CPU Temp spikes

[Protegit]

I noticed some strange temp spikes while my CPU is under load. Is this behaviour normal? If not, what could be the cause of these spikes? The CPU is not thermal-throttling. Case air flow is 9/10. The CPU (Overclocked R9 3900x) is cooled by an NH-D15 in a single CPU-fan configuration. The graph shows CPU temp, CPU all-clock, CPU fan speed vs time, where one grid unit equals 5 seconds. Max CPU temp in these spikes is 97 degrees Celsius and avg CPU temp under synthetic load is 87 degrees Celsius. In reality the CPU dosnt hit these kinds of temperatures. It happens only when benchmarking/stress-testing, but the spikes happen all the time, although they get dangerously high only under synthetic load.

[Jae Tee]

It's most likely the boost algorithm, how ever you prob could redo the thermal paste just to be sure.

[Protegit]

Just now, Jae Tee said:

It's most likely the boost algorithm, how ever you prob could redo the thermal paste just to be sure.

Already changed thermal paste. No change in results. If it was the boost algorithm, then we would probably see all-core or even single-core clocks rise at the spikes, but thats not the case. Good guesses tho.

[Chabax]

perfectly normal ryzen behaviour, 97c not good not bad

[Void Master]

1 minute ago, Protegit said:

Already changed thermal paste. No change in results. If it was the boost algorithm, then we would probably see all-core or even single-core clocks rise at the spikes, but thats not the case. Good guesses tho.

Is there any weird voltage spikes ?

[Protegit]

10 minutes ago, Void Master said:

Is there any weird voltage spikes ?

@Void Master Nope, stable 1.32V. Its all over the place when its idle, but thats how its supposed to be afaik. The 12V PSU rail is holding stable too, although at 12.192V.

[LogicalDrm]

Have you checked how it looks per core? The temps are on the high side overall considering max temp for Ryzen CPUs is 95C (Intel is 105C). Meaning that if it would be multiple cores, it should throttle or auto-shutdown.

[Protegit]

26 minutes ago, LogicalDrm said:

Have you checked how it looks per core? The temps are on the high side overall considering max temp for Ryzen CPUs is 95C (Intel is 105C). Meaning that if it would be multiple cores, it should throttle or auto-shutdown.

@LogicalDrm I cant seem to find per clock temps in AIDA64. Would this kind of information be available in different software? I thought there were no per-core temp sensors but I might be wrong. Unless you meant core clocks, then here, I did another test.

 

I already made a lot of test and dont seem to find any causes. The only thing that comes to my mind is cooler/heatpipe damage but there are no signs of damage on the outside of my cooler. There is no reason for me to believe that my cooler has been damaged aside from the temp spikes. I do not know how damaged heatpipes affect CPU temp, so Im only guessing. The cooler is slightly tilted to the side tho (about 2 degrees) but the base of the cooler is flat on the CPU and makes perfect contact so I always neglected this imperfection because it shouldnt affect anything. its the heatpipes that are bent this way. I could order another NH-D15 and check if the problem persisted and then just return it for free (the new NH-D15 I mean. The "old" one (the one I currently have) is 6 months old) (my local law allows me to return a product 14 days after I bought it without giving an explanation), but that would be a di*k move.

 

In this test I turned off GPU stress-test and the temps are about 7 degrees celcius lower spike or no-spike.

 

[LogicalDrm]

2 minutes ago, Protegit said:

@LogicalDrm I cant seem to find per clock temps in AIDA64. Would this kind of information be available in different software? I thought there were no per-core temp sensors but I might be wrong. Unless you meant core clocks, then here, I did another test.

I was gonna say OCCT, but I don't know how that works with Ryzen. HWiNFO64 should do the trick.

[Protegit]

9 minutes ago, LogicalDrm said:

I was gonna say OCCT, but I don't know how that works with Ryzen. HWiNFO64 should do the trick.

@LogicalDrm Also, here is a view from the top of my case on the NH-D15, so you can see the slight tilt.

 

[Bladzy]

Hello just wanted to post an article with the exact same title as yours before I realized there is a relatively new one. ^^

 

I hope you don't mind if I post here, but I've got a very similar problem with a Ryzen 3 3100. It on idle is temp spiking 24/7. Spikes are usually around 10C and even moving the mouse brings temps up by 10. The stock cooler with its thermal pad is being used, and this behavior occurs with and without overclock applied. Under load temps idle out nicely, but without load on 1-2% utilization the temps spike every 3 secs identical to your first picture (after your test was over). Even tho temps while on idle are in the safe range (40-50C) I can't imagine this being healthy at all, since this is literally a physical principle being used to destroy materials.

 

 

I've read this would be relatively normal for ryzen 3 cpus. But I'd like to be assured if anyone reading this knows for sure. I'am just a little nervous since my ryzen 5 2600 does'nt behave like that. This is the first time I encountered sth. like this.

[Protegit]

48 minutes ago, Bladzy said:

Hello just wanted to post an article with the exact same title as yours before I realized there is a relatively new one. ^^

 

I hope you don't mind if I post here, but I've got a very similar problem with a Ryzen 3 3100. It on idle is temp spiking 24/7. Spikes are usually around 10C and even moving the mouse brings temps up by 10. The stock cooler with its thermal pad is being used, and this behavior occurs with and without overclock applied. Under load temps idle out nicely, but without load on 1-2% utilization the temps spike every 3 secs identical to your first picture (after your test was over). Even tho temps while on idle are in the safe range (40-50C) I can't imagine this being healthy at all, since this is literally a physical principle being used to destroy materials.

 

 

I've read this would be relatively normal for ryzen 3 cpus. But I'd like to be assured if anyone reading this knows for sure. I'am just a little nervous since my ryzen 5 2600 does'nt behave like that. This is the first time I encountered sth. like this.

 

Interesting. So it would seem that my cooler is not the problem here. Can you make use though that yours is properly mounted?

 

I will give it more thought tomorrow. Goodnight for now.

[Protegit]

I've read more about this problem and it seems that it is in fact normal Ryzen behaviour...

It makes my fan curve adjusting 10x more challenging. Ehh... whatever. Kinda disappointed. I will be tweaking some BIOS settings to check if they change this behaviour and I will update this post if I find something worth mentioning.

[Bladzy]

14 hours ago, Protegit said:

Interesting. So it would seem that my cooler is not the problem here. Can you make use though that yours is properly mounted?

 

I will give it more thought tomorrow. Goodnight for now.

Hello, I just double checked, the cooler sits tight on all sides, maybe a little to tight I admit ^^ but it should make gread contact with the cpu. 

[Bladzy]

1 hour ago, Protegit said:

I've read more about this problem and it seems that it is in fact normal Ryzen behaviour...

It makes my fan curve adjusting 10x more challenging. Ehh... whatever. Kinda disappointed. I will be tweaking some BIOS settings to check if they change this behaviour and I will update this post if I find something worth mentioning.

Well that sucks, I hoped to hear sth. else, but if thats the case we can't really do anything can we? I messed around my with fan curves a lot and no matter what I try the spiking stays the same. It will fall down to a comfy 39C and jump for no visible reason straigt up to 49+ then fall down again in a few secs and continue the cycle. I'vee also tried the AMD specific power plans (you should try  this as well, according to the internet this helps in some cases) but that didn't work for me. I guess I will just leave it like that even tho I am not happy. I am almost certain this continuous spiking on idle will reduce lifespan, I mean frequent temp change is generally bad. I have no ideas left.

 

 Good luck to you!

[Protegit]

1 hour ago, Bladzy said:

I messed around my with fan curves a lot

@Bladzy You didnt mention it, but if you're having problems with fans ramping up and down with the temperature try changing the ramp-up and ramp-down time to 3 or more seconds. This smooths out the fan behaviour so you can forget that your temp chart is a bloody saw-wave lol. Also it doesnt hurt to have the fan curve flat right in the idle/light load temp oscilation area, so that the fans only ramp up when youre switching between idle, wathcing YT and gaming for example (3 levels of workloads). It does require a lot of testing tho, but I managed to calibrate it just right.

