My CPU is running hot... Is this normal or could something be set up wrong?

[Drakinite_]

I have a Rysen 5800X being cooled by a Corsair H115i RGB PRO XT, 280mm AIO radiator. If I recall correctly, the AIO radiator came bundled with the 5800X when I bought it. I have precision boost overdrive on and if I recall correctly, I've got it slightly overvolted. When I have Folding@home running (CPU only, no GPU), it goes up to 92°C and stays there, drawing around 105 W. With this cooler, is this type of temperature expected at this power draw, or might I have set something up incorrectly (e.g. bad thermal paste application, or incorrect pressure between the CPU and cooler, or have fans blowing in the wrong way)? I tried looking up the cooling capacity of the H115i Pro XT 280mm, but I couldn't find a simple "it's good for up to X watts" answer.

 

For additional context: When I start/stop intensive CPU activity, be it Folding@home or an AIDA64 stress test, the CPU temps rocket up / drop down to normal within a matter of seconds. I don't know the significance of this fact, though, nor do I know whether it's expected.

 

Here's a screenshot of HWMonitor while Folding@home is running:

[josefah]

Nothing is wrong. It's running hot while you're stress testing every component of the CPU and it's still under 100.

 

It will be more than fine under normal conditions. 

 

 

[NikolakiH]

Its not that concerning with a full tilt stress test like that. Its kind of weird that you overvolted it, though, since that just makes it use more power to achieve the same clock speed in most cases.

[Drakinite_]

7 minutes ago, NikolakiH said:

Its not that concerning with a full tilt stress test like that. Its kind of weird that you overvolted it, though, since that just makes it use more power to achieve the same clock speed in most cases.

After writing the post, I checked my overclocking settings:

 

- DDR4 is set to 3600 MT/s, which is the rated speed of my RAM, but i think that's still technically an overclock, right?

- Everything else in the "OC Tweaker" section of the BIOS, including voltage offset and CPU load-line calibration, is set to default (auto)

Advanced -> AMD Overclocking settings:

- ECO mode is disabled

- Precision Boost Overdrive is set to "Advanced", PBO Scalar is set to 3X, CPU Boost Clock Override is set to +100

So I misremembered, there was no overvolt.

 

To test further, I just changed the PBO settings to Auto and rebooted, and I still see the same behavior.

11 minutes ago, josefah said:

Nothing is wrong. It's running hot while you're stress testing every component of the CPU and it's still under 100.

 

It will be more than fine under normal conditions. 

The weird thing is, I get the exact same behavior when Folding@home is set to Light. Task Manager reports CPU usage as around 60-70% while AIDA64 and HWMonitor report CPU usage at 40-45% (weird, I don't know why there's a discrepancy), but the CPU still consumes 100-105 W and the temperature still goes up to 92­°C in that case. And it's still really sensitive, where the temperature shoots up within a few seconds when I resume folding, and then cools back down really quickly when I pause folding.

[GuiltySpark_]

10 minutes ago, Drakinite_ said:

Precision Boost Overdrive is set to "Advanced", PBO Scalar is set to 3X, CPU Boost Clock Override is set to +100

Looks normal especially since you haven't touched Curve Optimizer. (owned a 5800x for 3 years now)

[Beerzerker]

8 minutes ago, NikolakiH said:

Its not that concerning with a full tilt stress test like that. Its kind of weird that you overvolted it, though, since that just makes it use more power to achieve the same clock speed in most cases.

Not true.
It will use more power because it's at a higher clock when you OC it, not because of higher voltage alone.

The chip will consume what power (Wattage) it needs to operate at the speed you have it set for when it's OC'ed and operating temps will go up of course because of the combination of higher voltage and clockspeed.
Voltage alone doesn't mean more power is being used.
You can take a chip running at stock speed, measure the wattage it's using,  volt it up, re-measure what wattage is being consumed - It will be about the same everytime.
There will always be a little variance, depending on the load it's handling but nothing to even worry about if purely talking voltage - Operating temps however are a different story.

OP - You can try checking to see what minimum voltage the chip will run stable (For folding) with at the clocks you have it set for if you want but I agree with what was said earlier, it's fine as is right now. 
I know folding places a real demand on system stability but as is, you're about right in the ballpark.
 

[Drakinite_]

4 hours ago, GuiltySpark_ said:

Looks normal especially since you haven't touched Curve Optimizer. (owned a 5800x for 3 years now)

4 hours ago, Beerzerker said:

OP - You can try checking to see what minimum voltage the chip will run stable (For folding) with at the clocks you have it set for if you want but I agree with what was said earlier, it's fine as is right now. 
I know folding places a real demand on system stability but as is, you're about right in the ballpark.

Ok, thanks for the input - It still feels weird, but I'm glad to hear that these temps are expected. Is 92°C totally fine for a CPU, or will it slowly wear down if it's kept at that temperature for a very extended period of time?

