Why is thermal throttling a bad thing?

[Kucing]

Why is thermal throttling a bad thing? My CPU can run an all-core boost at 5ghz for a few seconds no problem, and it starts to thermal throttle after that. It seems like the universal piece of advice would be to decrease the turbo ratio limits to prevent that it never goes that high to begin with, but why? Why would I cap its performance in both short and long loads when it can cap its performance with thermal throttling in long loads and run much faster in short loads?

Edit: I have also been decreasing the prochot to 90c so it's not a degradation issue

 

[AleXXX666]

it's really personal preference - probably you won't use your cpu for example for ~10 years, so even this is "a bad" thing, like silicon could degrade and don't work well - it's up to you.

 

[kondamin]

Why is thermal throttling a bad thing? My CPU can run an all-core boost at 5ghz for a few seconds no problem, and it starts to thermal throttle after that. It seems like the universal piece of advice would be to decrease the turbo ratio limits to prevent that it never goes that high to begin with, but why? Why would I cap its performance in both short and long loads when it can cap its performance with thermal throttling in long loads and run much faster in short loads?

Why don't you just get adequate cooling so you don't throttle?

 

[Kucing]

Why don't you just get adequate cooling so you don't throttle?

Because it's a laptop

 

[AleXXX666]

Because it's a laptop

wow, that a different story! I remember back days when Apple used Intel chips - their temps there were like pretty ALWAYS 100 C. Well, the temp limits for laptop chips are higher than for desktop cpus. But, greedy laptop makers usually make crap cooling solutions so thermal throttling happens...
I'm also "wannabe gaming" laptop user currently, and I DON'T LIKE HEADPHONES. At all. So I limit this stupid thing power a bit - so the fans don't spin like a turbo in Subaru. I'm satisfied with FPS/other performance (well, except benchmarks LOL), so..

 

[Kucing]

wow, that a different story! I remember back days when Apple used Intel chips - their temps there were like pretty ALWAYS 100 C. Well, the temp limits for laptop chips are higher than for desktop cpus. But, greedy laptop makers usually make crap cooling solutions so thermal throttling happens...
I'm also "wannabe gaming" laptop user currently, and I DON'T LIKE HEADPHONES. At all. So I limit this stupid thing power a bit - so the fans don't spin like a turbo in Subaru. I'm satisfied with FPS/other performance (well, except benchmarks LOL), so..

I know it's a very unpopular opinion but I love fan noise lol, kind of like how car guys like loud engines I guess. It's just good to know my CPU is working as cool as it can be.

 

[AusWolf]

These days, it's not a bad thing, more like personal preference.

AMD CPUs are designed to run at 95 ˚C, Intel at 100 ˚C, unless your oversized cooling or settings keep it under during max boost. Some people assume that such temperatures are a safety limit, and the CPU will die if it runs so hot. That held true 20, or maybe even 10 years ago, but not anymore. Such temperatures are designated as the highest "safe operating temperatures", not a thermal shut-off point or something. Also, these modern CPUs include lots of safety features that old ones didn't. Your laptop may run a bit slower if it reaches the thermal limit, but it'll never burn itself down (unless for other reasons, like an electrical fault).

So basically, people who used to overclock their Athlon XPs and Pentium IIIs to the moon and back tend to think that "thermal throttling is bad". Otherwise, running your hardware as cool as possible is a good thing, but isn't strictly necessary.

 

[Chrispy_]

Boost throttling is by design. It's nothing to worry about.

If a laptop is given a 100W cooling solution, that means it can average 100W of heat dissipation but your CPU can boost so that it generates more heat than this temporarily, using the thermal mass of the silicon, heatpipes, and fin array to soak up the extra heat temporarily. Once it hits the maximum safe temperature of the silicon, it'll slow down the clockspeed and reduce power to ensure that the cooling can keep it at that maximum safe temperature (typically 95C to 105C depending on the specific CPU model).

Generally, boosting to maximum temperature is a great solution for laptops that enables them to perform temporarily faster than the limitations of a slim/small laptop cooling system will normally allow. You can't change the cooler to a better one in a laptop, so boosting to the throttle point is one way they can get you the (temporary) performance of a much thicker and heavier laptop with a massive cooling system.

If your CPU is always thermal throttling, that's a different thing entirely - that implies that your cooler isn't doing its job (poor contact, dried up thermal paste, vents clogged with dust, fans broken etc.

 

[Jacky_BEL]

It is not a bad thing , it is one of the mechanisms to protect components from damage while pushing the performance to the limit , even when the cooling capacity degrades over time
The components must be well designed for this , and then the parameters for this throtling must be well chosen.

 

[unclewebb]

Why is thermal throttling a bad thing?

I think this myth got started a long time ago. My Core 2 Duo E6400 rated at 2133 MHz could be overclocked to the moon. 3400 MHz was Prime stable. Even 3500 MHz or 3600 MHz was possible with enough voltage. Those were the good old days when overclocking was both fun and productive. What I noticed back then was that Core 2 Duo technology, when heavily overclocked, would start to lose stability at around 70°C to 75°C. Keep it at that temperature or less and it was rock solid stable even when overclocked a crazy amount.

When Core i technology was introduced, it became possible to run fast and hot while still maintaining stability. Intel bumped the thermal throttling temperature up to 100°C and has used that temperature for the vast majority of their Core i CPUs. The 3rd Gen got increased to 105°C but this was reduced back to 100°C when the 4th Gen was released. To me that is a good sign that Intel themselves have done plenty of testing and are quite comfortable that their CPUs can run both reliably and long term at temperatures much higher than most enthusiasts recommend. After 16+ years of Core i CPUs, the "high temperatures are bad" myth continues on some forums.

