Question on CPU temps

Tony EH

I've been doing quite a bit of encoding lately of DVDs and Blus using Handbrake to turn them into MKVs for digital storage. When Handbrake is running, the CPU temp rockets up to 90 c (yes I know video conversion uses a lot of power) and the fans spin like a propeller on a plane. The temp hovers there until the encode is finished. Then returns back to normal idle temp of about 25 to 30 c.

I can reduce the number of threads used by Handbrake to reduce the temp, but it increases the encode time.

But, my question is this - when encoding is finished, the core CPU temp immediately plummets down to about 30 c. That's an immediate reduction of 60 c in the blink of an eye.

How is that possible? Surely, the temp should gradually reduce over time and not in a split second.

I am using HWiNFO64 to monitor the temps.

thegreenbean

As soon as the task is finished the power draw cuts of like a light switch and the heat will disappear almost as fast.
I use core temp for monitoring temps. I find it fairly accurate when overclocking.

Tony EH

But to drop like a stone sounds odd from a physics POV, no? Temperature should reduce over time, not simply plummet in a second. I can understand it going from 90 c to 80 c to 70 c etc over the course of a minute even, but it loses 60 c literally in a couple of seconds.

Not saying you're wrong. It's just struck me as weird.

Might check out 'Core Temp' and see what happens.

thegreenbean

Tony EH wrote: »

But to drop like a stone sounds odd from a physics POV, no? Temperature should reduce over time, not simply plummet in a second. I can understand it going from 90 c to 80 c to 70 c etc over the course of a minute even, but it loses 60 c literally in a couple of seconds.

Not saying you're wrong. It's just struck me as weird.

Might check out 'Core Temp' and see what happens.

Yep, no harm in a second opinion. Let us know how you got on.

ED E

Forenote: Software is my area, my physics is limited to LC Honors.

The "heat capacity" of a processor is very low, its a tiny piece of silicone topped by a large chunk of alloy (heat spreader) topped by an even larger cooler. As its not a pure element I can't do any napkin maths but lets say the heat spreader has 4x the capacity of the chip. Then for it to raise by 1*C the processor needs to shed enough energy to raise itself by 4*C.

The reason the processor gets hot in the first place is it cant shed it fast enough but once it idles dumping the small "buffer of heat" it holds is childs play.


PS: Electrical engineers pls dont crucify me.

FanadMan

ED E wrote: »

Forenote: Software is my area, my physics is limited to LC Honors.

The "heat capacity" of a processor is very low, its a tiny piece of silicone topped by a large chunk of alloy (heat spreader) topped by an even larger cooler. As its not a pure element I can't do any napkin maths but lets say the heat spreader has 4x the capacity of the chip. Then for it to raise by 1*C the processor needs to shed enough energy to raise itself by 4*C.

The reason the processor gets hot in the first place is it cant shed it fast enough but once it idles dumping the small "buffer of heat" it holds is childs play.


PS: Electrical engineers pls dont crucify me.

That's a nice and simple explanation.....even I can understand it And also HL Physics in LC but that was years ago!

thegreenbean

ED E wrote: »

Forenote: Software is my area, my physics is limited to LC Honors.

The "heat capacity" of a processor is very low, its a tiny piece of silicone topped by a large chunk of alloy (heat spreader) topped by an even larger cooler. As its not a pure element I can't do any napkin maths but lets say the heat spreader has 4x the capacity of the chip. Then for it to raise by 1*C the processor needs to shed enough energy to raise itself by 4*C.

The reason the processor gets hot in the first place is it cant shed it fast enough but once it idles dumping the small "buffer of heat" it holds is childs play.


PS: Electrical engineers pls dont crucify me.

yes, this^^^
i gave a timey wimey explanation

Tony EH

Sorry for the thread necromancy...

But I recently solved my Handbrake/CPU overheat issue. After a number of times where the CPU (Intel i7 4790K) said F you and shut down due to overheating when encoding with handbrake, I tried shutting off the Turbo Boost in the BIOS. Lo and behold, that seemed to do the trick. Now, I am loath to touch anything in the BIOS, but this issue was really bugging me and with the CPU switching off due to heat, I reckoned that was a very serious issue for the lifetime of my chip.

Another way to achieve the result was to set the "Maximum processor state" to 99% in the power options in Windows, thus negating the need to go near the BIOS at all.

This has my machine rendering video in Handbrake at the same encoding time (give or take a minute) as before, without the overheating and the fans spazzing out. My PC purrs along at a very agreeable 65c when encoding now and when using the likes of Photoshop my fans don't keep spinning up either.

Happy days.

Just thought I'd pass on this, in case anyone else is having the same sort of issues.

unkel

I wouldn't turn off turbo boost. It will make CPU intensive tasks like your encoding slower.

You don't mention you have your system overclocked, so I presume you haven't. The problem is your cooling. It ain't working well. What kind of PC is it and what cooling solution do you have? Standard Intel CPU cooler by any chance?

Buying a better cooler will not cost a lot, and it will keep your CPU temps more reasonable (and avoid overheat crashes). It can also be a lot quieter. You didn't spend all that money on your expensive CPU for nothing, did you? No point in limiting its power. Up your cooling game!

TerrorFirmer

If your PC is shooting instantly up to 90c under stress and actually turning off while under load, there's something wrong. That should not happen even with the Intel stock cooler, which is really not terrible. It may not be properly installed, or else have either too much or too little thermal paste applied. A decent 3rd party cooler is a good buy anyway, but it shouldn't be shutting down.

Tony EH

This overheating thing is only really happening with Handbrake. With Turbo off, Handbrake encodes at roughly the same speed (it takes just a few minutes more) but without the chip going to 90 odd degrees and the fans going mental for the duration of the encode.

For everything else, turbo is left on and there's no issues.

Handbrake, for some reason, seems to want to use everything the PC has and more. It's the Nigel Tufnel of my programs.

Even Premiere runs fine with Turbo left on.

It's weird, to say the least.


EDIT:

I've no idea what cooler is in there, whatever Overclockers UK put in. But, the machine idles at around 20-25 degrees and doesn't heat up generally, even with many tasks, or intensive games running. I know when she's reached over 75 degrees (which somestimes can happen for short periods), because the fans will start to become a little louder, but not crazy. When it's over 95+ degrees (with Handbrake), it's like she's going to take off.

ED E

Its comparable to say your tyres come off at 120kph, but you only drive at 100kph. You might be getting away with it, but its not a good idea.

When we regulars or even Dell or Overclockers build a system we test at 100% load not any less than that. Its worth fixing it when you have the time, as the machine ages the average usage will increase so you may as well sort it sooner and prevent any potential problems.

unkel

And if you buy an expensive CPU, you want at least 100% of its capacity used. Most of us want more like 120%-130%

There's no such thing as "I gave it 110%". But there is with CPUs

Tony EH

ED E wrote: »

Its comparable to say your tyres come off at 120kph, but you only drive at 100kph. You might be getting away with it, but its not a good idea.

Um, to further that analogy, my situation would be, I can drive 120Kph on every road, except the "Handbrake" road, where I have to slow down a bit.