 

Theoretically, if I did something to improve the thermal contact between the CPU and cooler, would that maybe prevent the temps from going so high (or allow the CPU to boost for longer, consume more power, etc.)? Again I don't fully understand the science behind this stuff, but the fact that the CPU cools down really quickly after pausing folding seems to imply to me that the radiator is having no problems getting rid of the heat. I didn't check my coolant temps earlier today when folding had been running for hours, but right now the coolant is at a chilly 33.2°C.

 

4 hours ago, Beerzerker said:

Not true.

It will use more power because it's at a higher clock when you OC it, not because of higher voltage alone.

The chip will consume what power (Wattage) it needs to operate at the speed you have it set for when it's OC'ed and operating temps will go up of course because of the combination of higher voltage and clockspeed.
Voltage alone doesn't mean more power is being used.

Are you sure? I have a background in electrical and computer engineering, so I have a pretty strong grasp of the physics of transistors, but not much experience with the practical aspects of CPUs and overclocking. What I know for sure is this:

• Each transistor wastes some X amount of charge whenever it switches on or off, so the more frequently it switches, the more current it consumes over time. Which I think means: Current is directly proportional with clock speed
• Sometimes a higher voltage is needed in order to stably achieve a higher clock speed, because (slight oversimplification) the faster you switch, the more likely you are for little errors to add up and result to an incorrect result; and a higher voltage can help mitigate that result
• Power = current * voltage

What I'm pretty sure can be concluded from bullets 1 and 3 is that at the same clock speed, a CPU will consume more power with a higher voltage and less power with a lower voltage. That was actually my reason for doing a slight overvolt in the past: I actually wanted my CPU to consume slightly more power during the wintertime, to warm my room a little bit more, without caring about better performance. Not sure if that's a bad idea though?

 

And what I think can be concluded from the bullets is that at the same power, the CPU that's at a higher voltage would actually be running at a lower clock speed. Which if true, yes, would mean that it's silly for me to overvolt my CPU. And I think it means that NikolakiH was right? But I'm aware CPU frequency and clocking is really complicated compared to the fundamentals I know about transistors; please lmk if I missed something.

[GuiltySpark_]

We've been fielding this question for years at this point.

 

Normal things apply. Enough thermal paste, proper mounting pressure, pump on full, etc..

 

Robert Hallock a few years ago at launch:

 

Quote

"Yes. I want to be clear with everyone that AMD views temps up to 90C (5800X/5900X/5950X) and 95C (5600X) as typical and by design for full load conditions. Having a higher maximum temperature supported by the silicon and firmware allows the CPU to pursue higher and longer boost performance before the algorithm pulls back for thermal reasons,"

 

 

I have a Ryzen 7000 system now but my 5800x box is still up and running as a secondary. Custom per-core Curve Optimizer undervolt and I see ~85c or so under Cinebench R23 loads. 

[Mark Kaine]

dont *overvolt* a Ryzen cpu, leave it at default settings simply.

 

so, yes, this is totally normal with the wrong settings,  99% of ryzen cpus dont need any "tweaking" although an *undervolt* can be beneficial. 

[Mark Kaine]

19 minutes ago, GuiltySpark_ said:

We've been fielding this question for years at this point.

what astonishes me is that people try over volts or overclocks, apparently without cross testing,  and never read up on how much better most modern cpus and gpus perform with an undervolt. 

 

indeed, since years, a decade almost at this point? 

[Beerzerker]

27 minutes ago, Drakinite_ said:

Are you sure? I have a background in electrical and computer engineering, so I have a pretty strong grasp of the physics of transistors, but not much experience with the practical aspects of CPUs and overclocking. What I know for sure is this:

• Each transistor wastes some X amount of charge whenever it switches on or off, so the more frequently it switches, the more current it consumes over time. Which I think means: Current is directly proportional with clock speed
• Sometimes a higher voltage is needed in order to stably achieve a higher clock speed, because (slight oversimplification) the faster you switch, the more likely you are for little errors to add up and result to an incorrect result; and a higher voltage can help mitigate that result
• Power = current * voltage

What I'm pretty sure can be concluded from bullets 1 and 3 is that at the same clock speed, a CPU will consume more power with a higher voltage and less power with a lower voltage. That was actually my reason for doing a slight overvolt in the past: I actually wanted my CPU to consume slightly more power during the wintertime, to warm my room a little bit more, without caring about better performance. Not sure if that's a bad idea though?

 

And what I think can be concluded from the bullets is that at the same power, the CPU that's at a higher voltage would actually be running at a lower clock speed. Which if true, yes, would mean that it's silly for me to overvolt my CPU. And I think it means that NikolakiH was right? But I'm aware CPU frequency and clocking is really complicated compared to the fundamentals I know about transistors; please lmk if I missed something.

I can tell you this - It takes x amount of power to do x amount of work.

