I was wondering if CPU's degrade over time, particularly because of heat. MY IBM notebook is always on, and I am wondering if this has any negative affects on the performance of the CPU in the long run.
Do CPU's age?
- Thread starter iregula
- Start date
My CPU test score is still the same after 3 years
If you don't torture them with intense heat and voltage, the performance should be the same
Chemicals they use to make CPU aren't reactive at all. They don't rust, they don't crack
If you don't torture them with intense heat and voltage, the performance should be the same
Chemicals they use to make CPU aren't reactive at all. They don't rust, they don't crack
Jeff7
Lifer
- Jan 4, 2001
- 41,596
- 20
- 81
It won't slow down over time, however, after a number of years (like probably 10+) electromigration might become a problem - as I understand it, it's sort of like erosion caused by the flow of electrons. Eventually the processor shorts itself out internally and just stops working. But that takes a long time, even when the processor is fed extra voltage for overclocking.
I would say no. Ask this in the DC forum, where folks have machines up at 100% CPU usage for years.
Nope, not in the slightest. If CPU's did slow down with age, that means the CPU's in old consoles such as the NES would be running slower than they should be, causing many games not to work, and Mario to run around in slow mo like an old man.
I once (long time ago) heard an explanation that cpu is a single-state machine, and it stays the same always, i.e never slows down or anything unless you physicaly break it.
only a few years later during my CS study I realized that idiot had no idea what he was talking about, since a ssm (single state machine), is indeed a cpu type, but has nothing to do with its durability... heh who was the bigger idiot, me for beliving, or him for saying that crap... go figure...
Anyway, my guess, is that YES, eventually sh!t wears down. I don't know exactly how to measure fatigue, and slowness, but if you follow basic rules of physics, you'll understand, it must wear down (unless we're talking about a cpu kept in a safe box...)
only a few years later during my CS study I realized that idiot had no idea what he was talking about, since a ssm (single state machine), is indeed a cpu type, but has nothing to do with its durability... heh who was the bigger idiot, me for beliving, or him for saying that crap... go figure...
Anyway, my guess, is that YES, eventually sh!t wears down. I don't know exactly how to measure fatigue, and slowness, but if you follow basic rules of physics, you'll understand, it must wear down (unless we're talking about a cpu kept in a safe box...)
CPU's do age when in use. Manufacturer's rate them for an operating lifetime - how long depends on the manufacturer. If used continuously, eventually they will fail. Common mean time to fail ratings are on the order of 100,000 hours. 100k hours is over 10 years, so as far as the original question about turning off a laptop, I wouldn't worry about it.
They do not slow down in so far as performance is concerned except in highly unlikely circumstances. In general, they usually work until they fail - at which point, they do not work any more. In other words, a CPU that takes 10 seconds to calculate a spreadsheet in Excel when it's new, will still 10 seconds to calculate a spreadsheet in Excel after 10 years of use.
------------------ Highly technical details ---------------------
In reality, the transistors gradually slow down as well. There are a couple of mechanisms that cause transistors to slow down over time: PMOS NBTI (Negative Bias Temperature Instability), and nMOS hot-e. A computer user doesn't notice this effect, however, because CPU's are digital synchronous devices. In other words, this effect of transistors slowing down over time can be considered to be "noise" that is ignored by the CPU because it has a clock and it's a digital device.
These two effects together are one of the reasons why overclocking works - manufacturers set a limit on the frequency that is less than the part is ever likely to slow down to within a given time period. So a CPU might come out of the fab at 2GHz and then after a year work at 1.98GHz, and after another year be limited to 1.96GHz, etc. Despite the fact that the CPU works fine at 2GHz at the time it's tested, it's "guardbanded" to a lower frequency so that it works correctly 10 years down the road. That guardband might, picking a random number that's easy to work with, be 10%. So rather than sell the CPU at 2GHz, it is marked and sold as 1.8GHz so that in 10 years it is still working. Higher voltages accelerate the two effects, and low temperatures accelerate hot-e while high temperatures accelerate NBTI.
This subject is one that is fairly esoteric - any references that I might provide will be to fairly unreadable electrical engineering papers. If you want more information, googling "hot carrier" "hot electron" "NBTI" "reliability" can find some sites with information (as well as, I note with some amusement, my postings here at Anandtech).
One good site that Google turned up for hot-e is: http://www.semiconfareast.com/hotcarriers.htm
They do not slow down in so far as performance is concerned except in highly unlikely circumstances. In general, they usually work until they fail - at which point, they do not work any more. In other words, a CPU that takes 10 seconds to calculate a spreadsheet in Excel when it's new, will still 10 seconds to calculate a spreadsheet in Excel after 10 years of use.
------------------ Highly technical details ---------------------
In reality, the transistors gradually slow down as well. There are a couple of mechanisms that cause transistors to slow down over time: PMOS NBTI (Negative Bias Temperature Instability), and nMOS hot-e. A computer user doesn't notice this effect, however, because CPU's are digital synchronous devices. In other words, this effect of transistors slowing down over time can be considered to be "noise" that is ignored by the CPU because it has a clock and it's a digital device.
These two effects together are one of the reasons why overclocking works - manufacturers set a limit on the frequency that is less than the part is ever likely to slow down to within a given time period. So a CPU might come out of the fab at 2GHz and then after a year work at 1.98GHz, and after another year be limited to 1.96GHz, etc. Despite the fact that the CPU works fine at 2GHz at the time it's tested, it's "guardbanded" to a lower frequency so that it works correctly 10 years down the road. That guardband might, picking a random number that's easy to work with, be 10%. So rather than sell the CPU at 2GHz, it is marked and sold as 1.8GHz so that in 10 years it is still working. Higher voltages accelerate the two effects, and low temperatures accelerate hot-e while high temperatures accelerate NBTI.
This subject is one that is fairly esoteric - any references that I might provide will be to fairly unreadable electrical engineering papers. If you want more information, googling "hot carrier" "hot electron" "NBTI" "reliability" can find some sites with information (as well as, I note with some amusement, my postings here at Anandtech).
One good site that Google turned up for hot-e is: http://www.semiconfareast.com/hotcarriers.htm
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter