[SOLVED] 5800x Temp Spikes when launching programs.

Dec 4, 2020
9
0
10
As the title says, I recently built a system with a Prime B550 Plus motherboard, a R7 5800x and a Kraken X73 aio cooler. I noticed that when launching games my temperatures spike then go down and stabilize. For example, launching Far Cry 5 makes my temps rise to high 70s max 80C but then go down to around high 50s low 60s when actually playing the game. This happens with other games as well. Another example would be cold war where my temps rise to low 70s when launching the game and go down to med-high 50s and low 60s while playing the game. My cpu is oced at 4.7ghz. Are these spikes normal?
 
Solution
... I'm running at XMP, I can try underclock it to 4.5 or 4.6 to see if that would stabilize it.
...

Are you running a fixed CPU clock? I wouldn't do that, I'd run with clock in AUTO and VCore in AUTO. Fixed overclocking may be feasible but setting it so low is certainly killing game (lightly threaded) performance as a 5800X should boost on it's own to 4.7. And right now it's only making your system unstable in multithreaded work.

I've also read that they boost up to 4.9 or even 5.0 when gaming if all is ideal. Not sure how that works, not enough reports back from field but you have no opportunity to ever see that with a fixed frequency.

But there's also the problem of how susceptible Ryzen 3000 was to fixed voltage/clock...
Dec 4, 2020
9
0
10
I would say normal. CPU usage goes up that makes the temp go up. The heat transfer is not instant so the temp goes up then stabilizes.
OK weird after just launching both games neither went above 65 on launch but last time I launched far cry on launch it did hit the 70s and even 80 at one point.
 
What are you using to read out temperatures? I suggest getting the latest HWInfo64 and look at the Temp die (Average) readings.

The cpu has dozens of temp sensors scattered all over the die's and it reads out the highest one at the moment. When a core boosts that one tiny area of one core can spike to a high temp but it's not thermally significant until multiple cores are boosting and staying that way for a while. It's like a match in a tiny closet..the match is hot but the room isn't; light a couple dozen and keep them going for a while and the closet starts to warm up.

An averaging temp readout gives you a better idea of the thermal state of the CPU and how your cooling is performing.
 
Dec 4, 2020
9
0
10
What are you using to read out temperatures? I suggest getting the latest HWInfo64 and look at the Temp die (Average) readings.

The cpu has dozens of temp sensors scattered all over the die's and it reads out the highest one at the moment. When a core boosts that one tiny area of one core can spike to a high temp but it's not thermally significant until multiple cores are boosting and staying that way for a while. It's like a match in a tiny closet..the match is hot but the room isn't; light a couple dozen and keep them going for a while and the closet starts to warm up.

An averaging temp readout gives you a better idea of the thermal state of the CPU and how your cooling is performing.
I use core temp will try HWInfo64 but I can also hear my.aio fans ramp up when the temps do rise
 
...can also hear my.aio fans ramp up when the temps do rise

Since the temp spikes aren't really thermally significant I just set a custom fan profile that ignores them. For my 240mm AIO I set a fixed fan speed that's barely audible up to about 70C then start ramping it up to something that's not too bad until about 80C. It only really gets annoying at about 85-90C. I'm also using a long ramp-up delay and ramp-down delay so it doesn't 'pulse'.

The annoying thing is how BIOS' work since they monitor the instantaneous CPU temp reading that spikes for fan control. Another thing you can do is follow VRM temp for setting fan speed as it shouldn't spike assuming your MoBo lets you change it to something other than CPU temp.

Or, if your radiator has one, monitor liquid coolant temp. You don't have to have responsive fans with liquid cooling since the liquid has massive capacity to absorb heat. That's one of their advantages over air coolers which have very little thermal capacity so you have to ramp up fans quickly to keep the CPU cool.
 
Last edited:
Dec 4, 2020
9
0
10
After looking at HWInfo64, I can't seem to post my screenshot, the CPU Die avg is 35.1 and it hit 51.5 for maximum although I wasn't doing anything demanding other than opening tabs and installing gyazo. For my motherboard there are two variables named temp 2 and temp 5 which are very high the latter being constant 81 and former being constant 91. Are those a cause for worry?
 
After looking at HWInfo64, I can't seem to post my screenshot, the CPU Die avg is 35.1 and it hit 51.5 for maximum although I wasn't doing anything demanding other than opening tabs and installing gyazo. For my motherboard there are two variables named temp 2 and temp 5 which are very high the latter being constant 81 and former being constant 91. Are those a cause for worry?
The CPU's Tjmax is 90C (I believe) for Ryzen 5800X's so 51.5C certainly isn't troubling even though it seems a bit high for just opening apps or installing something. Still, I'm using a 3700X and don't know exactly what to expect. Also, installing something can be pretty intense as it's de-compressing files which can use all cores in that process.

How does CPU die (Average) behave when you do something extremely intense: like run the stress test in CPUz or let Cinebench23 run a full multi-thread benchmark? Where does it peak out then and does it stay at a fairly level temp?

Running a long stress test is the better way to proof out your cooling.
 
Dec 4, 2020
9
0
10
The CPU's Tjmax is 90C (I believe) for Ryzen 5800X's so 51.5C certainly isn't troubling even though it seems a bit high for just opening apps or installing something. Still, I'm using a 3700X and don't know exactly what to expect. Also, installing something can be pretty intense as it's de-compressing files which can use all cores in that process.

How does CPU die (Average) behave when you do something extremely intense: like run the stress test in CPUz or let Cinebench23 run a full multi-thread benchmark? Where does it peak out then and does it stay at a fairly level temp?

