CPU Life Expectancy

[kc8adu]

Re: CPU Life Expectancy

Yea, I should add that I agree that CPU's don't die of old age (if it wasn't clear by my post)
CPU's die of power spikes and ripple current
But usually the power circuit on the mobo dies before that happens anyway!

i4004; Ya, NASA uses them because their stuff does not require more power
But problem is even if they wanted more processing power it would be difficult (without a specialized design)
The small transistors of current CPU's have a problem in space; they are too suspectible to the noise caused by the sun in the form of electromagnetic radiation
The stuff used in space is 250nm or larger (250nm is AFAIK cutting edge)
What sits in my system right now is a 45nm CPU, quite the difference!
Stuff is space also have other materials to protect it

many of the cpu's spec'ed for space are radiation hardened too.
r&d to do that is costly so just use more of the old,slow but PROVEN tech.

[kc8adu]

Re: CPU Life Expectancy

I'm 6'3" and those damned suits never fit right.
Stand up too fast in one and my testicles get to say hi to my tonsils.

.

made for tiny japanese dudes.
i cant even get into one of those suits.
tried at hamvention one year where some used ones were on display.
if it could have been zipped up it would have split as soon as i moved.
not made for big uglys.

[sLowEnd]

Re: CPU Life Expectancy

I've never seen a CPU run on stock voltage throughout it's life die

In fact, I still have a few working Pentium 1's around

[kikkoman]

Re: CPU Life Expectancy

now if thermal cycles have an impact on a cpu (which i think is a valid point), wouldn't SpeedStep or CnQ at least potentially lead to a reduced life expectancy?
i've noticed that my CPU's internal temp sensors (it's an AMD X2, CnQ via RMClock) register rather big temperature changes when going from idle to full load.
RMclock's monitor shows steps as big as ~3-5°C/sec.

i'm not really surprised by these jumps since my CPU idles at 1.0 ghz@1.10V and goes to 3.1ghz@1.35V under full load (with intermediate steps) but i'm wondering if these hard transitions could noticeably reduce its life expectancy. (to be more precise, if this could kill the CPU faster than just running at full speed.)
also wondering if this could overly stress the VRM (and PSU).

[370forlife]

Re: CPU Life Expectancy

Thats why I turned cool 'n' quiet off of my 5000+, use to run at 1ghz at idle, and get back up to 2.6ghz when it was needed. Now it just runs at 2.6.

[zandrax]

Re: CPU Life Expectancy

@ kikko:
I experienced the same behavior with an old notebook equipped with a Mobile Athlon XP 2800+: temp spiked as the vcore rised, setting the same vcore on all multipliers lead to a slower variation.
I don't know if the thermal diode is correct or is affected by vcore nor if the thermal cycles can noticeably affect life expectancy; however I wouldn't disable C&Q or Speedstep on a notebook since most small heatsinks aren't usually designed to cope with an always loaded processor (dissipation is one aspect where manifacturers cut corners and usually fans are bios controlled: you can't set fan speed) and I wouldn't disable on a desktop either since dissipated power depends linearly on vcore and more heat means higher power draw and faster/noisier fans: main exception is to set a fixed multiplier both for overclocking and underclocking.

Zandrax

[PCBONEZ]

Re: CPU Life Expectancy

now if thermal cycles have an impact on a cpu (which i think is a valid point), wouldn't SpeedStep or CnQ at least potentially lead to a reduced life expectancy?

Thermal cycles have a negative effect.
Heat has a negative effect.

Which one is the bigger negative? - I dunno. That's COMPLICATED.
The duty cycles (how long it's left hot and how hot) vs (how often/far it's thermal cycled) would play a big part in the math for that.

I used to do similar calculations for fuel cell materials and even just bolts used in reactor cores as part on my job. It's quite complicated because even such things as how pure the metals/materials are comes into play. Impurities cause stress cracks in materials during thermal cycles and impurities can be 'acquired' over a components lifetime by physical contact with other materials. [This is one of the problems we had to deal with: http://en.wikipedia.org/wiki/Hydrogen_embrittlement - for us the hydrogen came from random 'free' hydrogen atoms natural in water which is used for coolant.]

Similar situations exist for silicon materials used in IC chips but I've never actually had to do the math for that sort of thing.

The basic idea is you test/analyze every part.
[In a CPU that would include each PN type junction (several per transistor) and each connection to the PN material (the connections that end up as pins outside the package) and any place where materials are bonded/formed onto the substrate.]
You find what's called "the Most Limiting Component" and base the rating for the whole chip on that.
The Most Limiting Components for heat-up, cool down, fast thermal cycles, slow thermal cycles, and prolonged heat exposure may be completely different parts. Also their effects aren't linear. [Heat-up from say 70F to 80F may be 1/10 as bad as heat-up from 80F to 90F.]

So, I dunno a solid answer, just the theory.

I'd have to say though that Speed-Step kicking in like every 1/2 minutes is [in materials theory] a lot worse for the CPU than it kicking in once every 10/20 minutes or so.

.

[gdement]

Re: CPU Life Expectancy

@ kikko:
however I wouldn't disable C&Q or Speedstep on a notebook since most small heatsinks aren't usually designed to cope with an always loaded processor (dissipation is one aspect where manifacturers cut corners and usually fans are bios controlled: you can't set fan speed)

I think I encountered that on my brother's HP laptop. We installed some memory in it, and I immediately had him run memtest86. After about 10 minutes the laptop suddenly turned off. I guess it overheated.

It's odd though, he said that's never happened with the long video rendering jobs he does on his laptop routinely. Seems that would hit the CPU just as hard and those jobs can take hours.

[tazwegion]

Re: CPU Life Expectancy

I've never seen a CPU run on stock voltage throughout it's life die

In fact, I still have a few working Pentium 1's around

Yeah I agree, our firewall system is running on a Cyrix 233MX flawlessly


Heat, Power irregularities & User abuse (overclocking) are the general causes of processor death IMHO