• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

CPU temperature spikes with Ryzen

I've seen these spikes on a bunch of ryzen systems as well, pretty sure its mobo or software related

NZXT CAM was really bad for it, seeing 5c spikes every few seconds and making the fans spaz out
 
im using the Asus Maximus VI Hero and the software that came with it AIsuite 3. I know a lot of people dont like it. I have 6 fans in my system and I swear by it. together with the amd profile my system is silent and isnt too bad in games. my cpu is never allowed to hit low 80's even....
I soley use the bios it works without any issues with no aisuite on, i did have less issues this generation but not none, the kicker was it stopped my pumps now again, not always permanently but once it did and it was gone.
 
I use both.... not sure if that helps....
 
Mine does it too. This is a threadripper rig. HOWEVER, its not as often. Just every 20m or so. Causes the temp alarm to go off as the spike reaches the 75-80C mark from 60-62 then drops in steps to 65 then tapers down over a 1-3m period.

Haven't ever narrowed it down. Could be a BIOS thing.
 
I think its just how ryzen works

b7fzp5tm1q.jpg

Corsair iCue, 2700x at idle
 
Ummm, I don't think so. Unless you are running a 3rd party app, fan speed control is done at the hardware (motherboard/chipset/BIOS) level, not Windows itself - at least for fans not powered directly from the PSU.
if m$ windows wasnt shit enough for all the effin bloatware, I just might believe you.
 
Windows Power Settings most likely. Also I wonder if Turbo Mode would spike that up too?
 
Hello,
here is a screenshot that tells a lot
127469


Things to note:
  • The spikes are there even at 0% utilization (yes literally zero)
  • The spikes are there at 100% utilization as well, but much less often (two spikes in the screenshot, right side)
  • The spikes do not seem to depend on other CPU stats like clocks, wattage or voltage
Tests system:
  • Fresh, from the store, Ryzen 3600 installation at stock params
  • Latest bios and chipset drivers for a Gigabyte mb
  • Ryzen Power Saver power plan (!)
  • 40C ambent
At this point I consider this either a bug or a quirk and AMD should comment on it.

In any case, the easiest way to avoid fan ramp up is to use a flat curve for both points it jumps between.
 
Hello,
here is a screenshot that tells a lot
View attachment 127469

Things to note:
  • The spikes are there even at 0% utilization (yes literally zero)
  • The spikes are there at 100% utilization as well, but much less often (two spikes in the screenshot, right side)
  • The spikes do not seem to depend on other CPU stats like clocks, wattage or voltage
Tests system:
  • Fresh, from the store, Ryzen 3600 installation at stock params
  • Latest bios and chipset drivers for a Gigabyte mb
  • Ryzen Power Saver power plan (!)
  • 40C ambent
At this point I consider this either a bug or a quirk and AMD should comment on it.

In any case, the easiest way to avoid fan ramp up is to use a flat curve for both points it jumps between.
Your motherboard and it's vendor control the fan speed, not Amd.
Anyway as you note manual fan curves or hysteresis settings got me passed this issue on a ryzen 2600, see it's not a new none issue I've seen other chips do similar.
 
First, garfield is not the OP and the OP made no reference to the Ryzen 3600. So not sure what is happening here.

40C ambent
Huh? Ambient typically refers to the room temperature. If your room is 40°C (104°F), you need to shut down your computer and wait until the middle of the night to use it.

The spikes are there even at 0% utilization (yes literally zero)
No, not literally zero. The only time a CPU is at literally 0% is when totally powered off. Even if sitting idle in the BIOS setup menu, the CPU is being utilized. It may be less than 1% and therefore not registering on the chart, but it is not literally at 0%. And for sure, if booted into an OS, some housekeeping tasks are being performed, if nothing else - to include such tasks as monitoring CPU utilization.
 
I had this ever since I built this Ryzen system 1+ year ago with the 1600x.

Thought its normal for Ryzen so I did not care much about it since the spikes are really short so the CPU fan doesn't have time to ramp up anyway,not in my case at least.

From what I noticed its something about that single core precision boost,when I monitor my CPU usage it ramps up to ~4100 Mhz for a split second and thats whe the temps rise too 'obviously'.

Catched it there but it shows on every monitoring software I tried:

127476



I'm using the Ryzen power plan too cause I had more issues if I turn it off.
 
Last edited:
New AMD power plans available this morning... :toast:
 
I wasn't sure when I posted, thats why I didnt add any info...

ive since sat and thought about every change I made to my pc yesterday.

its almost certainly from the chipset driver I reinstalled for my board, although I cant be positive as it literally didnt show up until this morning.

personally I thought it had been through a windows update... :toast:

edit' ok i googled it and found a few links to possible reasons for the new power plan. https://www.bit-tech.net/news/tech/cpus/amd-ryzen-win10-power-plan/1/
 
You may wantto adopt typical water cooling strategies to address these cyclic bumps.... when using water cooling, unless you use your own temp sensors, the system is responding to instantaneous CPU loads and temps rathern than the more appropriate coolant temp. Many of the utilities provided with medium to high end MoBos include fan control software that allows you to:

a) Turn the fans off ... At idle, or even like this when typing forum messages, w/ CPU load under 0 - 10% the fans generally don't even turn on ... the heat in the coolant heats the air between the rad fins, it rises and leaves the case passively.

b) Ramp Down the fans - when gaming, as soon as you get to a lull, the CPU / GPU is unloaded and it drops to normal temps.... but your coolant is going to be 8 - 10C warmer (16 - 20C in an AIO) and it's advantageous to bring the coolant temps down ... also when gaming between active periods and relatively inactive periods, you can avoid the whiring sounds by gradually damping any change in rps to occur over an interval you set ... 30 - 90 seconds.

c) Ram Up the fans - Those aforementioned little thing that Windows does that require CPU activity, will cause the problems you describe ... ramp these up over a priod of 1/2 the time cycle and ypou shouldn't notice them any more.
 
