Strange temperatures readings

[fola]

Dear PCS Community,

I am the mildly happy owner of a Recoil 15' laptop (bought in June '21). I have no complaints about performance, but I am beginning to question the overall quality and stability of the system.

Around 2 months ago I opened an RMA because the CPU fan was always stuck at 100% no matter the temp or load. I sent the laptop and got it back a couple of weeks ago, with the fan issue apparently solved.
During those last couple of weeks I was totally submerged in work, so I did not use the laptop very much, but now that I am using it a little more I noticed something very disturbing: the CPU temperature readings (and consequently the fan speeds) look really unreliable to me.

Attached a plot (from HW Monitor PRO) of the CPU package temperature. It spikes from 50°C to more than 90°C within 2 seconds and then back again super fast. And this is in Office more with VERY little load (a couple of cores around 20% load, negligible GPU load) for some office apps.
Are those spikes even possible? I honestly don't think so.
Moreover, the fans never ever go below 30% (which results in idle temps of around 50°C) which is strange too... in the past, Office mode was almost always really quiet even under mild load (and by "quiet" I mean no fan at all, or very little).

What do you think? Should I RMA this once again?

 

[SpyderTracks]

Dear PCS Community,

I am the mildly happy owner of a Recoil 15' laptop (bought in June '21). I have no complaints about performance, but I am beginning to question the overall quality and stability of the system.

Around 2 months ago I opened an RMA because the CPU fan was always stuck at 100% no matter the temp or load. I sent the laptop and got it back a couple of weeks ago, with the fan issue apparently solved.
During those last couple of weeks I was totally submerged in work, so I did not use the laptop very much, but now that I am using it a little more I noticed something very disturbing: the CPU temperature readings (and consequently the fan speeds) look really unreliable to me.

Attached a plot (from HW Monitor PRO) of the CPU package temperature. It spikes from 50°C to more than 90°C within 2 seconds and then back again super fast. And this is in Office more with VERY little load (a couple of cores around 20% load, negligible GPU load) for some office apps.
Are those spikes even possible? I honestly don't think so.
Moreover, the fans never ever go below 30% (which results in idle temps of around 50°C) which is strange too... in the past, Office mode was almost always really quiet even under mild load (and by "quiet" I mean no fan at all, or very little).

What do you think? Should I RMA this once again?

Have you serviced it since you've had it?

Any performance laptop will need repasting and fans and heatsinks cleaned of dust about once a year otherwise it will start thermal throttling.

 

[fola]

Hey Spyder

I cleaned the sinks and fans, but did not repaste it.
It was not really clear (will need to ask more more details) but I think that they replaced the whole motherboard during the first RMA (at least, the motherboard was in the list of components they sent me they need to order because not in stock) so this would include a repaste.

Anyways, my question was more like "is that even possible for a whole CPU package to go from 50 to 90 and back to 50 within 2 seconds?". Because it never did that, and the real sudden changes for no apparent reason look fishy to me. I may of course be wrong and I just need to repaste...

 

[SpyderTracks]

Hey Spyder

I cleaned the sinks and fans, but did not repaste it.
It was not really clear (will need to ask more more details) but I think that they replaced the whole motherboard during the first RMA (at least, the motherboard was in the list of components they sent me they need to order because not in stock) so this would include a repaste.

Anyways, my question was more like "is that even possible for a whole CPU package to go from 50 to 90 and back to 50 within 2 seconds?". Because it never did that, and the real sudden changes for no apparent reason look fishy to me. I may of course be wrong and I just need to repaste...

If the paste is old then yes, absolutely. It can reach thermal throttling within a fraction of a second.

 

[SpyderTracks]

Since about 9th gen Intel's have run extremely hot to the point where on 10th and 11th gen Intel said thermal throttling was a feature because they couldn't get reasonable temps.

How Intel laughs at thermal throttling, and why it shouldn't | Digital Trends

Intel's 16-core HX processors are the best of the best, but that comes at the cost of pushing against a thermal wall that could become a problem in the future.

www.digitaltrends.com

You may find once thermal paste is reapplied that you need an undervolt to bring temps properly inline. It's basically a given on an Intel chassis.

Bit start with a repaste.

 

[SpyderTracks]

That temperature graph reads like a flaky sensor to me.....a sudden 2 second spike from 50 to 90+ (and then down again after 30 seconds) doesn't read right to me (then again I'm not a laptop person)

It's thermal throttling, it's normal (for an overheating processor)

Also, just because you may only have a word app open or something doesn't mean that windows isn't doing background tasks. The way to be sure is open taskmgr and sort by cpu usage. There will have been a task running that corresponded with that spike.

Something like a virus scan or windows updates can spike CPU usage.

 

[SpyderTracks]

Fair enough. I still think, even with thermal throttling, a 40 degree drop in temperature in 2 seconds doesn't sound right...even in a desktop, I wouldn't expect heat to be dissipated that quickly, let alone a laptop

Very normal. CPU temps rise and fall in a matter of milliseconds. From a cold boot if you apply a load of 100% from something like cinebench, and monitor temps you'll see them go from idle straight to high 90's or whatever your peak temp performance is pretty much instantly from a cold boot.

And that's on a healthy system. It doesn't take a while to build up temperatures.

When paste or heatsink isn't correctly applied (or dried out) there is less contact between the heatsink and CPU IHS, so effectively at a microscopic level, there will be areas of the IHS that only have ambient cooling, and those sections will instantly temperature throttle.

The CPU package temp (which is what that graph is showing) will account for all sensors within the CPU die so if one of them is throttling, it will report the die as throttling.

It is possible that the heatsink isn't correctly applied or even warped (which we've had before) bit a repaste would still be the first port of call to unearth that.

 

[ubuysa]

As a non-hardware expert, what stood out for me in that graph was how fast the CPU cooled down. That's a really good sign IMO. Anyone can heat up a CPU, but the speed at which it cools is an excellent indicator of how good your cooling is.

 

[SpyderTracks]

So this is a typical temperature graph for the Linpack test from Puget systems

Impact of Temperature on Intel CPU Performance

While we all know that modern processors need active cooling, there is actually very little official information on how temperature affects a CPU's performance. Do you really need a high-end liquid cooled setup to get peak performance, or is the little stock cooler that comes with most CPUs...

www.pugetsystems.com

Just an example of ramps up and down based on load.

 

[fola]

I could of course be mistaken, but to me "throttling" means limiting CPU power to avoid spiking the temperature. Which is not really the case here since I get the spikes even with arond 20% CPU load in Office mode. I get no sign of throttling in CPU speed or voltage.

Moreover, a bad or deteriorated paste job would actually hamper the ability of the package to cool down (that's the whole point of paste and thermal dissipators after all), and for this reason I look with extreme suspicion at those 40°c drops in temperature within 1-2 seconds.
I would understand it better should the CPU load drop from 90% to 0%, but here we are talking of loads that never ever go over 20%...
Honestly, I tend towards an erratic temp sensor myself...

In any case thanks for the help and suggestions. I sent a message to PCS to have another opinion on that. Will keep you posted if you are interested.