My Ryzen 7 1700 idles at around 30°C. Every 5-10 seconds, it spikes up to 40°C. There are no background processes causing this because it also happens with a fresh Arch Linux installation. Overall CPU usage stays between 3% and 5%, only the temperature spikes up.
I came across several posts on Reddit and Tom's Hardware where other users have the same issue with all Ryzen models, motherboard manufacturers and chipsets. Most answers point out to background processes or "usage of the mouse". Well, my old FX-8350 did not go up 10°C when using the mouse and temperatures only went up when the CPU was under load.
I opened a ticket with ASUS since I thought it had to do with the motherboard / BIOS. They recommended a CMOS reset which I did, but this did not change anything.
As soon as I put a little load on the CPU, these spikes disappear. For example, compiling software on Linux with one thread under full load results in a constant temperature of 36°C without any fluctuations.
The annoying thing about this is that my computer "breathes" due to a custom fan curve of the CPU cooler. Of course I could make a lesser agressive fan curve, but I would rather like to know what causes theses spikes and how to get rid of them. I have read a lot of guessings but no actual explanation so far.
Can anyone help?
Does your board have Bluetooth that your not using? I had the same situation and it was my bluetooth searching for devices to connect to. It didn't show any cpu usage. I disabled the bluetooth and it went away. I could also see it in my Kill-a-watt meter.
I guess you are running default settings, right?
Try fixed multiplier or fixed voltage and see what happens.
Personally - just ignore those mini temp jumps. Even if you think your linux is not doing anything it is doing losts of stuff in the background.
A naked Arch installation without desktop envirnoment does nothing I am familiary with that causes no measurable cpu load but a 10c spike. I cannot ignore this because it seems hardware related.
My R5 1600 has the same behavior and as far as I know this is normal behavior for AMD CPU's at idle clocks at the Tctl/Tdie sensor. The socket temp(CPU) temp is more accurate for idle temperatures but does not scale well under load, normally the socket is 10c-15c cooler than Tctl/Tdie at full load. My FX-8310 cpu had the same behavior also. You can try
-adjust the fan curve to account for this
-use the CPU sensor for the fan curve(if motherboard BIOS/UEFI allows)
-research, buy and apply your own sensors exactly where you want them on the motherboard
Temperatures And Many Questions - AMD Ryzen 5 1600X CPU Review This article is about R5 vs R7 temperatures and the offset the R7, but also explains how at idle voltages and clocks the temperature readings were "nonsense" for both R5 and R7 CPU's.
It's probably Linux doing something that you're not seeing.
Open a terminal window and run:
sudo top -d .09
TOP will report much quicker, and you'll probably see the processes that are causing this issue.
I wonder if this is caused by defective CPU chips. My new 3900x is doing it on Windows 10. I can't believe that EVERYONE has to modify their fan curves to prevent annoying fan oscillation between 30% and 80% every few seconds. Also, there would be no reason for AMD to use a Tctl that spikes that far higher than CCD 1 & 2 Tdie. There no reason for it since no cooling system is going to make a difference in 1 second. Tctl would almost have to be an moving average.