OK, I believe to have found a fix (at least for me). It seems that the CPU is overcompensating the voltage with any workload spike (any that is too abrupt) and is causing the Temp spikes. This can be fixed (again, in my case) by going into the bios and turning off the automatic boost clocks setting and turning the clock multiplier to your desired speed (in my case 4.0GHz w/ R5 3600, Gigabyte B450M, Corsair H60). This should stop the dumb over compensation and causing the heat spikes to decrease (because it does not have to anticipate that, "OH, your work load is increasing? well I better prepare for it to increase further..." thus generating more heat. I am bey far no expert, but it fixed my problem without any Blue Screen threat.
 
Not sure overcompensating is the right word. It seems to me thermal protection would over-ride any feature that increases performance to the point it causes heat issues. And if overcompensating, it would reduce voltages and clock speeds too much and/or increase fan speeds too much.

In any case I am glad you got it sorted out and thanks for posting your solution! :)
 
Not heat issues per say, because it would still operate within a safe temp( only some times getting as high as 50c). When I opened hwzmonitor, it would show the clocks and heat bouncing up and down coourdinatly. It would cool down then periodically bounce back up again. Turning of the auto boost setting does however fix the problem, however you would need to set a good stagnant clock.
 
Well, the thread is about temperature spikes so we may not be dealing with the same thing.
 
Not heat issues per say, because it would still operate within a safe temp( only some times getting as high as 50c). When I opened hwzmonitor, it would show the clocks and heat bouncing up and down coourdinatly. It would cool down then periodically bounce back up again. Turning of the auto boost setting does however fix the problem, however you would need to set a good stagnant clock.
Seems like normal use to me...
 
OK, I believe to have found a fix (at least for me). It seems that the CPU is overcompensating the voltage with any workload spike (any that is too abrupt) and is causing the Temp spikes. This can be fixed (again, in my case) by going into the bios and turning off the automatic boost clocks setting and turning the clock multiplier to your desired speed (in my case 4.0GHz w/ R5 3600, Gigabyte B450M, Corsair H60). This should stop the dumb over compensation and causing the heat spikes to decrease (because it does not have to anticipate that, "OH, your work load is increasing? well I better prepare for it to increase further..." thus generating more heat. I am bey far no expert, but it fixed my problem without any Blue Screen threat.
You should not run the CPU in a fixed speed just because you think it does not work well on auto unless you want a fixed speed for other reasons. You should take your time and read the posts from threads below. After you read them you will understand better you CPU and its behaviour. There is no need for what you I've set it. And there is not any dump compensation either. On all auto settings regarding the voltages, speeds and multis the CPU is doing great job. The way is working the Ryzen 3000series is new and unfamiliar to most people. This is why the ones who doesnt own a Ryzen 3000, unless they have done a nice research about it, they will give you false advise. And a lot that do own one and dont know the mechanics and the facts of the operation/readings reporting, will also mislead you.
Please give it a read to get to know your nice and capable CPU...

From this all the posts

..and this, you dont need to, but if you like from #17 to #28
 
Last edited:
OK, I believe to have found a fix (at least for me). It seems that the CPU is overcompensating the voltage with any workload spike (any that is too abrupt) and is causing the Temp spikes. This can be fixed (again, in my case) by going into the bios and turning off the automatic boost clocks setting and turning the clock multiplier to your desired speed (in my case 4.0GHz w/ R5 3600, Gigabyte B450M, Corsair H60). This should stop the dumb over compensation and causing the heat spikes to decrease (because it does not have to anticipate that, "OH, your work load is increasing? well I better prepare for it to increase further..." thus generating more heat. I am bey far no expert, but it fixed my problem without any Blue Screen threat.

the ryzen CPU's use high idle voltages when load is low, as they only care about the total amperage to the CPU. As the load increases, the voltages (and the associated clocks) are reduced to keep the amperage per core/CCX and for the entire CPU within reasonable levels.

All you've done is hurt your single core performance and idle wattages, the flickering temp readings is definitely annoying but its not a sign of an actual problem - the CPU temp readings show the reading from the hottest core, throw in a measuring delay and it bounces around a bit. base your CPU cooling off some other reading or change your fan profiles.
 
I have the MSI B450 Tomahawk and a ryzen 3600x. The spike is possibly just because the TDie sensors are not averaged for Ryzen, and they are with other processors. If you can see my picture below there is another sensor in HWinfo, that has the temp averaged that has smoother curves. I think it has to do with how the processor bounces the tasks and sleeps certain cores then wakes them up so often but thats speculation. I feel that in my MSI afterburner which also showed the temp sawtooth spikes that there were power plan settings or MSI/AMD software/driver combinations that limited the sawtooth pattern. But My memory is so bad I may have been imagining it. You could uninstall all the amd/motherboard cpu related software and reset the powercfg -restoredefaultschemes in command line and see if that helps, But it will probably hinder performance elsewhere.

P.S. I'm not an expert by any means. This is just a bad recollection of my experiences.

1586918028901.png
 
Back
Top