Intel increased the thermal throttling temperature up to 110°C for the Core Ultra 9 185H.

Intel® Core™ Ultra 9 Processor 185H (24M Cache, up to 5.10 GHz) - Product Specifications | Intel

Intel® Core™ Ultra 9 Processor 185H (24M Cache, up to 5.10 GHz) quick reference with specifications, features, and technologies.

www.intel.com

The Core Ultra 9 275HX is at 105°C.

Intel® Core™ Ultra 9 Processor 275HX (36M Cache, up to 5.40 GHz) - Product Specifications | Intel

Intel® Core™ Ultra 9 Processor 275HX (36M Cache, up to 5.40 GHz) quick reference with specifications, features, and technologies.

www.intel.com

This new HX is unlocked and can be overclocked. Perhaps the Intel thermal engineers decided to give the HX a little extra temperature headroom.

I have also been decreasing the prochot to 90c so it's not a degradation issue

You can run your Intel CPU at any temperature you like. I do not think degradation at high temperatures is much of an issue. I know in theory it is but practically speaking, it does not seem to be an issue. Too much voltage is likely a bigger problem when it comes to degradation compared to running an Intel CPU within its temperature spec.

A lot of laptop manufacturers reduce the thermal throttling temperature down to about 95°C. I think they are more concerned about customer's laps then they are concerned about degradation. About the only thing that is likely to degrade at a laptop CPU temperature of 100°C is one's ability to reproduce.

 

[Hardcore Games]

Turboboost is a more recent move to get more performance from processors. My Dell is 20W but the boost can take it much higher towards 55W which is about what my CPU propeller can handle. The GPU is another heat source. So I need to be wary about heat.

 

[Lew Zealand]

FWIW, my experience with running laptop-class CPUs at ~100C for the following duty cycles is this:

Core i7-2635QM for about a year at ~50% duty cycle - no damage
Core i7-3720QM for about 2.5 years at ~50% duty cycle - no damage
Core i7-3610QM for about 1 years at ~40% duty cycle - no damage
Core i7-5557U for about 3 years at ~15% duty cycle - electromigration damage noticed after 4+ years when redeploying

The no-damage quad cores are all in Mac Minis, it could be that Apple secured better quality parts somehow while the 5557U is in an Intel NUC. Now why wouldn't Intel have given themselves the best? Well, the only other major customer for the 5557U was: Apple.

Yeah that 3720QM was beaten to a pulp but still works as new today though it doesn't get much use any longer of course. Once I learned about the dangers of electromigration and long-term damage I started running it with turbo off for another year and it would plateau at only ~80C. Later on I finally did the right thing and built a PC with adequate cooling to run 100% CPU at max turbo at 70C.

 

[AleXXX666]

I know it's a very unpopular opinion but I love fan noise lol, kind of like how car guys like loud engines I guess. It's just good to know my CPU is working as cool as it can be.

same here to me. Though I like it more in desktops, on laptops they are good as soon as they just "informs" me that thing "works" lol

 

[Toothless]

I paid for the performance so I want the performance. If I have to run a laptop by all means there will be a method found to undervolt/cooling pad/satanic magic to keep those turbo clocks.

 

[Vayra86]

I paid for the performance so I want the performance. If I have to run a laptop by all means there will be a method found to undervolt/cooling pad/satanic magic to keep those turbo clocks.

So this is your man cave

FWIW, my experience with running laptop-class CPUs at ~100C for the following duty cycles is this:

Core i7-2635QM for about a year at ~50% duty cycle - no damage
Core i7-3720QM for about 2.5 years at ~50% duty cycle - no damage
Core i7-3610QM for about 1 years at ~40% duty cycle - no damage
Core i7-5557U for about 3 years at ~15% duty cycle - electromigration damage noticed after 4+ years when redeploying

The no-damage quad cores are all in Mac Minis, it could be that Apple secured better quality parts somehow while the 5557U is in an Intel NUC. Now why wouldn't Intel have given themselves the best? Well, the only other major customer for the 5557U was: Apple.

Yeah that 3720QM was beaten to a pulp but still works as new today though it doesn't get much use any longer of course. Once I learned about the dangers of electromigration and long-term damage I started running it with turbo off for another year and it would plateau at only ~80C. Later on I finally did the right thing and built a PC with adequate cooling to run 100% CPU at max turbo at 70C.

Different node & different architecture, right?

 

[Apocalypsee]

Thermal throttling isn't a bad thing, the only bad thing about it is most laptop manufacturer usually has one-size-fits-all cooling solution that fits both higher-end and lower-end config, where the higher-end config usually suffers because the cooling needs to cope with both CPU and GPU heat, usually throttles too much and hamper performance. I also don't like to have vary performance throughout my intended usage; I get good performance when the laptop is cool but then it started to slow down when it heated up especially gaming. End user usually doesn't get to control this, and that's why utility like MSI Afterburner and Throttlestop is popular.

 

[Lew Zealand]

Different node & different architecture, right?

Yup.

2635 - 32nm Sandy Bridge
3610, 3720 - 22nm Ivy Bridge
5557 - 14nm Broadwell

Early 14nm was problematic for Intel and Broadwell was a narrow launch with few CPUs. But then Ivy Bridge was also the first 22nm so you win some, you lose some.

 

[eidairaman1]

Why is thermal throttling a bad thing? My CPU can run an all-core boost at 5ghz for a few seconds no problem, and it starts to thermal throttle after that. It seems like the universal piece of advice would be to decrease the turbo ratio limits to prevent that it never goes that high to begin with, but why? Why would I cap its performance in both short and long loads when it can cap its performance with thermal throttling in long loads and run much faster in short loads?

Edit: I have also been decreasing the prochot to 90c so it's not a degradation issue

Yet no system specs