's often have issues with current draws below 20% and over 80%. So going green could be an issue for me as a gamer. I'd love to know the state of new PSU's. With graphics cards using more power and with the desire to not use more power then needed. Could I have issues with my computer say using very low watts normal and high watts loaded? My guess is this could cause some issue for some gamers, but this is an uneducated guess. I guess what I wonder is if stable voltage with a wide difference in power needs from the same PSU has been considered.

 

[mindless1]

Originally posted by: Abram730
The biggest degradations as I understand are from capacitors so getting components with high grade capacitors will extend the life of a computer. Also fans running according to temperatures helps.

False, that only reduces noise. Since the fans have to spin fast enough at higher heat levels either way, reducing fan speed at lower heat levels just means the system runs a little hotter at lower loads than if they were fixed speed fans. However, it is a reasonable compromise to make if you value lower noise levels.

A system that can run fans after the computer is shut down will last longer. I have seen 20% longer life claimed before.

False, it is a marketing department white lie that was spread as an urban myth. The system ran at "X" temp for quite a while and once turned off all heat production besides a trival level in PSU 5VSB circuit is gone. Thus all parts cool down below the former temp at which they ran, quickening that has no useful purpose as it is a negligable difference if a part temp takes 30 seconds or 4 minutes to practically reach room temp because it is such a short period of time either way and it is always getting ever cooler than it's prior running state. It could actually be worse to continue actively cooling parts no longer producing heat in that it is a more rapid thermal, contraction change.

I thought of a possible issue with PSU's. PSU's often have issues with current draws below 20% and over 80%. So going green could be an issue for me as a gamer. I'd love to know the state of new PSU's. With graphics cards using more power and with the desire to not use more power then needed. Could I have issues with my computer say using very low watts normal and high watts loaded? My guess is this could cause some issue for some gamers, but this is an uneducated guess. I guess what I wonder is if stable voltage with a wide difference in power needs from the same PSU has been considered.

With systems using a lot of power from active devices such as GPU or CPU that have large changes in stately current consumption, having a larger PSU transformer for rapid recovery times, and quality output subcircuit filter caps to survive the larger ripple will help promote longer life and stability. It is a bit beside the point though that these things are inherant in quality, accurately rated PSU rather than the budgetized generics which tend to cut corners on parts like capacitors, and tend to overoptimistically rate output current capability.

 

[Rastus]

Originally posted by: Zorba

Originally posted by: DrPizza

Originally posted by: Special K

Originally posted by: alimoalem

Originally posted by: CTho9305
How long does the computer really need to last?

if you want it to go for 5 years, you should be fine

On a related note, all of the academic literature I have read on CPU reliability says that 30 years is the target lifetime for modern processors. Why would manufacturers need their chips to last even half of that time? Who is still going to be using the same chip in 30 or even 15 years?

Uhhh, have they ever upgraded the computers in the space shuttle?

Yeah, they updated them a few years back. But they still upgraded them with old a$$ stuff .

It takes years and lots of money to qualify new hardware to retrofit old hardware that meets rigorous specifications. Some obsolete processors stay in production long after the commercial market dries up to meet contractual obligations.

I've seen 1971 computers still in operation in the late 1990's.

 

[mindless1]

NASA does not play Quake 7. The frame of reference is entirely different. There must be two priorities, #1 being extremely robust, long lived parts. #2 must be low power. Towards these ends, a high IPC and lower current processors are good choices. Remember that gaining some performance that you don't need, then having to spend $2 million to replace something is crazy.

Given today's technology, you can build a <= 80486 class system that will last more than twice as long as modern hardware will unless the size and research increased by an order of magnitude. Perhaps we humble PC users would benefit from NASA's research in this area but they would not until they need higher performance levels.