[Protegit]

@Bladzy So I found the culprit. But dont celebrate yet. Its in the BIOS settings and its called Core Performance Boost and you have to disable it. It gets rid of all the spikyness but also caps your clock speed at stock speed which is 3.8GHz (at least for my 3900x). So the first answer @Jae Tee was right. It is the algorithm. It just didnt show in the clocks. I dont know what MB youre running, so I dont know where in the BIOS you might have this setting, but I have an MSI x570 Gaming Plus and it was under OC>Advanced CPU Configuration>Core Performance Boost>Disable. I will do some benchamrking to see whether its worth leaving it disabled, but the temps sure are much lower and much smoother. I didnt realise that the auto overclocking AMD put in their BIOS was SO aggressive. You dont even have to do manual overclocking nowadays lol. And it was enabled by default. I will come back with benchmark results once I do the testing. For now, here are the temps after disabling Core Performance Boost. Barely exceding 70 degrees Celcius lol. I did not adjust my fan curves for this test tho, so I could lower the temps even more.

[Bladzy]

4 hours ago, Protegit said:

@Bladzy You didnt mention it, but if you're having problems with fans ramping up and down with the temperature try changing the ramp-up and ramp-down time to 3 or more seconds. This smooths out the fan behaviour so you can forget that your temp chart is a bloody saw-wave lol. Also it doesnt hurt to have the fan curve flat right in the idle/light load temp oscilation area, so that the fans only ramp up when youre switching between idle, wathcing YT and gaming for example (3 levels of workloads). It does require a lot of testing tho, but I managed to calibrate it just right.

Hello again, thanks for the fast response. Fan speeds are not a problem, the stock cooler is suprisingly quiet and it is more of a budget system with every last cent put into performance over noise or looks (Its an upgrade rig for my brother, he ran a horrible old cpu and slow ddr2 ram with an almost broken hdd but a sujprisingly "good" graphics card (gtx 1050). The card was bottlenecked by at least 70% xD. Now it is the opposite. He is supposed to swap graphics card and monitor aswell, as soon as he has the money, but until that he at least can game stable at 60+ fps in almost any game). The motherboard as far as I am concerned doesn't support adjusting ramp down/up time. It's a budget one from ASRock (b450m pro 4f) fancurve adjustment on its on is a hussle  (it only allows setting 5 control temps and those have to be at least 5C apart from each other) But maybe I've missed sth. I will look into it again. But the board in general is missing some nice to have features, such as LLC. Should have thought about that maybe when buying... Building on a tight budget is always challengign ._.

[Bladzy]

4 hours ago, Protegit said:

@Bladzy So I found the culprit. But dont celebrate yet. Its in the BIOS settings and its called Core Performance Boost and you have to disable it. It gets rid of all the spikyness but also caps your clock speed at stock speed which is 3.8GHz (at least for my 3900x). So the first answer @Jae Tee was right. It is the algorithm. It just didnt show in the clocks. I dont know what MB youre running, so I dont know where in the BIOS you might have this setting, but I have an MSI x570 Gaming Plus and it was under OC>Advanced CPU Configuration>Core Performance Boost>Disable. I will do some benchamrking to see whether its worth leaving it disabled, but the temps sure are much lower and much smoother. I didnt realise that the auto overclocking AMD put in their BIOS was SO aggressive. You dont even have to do manual overclocking nowadays lol. And it was enabled by default. I will come back with benchmark results once I do the testing. For now, here are the temps after disabling Core Performance Boost. Barely exceding 70 degrees Celcius lol. I did not adjust my fan curves for this test tho, so I could lower the temps even more.

