[SOLVED] How does age and sustained high temperatures affect the performance of a CPU/GPU?

[ShangWang]

How does heat damage a CPU/GPU overtime? What changes in functionality occur when a CPU/GPU become "old" or "worn" out?

I'm not talking about thermal throttling, but if a CPU does age/has sustained high temperatures over the years, does that affect when the threshold for throttle in the future?

Will high temperatures overtime actually cause the CPU/GPU to drop in performance permanently and how?

When people refer to "wear" or "damage" to a CPU, they're only referring to how capable the CPU is of overclocking? If this is the only thing that affects a CPU from age/high temperature sustain, how does heat/age affect overclocking?

Regular frequencies/max turbo frequencies/performance are not affected by this whatsoever as a CPU/GPU ages/is stressed by heat?

I'm assuming a CPU/GPU can only "die out" from age, and not actually suffer dramatically from wear and high temperatures unless it was above 100c or something like that?

 

[JohnBonhamsGhost]

modern CPUs will not reach harmful temperatures.
they will either throttle or shutdown completely before this can happen.

it may be possible in a customized scenario to run at >110°C for some time but the damage to the CPU would be impossible to predetermine.
some may just burn out and die, others may start experiencing issues processing, physical damage could occur and damage certain built-in controllers, etc.

GPUs would be similar but i've actually seen some of these that were physically damaged by high temperatures and high voltage for extended amounts of time due to high overclocking.
some that just crashed and never functioned again, others that would start showing artifacts and color corruption, a couple that functioned fine during minimal use but as soon as the clocks ramped up it would show artifacts and/or cash.

i'm sure you could find some articles or maybe videos of experiments into this realm where users are intentionally damaging hardware with heat and extreme usage.

 

[ShangWang]

i'm sure you could find some articles or maybe videos of experiments into this realm where users are intentionally damaging hardware with heat and extreme usage.

Yep, seen the one that shows a CPU without a fan was pretty interesting.

What I'm asking is why people say heat causes more "wear" on a CPU, I'm wondering what wear/age refers to and how it affects performance over the years.

I got advice before from someone who said it will affect how much you can overclock CPUs, but never understood why that would happen from heat stress/age. Is that the only thing that can happen, and can you explain why?

The other other thing a CPU can be affected by age/heat stress is simply dying out which would probably take over 10 years even when used at 80-90 degrees constantly?

 

[JohnBonhamsGhost]

why people say heat causes more "wear" on a CPU, I'm wondering what wear/age refers to and how it affects performance over the years.

it would be a case by case basis in the affect it may have on a particular processor.

some would show zero symptoms of damage,
some would not accept as high clock rates anymore,
some would cause errors and/or corruption of processed data,
some would just die permanently...

there are multiple types of damage that could occur.
there is no way to determine what exact damage that may be.

these components are manufactured out of multiple conglomerate elements.
each element may be affected by temperature in a different way each time they are exposed.
so where one chip may sustain damage to it's onboard memory controller another may not have proper connection with it's pin contact plates anymore.

you'd never know for sure what the end result may be until you put each chip through similar torture.

 

[ShangWang]

you'd never know for sure what the end result may be until you put each chip through similar torture.

Thank you, I have never seen anyone who has had issues from an old CPU so I was just curious. In your own experience, is the most likely thing that will happen is just overclock capabilities?

Generally normal frequency functions like max turbo frequencies aren't usually affected, and you would for the most part just see a CPU that has suffered from age/heat when it completely stops working?

 

[drea.drechsler]

How does heat damage a CPU/GPU overtime? What changes in functionality occur when a CPU/GPU become "old" or "worn" out?

I'm not talking about thermal throttling, but if a CPU does age/has sustained high temperatures over the years, does that affect when the threshold for throttle in the future?

Will high temperatures overtime actually cause the CPU/GPU to drop in performance permanently and how?

When people refer to "wear" or "damage" to a CPU, they're only referring to how capable the CPU is of overclocking? If this is the only thing that affects a CPU from age/high temperature sustain, how does heat/age affect overclocking?

Regular frequencies/max turbo frequencies/performance are not affected by this whatsoever as a CPU/GPU ages/is stressed by heat?

I'm assuming a CPU/GPU can only "die out" from age, and not actually suffer dramatically from wear and high temperatures unless it was above 100c or something like that?

Heat is the destroyer of all semiconductors. One reason that especially affects extreme large scale IC's with very small geometries is an effect called electromigration...that's where electrons rushing through the conductors in the IC actually 'push' the conductor's atoms into the surrounding silicon. It occurs all the time but the rate is much faster with higher temperatures and current through the conductors.

The effect of pushing it's atoms out is it increases the resistance of the conductors. As resistance increases the internal voltage drops will also increase, i.e., the internal voltage decreases even though the voltage from the VRM remains the same, and it becomes unstable. This is called degradation.

The way to 'fix' that is you have to increase operating voltage so it returns to stability. The side effect is that increases temperature so it degrades even faster now so you have to increase voltage again...and sooner...which increases operating temp even more. And so on.

Assuming you can keep cooling it as you increase voltage to return stability you might come upon the dielectric limit of the silicon. That defines the highest voltage the manufacturing process allows without basically arcing through the insulators with immediate death resulting. That's a quick death, very much avoidable and not really heat related. Electromigration is the slow death that comes with operating at really high temperatures for extended periods.

The reason modern CPU's and GPU's don't really degrade at the high temperatures they seem to operate at is they don't do it constantly. They only run at high clocks with high temperature for very short duration on one core, literally milli-seconds at a time, then run the clocks down quickly and move the process to another core. That way the high temp loads move between cores so any one isn't being degraded. But we see high temp spikes and assume that's the constant temp on all the cores. But it's not: you have to find average temps across all the cores to get a true picture.

The other way to return stability is simply lower the operating frequency. Ryzen's boost algorithm does that automatically so if temperatures are running hotter than normal it will naturally tend to boost less aggressively with degraded performance especially in heavier all-core processes. This might be the result when one is returned to full-stock settings after an aggressive all-core, fixed overclock when run at extremely high temperatures for long periods since it can never lower voltage and clocks to reduce temp on a core as it gets hot.