What should I upgrade on my laptop ?

[cyberfish]

The short version is that a product generally lasts a given average number of hours... how quickly you use them up, is entirely up to you. Having your machine on *all the time* is going to use those hours up --substantially increasing the likelihood of a failure-- probably two or three times as fast as someone who has to common sense to turn things off when they're done with them.

For things like high pitch BGA chips mounted on PCBs (like everything in a modern PC), the thermal stress of turning on and off is much more damaging than having it running all the time. They are just like light bulbs. On/off stress is much higher than constant-on stress.

MTBF only makes sense for things like fans, which are cheap to replace, and last a very long time anyways. I keep all my desktop computers running all the time. With high quality brushless DC fans, they wear out in about 5-7 years. They only cost $4 each.

[zakiuz]

I didn't check on this tread for a while now ! It has gotten pretty big. But in case anyone is interested, I sold my Toshiba and bought an Acer Aspire 5750.

Intel Core i5 2430m 2.4ghz up to 3ghz
4GB of RAM
500GB HDD 5400RPM
Intel HD Integrated 3000

So far I love it, I have 64bits windows 7 and I'll get ubuntu 10.10 64bits as well. The quality is much better than the Toshiba and the battery life goes up to 5 and half hours with only a few tasks !

[MK27]

I only read the last two posts because all I want to replace in my laptop is...my laptop But:

For things like high pitch BGA chips mounted on PCBs (like everything in a modern PC), the thermal stress of turning on and off is much more damaging than having it running all the time. They are just like light bulbs. On/off stress is much higher than constant-on stress.

While no doubt it is true that "on/off stress is much higher than onstant-on stress", my layman-interested-in-science mind tells me that this difference is finite, ie, things like oxidation etc, increase with molecular excitement, aka heat, and that it cannot be infinitely true that leaving something on hot will never do more damage than turning it off once, then turning it on again, duration X later.

I keep all my desktop computers running all the time.

Hmmm. Well honestly I do not believe that leaving a desktop on 5pm - 9am makes the machine last longer, but if it did, is this not about the "sudden stress"? If you are working a machine up until 5pm, then you walk away, it cools down quickly, but not as quickly as if you turned it off. I'd assume the real damage is done at start-up, when you take a cold thing and apply suddenly max current.

Processors work at a fixed voltage etc, so I do not think it would be possible to come up with a software solution to this. But could some kind of hardware be designed so that you could (optionally) slow start/slow stop the system, and thereby reduce the sudden thermal stress?

[cyberfish]

That's true. If you leave it on indefinitely, it will breakdown at some point. Most likely due to electromigration (material shift due to current flow). But that won't happen for a very very long time. On/off shock is so great that X would probably be on the order of years.

When the machine is on, loading and unloading does induce thermal stress, too, just not as great as on/off.

An example of this is with a batch NVIDIA mobile GPUs a few years ago - Why nvidia's chips are defective- The Inquirer

I am not aware of something that can slow start a digital circuit without causing problems. It's easy to do a slow voltage ramp-up, but digital circuits will do calculations wrong at too low voltage. That's why they usually have a power-on reset circuit that holds the chip in reset until voltage raises to a certain level.

Either way, when computers "break", it's usually not electronics that break, but a motor. Most often fan motor (which causes overheat), and harddrive motors. And in the past few years, blown low quality capacitors (there's a huge scandal around that, but that's probably history now). Circuit failures, while not unheard of, is comparatively very rare. And for motors, as you probably know, on/off stress is much higher than operating stress. It is the same for capacitors (surge current on startup), which are the most common failure points for circuits.