I don't know if this will change anything for me, since the overclock the CPU is locked at 4.2 ghz all time. So it is never boosting. This is one reason why I am so clueless. I thought after overclocking I'd get rid of this weird temp fluctuation. But at least I think I have those settings, I will definitally try it out, maybe it keeps the Overclock? But actiuvating them means in my case I am loosing 600mhz which is a lot when it comes to gaming. And I specifically advised for one of the newer CPUS because they overclock so well   

 

Anyways thanks a lot for the tipp, I will try it out tomorrow in the morning when I have got a little bit of time. Nice to see your temps balance out, but I think we should not be forced by AMD to underclock our hardware. This sounds odd to me  

 

Have a nice evening o/

[Protegit]

5 hours ago, Bladzy said:

I don't know if this will change anything for me, since the overclock the CPU is locked at 4.2 ghz all time. So it is never boosting. This is one reason why I am so clueless. I thought after overclocking I'd get rid of this weird temp fluctuation. But at least I think I have those settings, I will definitally try it out, maybe it keeps the Overclock? But actiuvating them means in my case I am loosing 600mhz which is a lot when it comes to gaming. And I specifically advised for one of the newer CPUS because they overclock so well   

 

Anyways thanks a lot for the tipp, I will try it out tomorrow in the morning when I have got a little bit of time. Nice to see your temps balance out, but I think we should not be forced by AMD to underclock our hardware. This sounds odd to me  

 

Have a nice evening o/

Its not really underclocking for me. Its just disabling the boost feature. The clock runs at 'stock' speeds now. I did some testing in Cinebench R20, and the result was a 4.67% drop (or a 4.9% boost if you wanna view it the other way around) in performace after adjusting fan curves (I did 7 tests with the Core Performance Boost enabled, avaraged the results and did the same after diasbling the Core Performance Boost feature and then compared the results). I decided to keep the boost feature turned off for now, but I will have to do more single threaded benchmarking, since R20 is heavily multi-threaded. I dont expect a more than 5% performance drop after seeing the multi-threaded numbers tho.

 

Also, remember that when you disable this feature you can still overclock your CPU manually. Try it, Im curious how it will affect your temps and perf. Also, congratulations on overlocking by +600MHz lol. That a pretty big leap. Didnt know those low/mid-range chips were so overclockable. Did you do any benchmarks to measure the performance improvement? I might mess with some more manual overclocking too.

I willl soon be building a budget rig as well, but Im going with R3 3300x since its bang4buck is spectacular.

One more question. Why do you want LCC so bad? I have it and dont use it. Maybe I just dont understand how it works or how to set it up properly.

[Bladzy]

13 hours ago, Protegit said:

Its not really underclocking for me. Its just disabling the boost feature. The clock runs at 'stock' speeds now. I did some testing in Cinebench R20, and the result was a 4.67% drop (or a 4.9% boost if you wanna view it the other way around) in performace after adjusting fan curves (I did 7 tests with the Core Performance Boost enabled, avaraged the results and did the same after diasbling the Core Performance Boost feature and then compared the results). I decided to keep the boost feature turned off for now, but I will have to do more single threaded benchmarking, since R20 is heavily multi-threaded. I dont expect a more than 5% performance drop after seeing the multi-threaded numbers tho.

 

Also, remember that when you disable this feature you can still overclock your CPU manually. Try it, Im curious how it will affect your temps and perf. Also, congratulations on overlocking by +600MHz lol. That a pretty big leap. Didnt know those low/mid-range chips were so overclockable. Did you do any benchmarks to measure the performance improvement? I might mess with some more manual overclocking too.

I willl soon be building a budget rig as well, but Im going with R3 3300x since its bang4buck is spectacular.

One more question. Why do you want LCC so bad? I have it and dont use it. Maybe I just dont understand how it works or how to set it up properly.

Good afternoon, I am done testing, but first of all:

 

I calculated my Overclock refering to its stock speeds which are 3.6 Ghz, if we calculate it considering its ability to boost (to 3.9 Ghz), its only a 300MHz overclock. But those chips can be pushed higer  if you have a good cooling solution at least, unfortunally the cooler we already had was not backwards compatible. Thats why we had to use the stock cooler, otherwise we would have overclocked to ~4.5 Ghz. We could probably do that even with a stock cooler, the system is never hit with an all core workload. Its preatty much gaming only. But it was not worth the hussle for me. 