I believe you are referring to what's called "Walling" in that once the chip gets to a certain speed it starts resisting going faster, taking more voltage to push it further which will increase wattage consumed too.
I've dealt with that for a very long time and knowing how to tell if a chip is doing that plays a part in OC'ing in general.

BTW you have a PM.

 

[Mark Kaine]

50 minutes ago, Drakinite_ said:

please lmk if I missed something.

sure. you're missing the important part. modern cpus/gpus have something called  "boost algorithm" and that depends mostly on *temps* so if you put more "voltage" they will run hotter and *downclock* because of that, if you really wanted to take advantage of the higher power consumption you would need *much* better cooling,  we talking sub 0C here... aka LN2

 

 

its also wrong that more v automatically means more w... in many cases less v will allow the chip to run cooler and thus draw *more* wattage. 

 

 

this is all well known,  things have changed since the 90s, my friend.  

[Drakinite_]

Thanks for the informative responses, all.

 

46 minutes ago, GuiltySpark_ said:

Normal things apply. Enough thermal paste, proper mounting pressure, pump on full, etc..

How do I know how much mounting pressure is right? This is all the information on mounting pressure that my cooler's manual provided (aka nothing):

 

 

4 minutes ago, Mark Kaine said:

sure. you're missing the important part. modern cpus/gpus have something called  "boost algorithm" and that depends mostly on *temps* so if you put more "voltage" they will run hotter and *downclock* because of that, if you really wanted to take advantage of the higher power consumption you would need *much* better cooling,  we talking sub 0C here... aka LN2

 

its also wrong that more v automatically means more w... in many cases less v will allow the chip to run cooler and thus draw *more* wattage. 

Interesting. The thing about temps makes sense, especially after I saw my CPU power slowly drop from 105W to as low as 80W, when i set my cooler to a "Zero RPM" preset, just to see what would happen.

 

Would "in many cases less v will allow the chip to run cooler and thus draw *more* wattage" only be applicable if I'm actively running something stressful (e.g. Folding@home) or is it also true if I'm just doing everyday tasks? For example, say the CPU is just idling, or doing a couple things in the background, so the cooler isn't the bottleneck. Might an overvolt lead to higher power draw in that case?

 

21 minutes ago, Beerzerker said:

I believe you are referring to what's called "Walling" in that once the chip gets to a certain speed it starts resisting going faster, taking more voltage to push it further which will increase wattage consumed too.

Definitely wasn't referring to that.

 

I think Mark Kaine's response explained best what I was missing. My electrical engineering / computer engineering classes are all about the fundamentals, going into details about how transistors n shit are made and how computing systems are designed, but they never talked about the ways modern CPUs manage their own performance. Probably for three reasons: (1) i'll bet a lot of the tech behind that is proprietary and secretive; (2) CPUs are like a fuckin labyrinth down there, really advanced stuff for undergraduate classes; and (3) my school doesn't do any research in the field, so it's all based on old textbooks rather than current researchers or industry professionals.

[Mark Kaine]

very simple example:

 

r5 3600 stock : 65w

 

r5 3600 with PBO negative voltage offset (that's a minus) 200 : 130w

 

about same temps.  and guess what? 5-10% more performance. 

as said you can oc, but your cooling game has to be on point,  at least for Nvidia and amd, they typically start lowering the max boost from around 50c (yes) so that's the ideal temp really. 

[Drakinite_]

That doesn't directly answer my question. It sounds like you're still talking about max power draw, but my question was about normal use, i.e. when CPU is mostly idle but might be doing a couple tiny things (e.g. web browsing, code editing, playing music, that kinda stuff where only a couple percent is used). My 5800X caps at around 115W but in the use case I'm describing, it varies from like 40-70W depending on what it's doing.

[Mark Kaine]

8 hours ago, Drakinite_ said:

For example, say the CPU is just idling, or doing a couple things in the background, so the cooler isn't the bottleneck. Might an overvolt lead to higher power draw in that case?

well during light tasks it shouldn't really matter, but medium/ high loads it does...

 

as said typically it always makes more sense to undervolt,  as said mostly because of temps - it also depends on silicone lottery,  as in not all chips will be able to even run an undervolt (also traditionally ryzen cpu just dont overclock well)

 

tldr, as long your cooling is good its best to run things at stock except the ram/xmp.

 

you're overthinking all of this (imho)

[Mark Kaine]

ps: but since your cpu already seems to run hot, undervolting is your only feasible option,  overvolt would make zero sense it'll only run even hotter??

[Mark Kaine]

8 hours ago, Drakinite_ said:

My electrical engineering / computer engineering classes are all about the fundamentals, going into details about how transistors n shit are made

exactly,  this has all very little to do with how modern boost algorithms work...

 

basically eli5 explanation:  most chips draw too much voltage from factory  - that's actually a safety margin from manufacturers so the chips dont starve from too little power. hence *most* chips can be easily undervolted and thus will run cooler and boost higher / longer.

 

and all that really only matters if your system is already running too hot - which seems to be the case in your case.