Running a long stress test is the better way to proof out your cooling.
I tried running cinebench and on single core it didnt go above 52C but on multi-core although it was under 75 trying it twice both times I got an application error and the software crashed. I had tested multi-core yesterday and throughout the whole process it didn't go above 80C so im assuming that's alright compared to other temps i've seen online. Running a stress test on CPU-Z the temp didnt go above 70C but I only did it for a couple seconds which also seems fairly good. Problem is the random spikes in temp when opening programs, launching games and sometimes with background apps the temps also spike which I'm assuming is normal. Should I do the stress test for longer to net more accurate results?
 
I tried running cinebench and on single core it didnt go above 52C but on multi-core although it was under 75 trying it twice both times I got an application error and the software crashed. I had tested multi-core yesterday and throughout the whole process it didn't go above 80C so im assuming that's alright compared to other temps i've seen online. Running a stress test on CPU-Z the temp didnt go above 70C but I only did it for a couple seconds which also seems fairly good. Problem is the random spikes in temp when opening programs, launching games and sometimes with background apps the temps also spike which I'm assuming is normal. Should I do the stress test for longer to net more accurate results?
You have to run it for more than just a couple seconds to proof out cooling, especially an liquid cooling. I'd run it for at least 1/2 hour, but if you use your machine for something really intense (like rendering video or images) I'd run it at least twice as long as your longest render.

But Cinebench multithread should not crash your system. Have you made any changes to setup? are you running memory XMP settings for instance? you should try it with pure default BIOS settings to establish a baseline and test each change you make, like after enabling XMP.

Also make sure you've set up the machine: have you installed the latest AMD chipset drivers? Get them from AMD support site to be sure they're latest. Also check if you have a BIOS update available as some boards are getting them and they're significant for Ryzen 5000.
 
Last edited:
Dec 4, 2020
9
0
10
You have to run it for more than just a couple seconds to proof out cooling. I'd run it for at least 1/2 hour, but if you use your machine for something really intense (like rendering video or images) I'd run it at least twice as long as your longest render.

But Cinebench multithread should not crash your system. Have you made any changes to setup? are you running memory XMP settings for instance? you should try it with pure default BIOS settings to establish a baseline and test each change you make, like after enabling XMP.
I'll run it for half an hour like you said but I did not change anything since last night and when i start the multi core test cinebench just gives an error and closes. I'll give it another try.
 
Dec 4, 2020
9
0
10
I'll hold off of the drivers for now as im upgrading my storage tomorrow so I can install all the motherboard and chipset drivers tomorrow although i already do have the ryzen chipset drivers installed. I'm running at XMP, I can try underclock it to 4.5 or 4.6 to see if that would stabilize it. I only have opera and discord running in the background while doing the cinebench test and it gives an application error and refers me to a folder with a text file which just has a lot of text i don't understand. I'll run 2 stress tests for 15-30 minutes each on 4.7 and go down to 4.5 4.6 to see if those frequencies would be better. Would changing the voltage do anything? I see alot of people online running at 1.325 or 1.35 volts while mine fluctuates between 1.2 and 1.3.
 
... I'm running at XMP, I can try underclock it to 4.5 or 4.6 to see if that would stabilize it.
...

Are you running a fixed CPU clock? I wouldn't do that, I'd run with clock in AUTO and VCore in AUTO. Fixed overclocking may be feasible but setting it so low is certainly killing game (lightly threaded) performance as a 5800X should boost on it's own to 4.7. And right now it's only making your system unstable in multithreaded work.

I've also read that they boost up to 4.9 or even 5.0 when gaming if all is ideal. Not sure how that works, not enough reports back from field but you have no opportunity to ever see that with a fixed frequency.

But there's also the problem of how susceptible Ryzen 3000 was to fixed voltage/clock OC's as it degrades the processors. Nobody knows if Ryzen 5000, using the same 7nm TSMC process, is any better but I'd have to say it's really risky to do a fixed overclock with a processor that's so hard to get hold of right now even if you don't mind risking the cost of replacement.
 
Last edited:
Solution
Dec 4, 2020
9
0
10
Are you running a fixed CPU clock? I wouldn't do that, I'd run with clock in AUTO and VCore in AUTO. Fixed overclocking may be feasible but setting it so low is certainly killing game (lightly threaded) performance as a 5800X should boost on it's own to 4.7. And right now it's only making your system unstable in multithreaded work.

I've also read that they boost up to 4.9 or even 5.0 when gaming if all is ideal. Not sure how that works, not enough reports back from field but you have no opportunity to ever see that with a fixed frequency.

But there's also the problem of how susceptible Ryzen 3000 was to fixed voltage/clock OC's as it degrades the processors. Nobody knows if Ryzen 5000, using the same 7nm TSMC process, is any better but I'd have to say it's really risky to do a fixed overclock with a processor that's so hard to get hold of right now even if you don't mind risking the cost of replacement.
I have my core ratio set to 47 in my bios if that's what you are referring to by fixed clock. I don't really understand how auto would work but if I were to try that would it be done through the BIOS or ryzen master. Finally concerning the voltage do you recommend I just leave it as it is?
 
I have my core ratio set to 47 in my bios if that's what you are referring to by fixed clock. I don't really understand how auto would work but if I were to try that would it be done through the BIOS or ryzen master. Finally concerning the voltage do you recommend I just leave it as it is?
That's exactly what I mean...core ratio at 47, or 47 ratio * 100mhz FSB = 4700Mhz fixed clock speed. In that place, in BIOS, where it has the CORE RATIO, just click in it and type AUTO. Or click in there and then hit + or - to increment and decrement till it says AUTO. How it works depends on BIOS.

For Core Voltage (I don't know what it is set for right now) do the same...if it doesn't say AUTO click in it and type AUTO or the +/- thing again.