 

I wish I had LLC because Voltages spiked from 1.336V up to 1.48V sometimes even surpassing 1.5V, which is bad. And LLC could have helped there. But after tweaking the voltages I found that at 1.325V it is barely Volt spiking at all, but this took some time . (CPU Voltage usage spikes between load and idle states, a low LLC level helps keeping those spikes small)

 

And indeed tuing helped, it now outperforms a I7 7700K in multi and singlethreded performance

 

Have fun building your budget rig! I wanted my brother to go for the R3 3300x as well... but he was to greedy for the extra 20€

 

But now the testing:

 

As you said disabling core performance boost on stock speeds eliminates the temp spikes entirely, but it also sent us back ~200 points behind the stock version with cpb (ill call it like that from now on) enabled:

 

Everything on stock settings with cpb enabled looks in comparison not even that bad, but it still spikes so much that I barely can tell the difference, unfortunally I messed sth. up with the pictures and cant find it ._. You have to take my word for that.

 

Deactivating cpb and overclocking to 4.2Ghz had no effect on curves , but it also did not damage performance, since it was already running higher than its usual boostclocks:

 

 

Btw. don't be afraid of it hitting 96C, the PC will only be used for gaming and that never stresses all the cores 100%, at least as far as I am concerned. 

 

Just because I like how it looks, here is the overclocked version after the single threaded test:  

 

close but we got em

 

I found out sth really interesting while testing, seems like ryzen master and Afterburner as well as hardware info do show conflicting values when it comes to temperature, but both report spikes. Ryzen master only displays them smaller:

 

 

And last but not least. I found a sort of weird but working soluton, I thought to myself: when the systems temps are stable under load we maybe have to put it under continuous load. I achieved this with Wallpaper engine (a tool used to display animated wallpapers). This sets our CPU usage to 6% and temps up by 10C to a continuous ~52C with only occasional spikes occuring. It is not the best solution but it is a solution for me at least lol

 

I also wrote to AMD, maybe they will answer, who knows. 

 

Sorry for spelling mistakes, I know there are some hidden in my texts, but I am in a hurry and english is not my mother tongue. So please forgive that. 

 

have a nice day, and fun with your rigs! I think I will just leave it at 4.2Ghz until AMD responds ^^ 

[Bladzy]

If you can take the hit of a 5% performancve drop sure. I'd also say ist is the best for the CPUs longevity. It will take longer for workloads to complete which in my mind at least wouldnt be that big of a deal. But since you bought such a good CPU I guess it matters to you that it completes its tasks as fast as possible. 

 

We only want to game on it ^^ therefore clock speed matters a lot. and being limited to stock speeds is scary in that sense. Ofc rn. the speeds are not that important since its GPU is still fairly old. So we could turn it off until he gets a better one.

 

I dont know, does gaming matter to you? If so you should see the FPS difference between cpb of and on and decide if it matters. If not leave it disabled. Id say thats the best solution in that case. 

[Protegit]

On 6/12/2020 at 4:10 PM, Bladzy said:

And last but not least. I found a sort of weird but working soluton, I thought to myself: when the systems temps are stable under load we maybe have to put it under continuous load. I achieved this with Wallpaper engine (a tool used to display animated wallpapers). This sets our CPU usage to 6% and temps up by 10C to a continuous ~52C with only occasional spikes occuring. It is not the best solution but it is a solution for me at least lol

 

I also wrote to AMD, maybe they will answer, who knows. 

 

Sorry for spelling mistakes, I know there are some hidden in my texts, but I am in a hurry and english is not my mother tongue. So please forgive that. 

 

have a nice day, and fun with your rigs! I think I will just leave it at 4.2Ghz until AMD responds ^^ 

I lolled at that wallpaper engine XD

 

If you get a reply from AMD, please share it here

 

I dont think anyone cares about spelling mistakes on this forum. As long as its not hindering communication. Hell, Im not even an English native speaker myself.

 

I used to game a lot, now I only game casually, but usually fine tune everything to get at least 144 FPS in order to get the screen to be the bottle-neck.