[Valmar]

> If you want really long battery life and no heat at all, you can downclock all the way to something like 200-400MHz. A recent CPU at that speed is actually quite usable for things like text editing and reading documentation.

Linux does grant the user that flexibility, so if someone actually wants that, they can have it.

The max non-boost frequency is usually the sweet spot for performance and efficiency.

> On the other hand, if it's plugged in much of the time, then let it boost as much as it can, with speed only thermally limited. Otherwise you're not getting the true performance you paid for.

If the user wants to live with a potentially reduced laptop lifespan, sure thing. But it's just not worth it for a laptop, frankly, given their limited thermal cooling capacities. That CPU will degrade over time when run at that level of heat.

[userbinator]

That CPU will degrade over time when run at that level of heat.

If Intel warrants their CPUs to be at TjMax 24/7, it's a good sign that it shouldn't be a problem. I have not heard of overheating killing CPUs since the days when AMD's didn't have any thermal protection[1], and I've cleaned out machines which were heavily clogged with dust and thermally throttling all the time for many years (the service prompted by their owners complaining about their computers being slow.) In one memorable case the push-in heatsink pins must've been originally not fully inserted, since they came out at some point and the heatsink was not even touching the CPU anymore, yet the CPU kept running for years in that state.

[1] There's a famous TomsHardware video about that: https://www.youtube.com/watch?v=y39D4529FM4

[InvaderFizz]

> If Intel warrants their CPUs to be at TjMax 24/7, it's a good sign that it shouldn't be a problem.

It certainly does work. I was working on some manufacturing equipment in 2018. The 2010 release 1st Gen i3 had a centimeter gap between the IHS and the HSF. The Intel HSF thermal compound had not been touched, it was like new. The CPU had run thermal throttled to about 700MHz for eight years, continuously. Properly attached the HSF, the slowness and thermal throttling went away.

[NavinF]

> That CPU will degrade over time

I overvolted and overclocked pretty much every chip I've owned from my 20 year old Athlon 64 to my 0 year old RTX 4090. None of them have degraded. If you watch overclocking livestreams you'll see just how much abuse it takes to get any sort of reaction from silicon.

[coretx]

Same here, if within sensible ranges it won't harm. It's the current that destroys the ( longevity ) of the silicon, not so much the clock or voltage.

[NavinF]

I'd say it's neither. The only failures I've seen in the data center were caused by differential thermal expansion cycles and broken solder balls. Same phenomena that kills game consoles but not ML/mining GPUs that spend all day and night at max power/current and constant temperature.

[coretx]

Until you power cycle them a couple of times... My old AMD GPU even held a record for how "good" the ASIC still was or was not. TL;DR Yes, your GPU wears and becomes slower over time; eventually it brakes. Not because of bad thermal coefficients but because of the current... Even faster so when OC'ing because U=I*R. You can benchmark it yourself. The broken solder balls of the past where attributed to the transition from leaded to lead free solder. If I recall correctly, it was mostly NVIDIA and Apple who suffered from this and only temporarily / 1 or 2 generations.