Question Ryzen 5 3600 high temperatures

cavini

Junior Member
Jan 24, 2021
3
2
36
Hi.

I've recently bought a Ryzen 5 3600 (a few weeks ago), but I've noticed it is heating so much.

During stress tests (Prime95 Small or Aida64 FPU), my processor goes up to 95C-98C (with stock cooler) peaks, with an average of 92C. I had home a 120mm AIO watercooler from deepcool (maelstrom 120T), and with it, on same tests, the temperature peak drops to around 85C-88C, with average of 83. Because I plan to do some overclock (though I still didn't do any overclock), I bought a Raijintek EOS 240 AIO Watercooler, and for my surprise, the temperatures peaks stay the same, around 85C-88C, also the average stay around 83C on stress test.

I've used watercooler on my systems for more than 10 years by now, and even a 120mm AIO should be enough to keep this processor cool, since it have only 65w TDP. The same Maelstrom 120mm AIO could keep my Phenom II x4 955 (125w TDP) around 70C under full load. Unfortunately, I don't have the Phenom anymore, so I couldn't test it with the Raijintek EOS.

I'm using a Cooler Master thermal paste (Cooler Master E2 IC Essential), the same I've always used, even on my Phenom II x4 955 which had 125w TDP.

One thing I noticed, the radiator and the CPU block/pump from AIO don't get hot, even after 30 minutes of prime95 (which gave an average temperature of 83C). I've checked the radiator and it was cold (not even warm). The tubes from AIO also didn't heat (though when I touch them, I can feel the liquid moving, it wasn't any hot).

I already tried remounting the AIO around 3 times, cleaned and reapplied thermal paste, always same results.

What could be giving it so bad temperatures? My room have air conditioner set to 22C on all tests.

My full PC specs:

Motherboard: Gigabyte Aorus Elite B550M
RAM: 2x8GB DDR4 3000MHz Gloway
VGA: Asus Phoenix Geforce GTX 1050TI (yes, need to upgrade)
Power Supply: Corsair CX 450
CPU Cooler: Raijintek EOS 240 mounted on front (already tried with top mounting, same results)
Case: Deepcool Matrexx 550 with 3 exhaust fans (2 on top, 1 on back)
HD: 1 SATA II 80GB HD, 1 SATA III 1TB, 1 SSD SATA III 120GB


Thanks in advance.
 

dr1337

Senior member
May 25, 2020
303
498
106
How do you have the cpu setup, Manual OC? PBO? And either way, what do your clocks and voltages look like when doing these tests? Also what are you using to monitor the temperature? Have you used more than one app to verify its not some software misreading sensors?
 

cavini

Junior Member
Jan 24, 2021
3
2
36
I don't have manual OC. The bios is on its default settings (PBO and Core Performance Boost on auto). During Prime95 Small FFTs, the clocks are oscilating from 3967MHz to 4067MHz on all cores, and VCore varies from 1.237V to 1.287V.

The peak wattage on CPU reported by HwInfo64, CPU Package Power (SMU) was 87.859W under full load on prime95.

I don't use two app at same time to check temperatures, but already compared the temperatures on HwInfo64 and Ryzen Master software, which reports the same temperatures.
 

dr1337

Senior member
May 25, 2020
303
498
106
hmm I'm gonna have to go ahead and say this looks normal actually, the numbers on your setup sound right. I have never seen my own 3600 get above 75 during normal use, but I also don't monitor the peaks. However after installing prime95 and running it for funsies, im getting about 85 degrees at full load with a nh-d15 and max fan speed.

My best guess is that p95 stresses ryzen in just the right way to actually take it up to peak temperature; prime 95 its not a thermally constrained workload for zen 2. Also after running p95 and stopping all workers, my temps immediately dropped instantly to 55 and they're on the way down to back to normal really quick. I could be completely talking out of my behind, but I'd assume that the temperature being reported during p95 is more like a hotspot temp and not necessarily like the whole package, hence why it drops super fast. Adding on to this, the ryzen master stress test doesn't take my 3600 past 60, despite it reportedly maxing out my cpu as well.
 
  • Like
Reactions: moinmoin

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
28,223
19,708
146
  • Like
Reactions: Captante

Captante

Lifer
Oct 20, 2003
30,239
10,745
136
Ok, thanks for the explanations. I just thought that a 240mm AIO Liquid Cooling would be able to keep this CPU with lower temperatures than the ones I've got, even on stress tests.


Passing stability-tests is by far the most important thing and provided you have things should be fine. However those temps do seem pretty high to me based on my experience below.

I have pretty close to the same setup except with a Gigabyte B450 Pro wifi MB and a Corsair 220-RGB case that has a mesh front with 3x Corsair 120mm "mag-lev" fans blowing in. Also note I'm running overclocked to 4.4 ghz on all 6 cores. H100i is set fans on "quiet" and pump on "balanced" voltage at defaults.

I can loop Furmark and an older version of 3d-Mark (2011 Pro) at the same time for hours and never see above the high 70's/low 80's. This test was done last during the summer with no AC plus using the included Corsair thermal-compound which is just okay.

Seriously considering a 5900x upgrade in the near-future! :p
 

dr1337

Senior member
May 25, 2020
303
498
106
Also note I'm running overclocked to 4.4 ghz on all 6 cores.
I wonder if the temperature difference is also due to a matter of bin. my 3600 wont crack 4.3ghz let alone 4.4 on any core.
I can loop Furmark and an older version of 3d-Mark (2011 Pro) at the same time for hours and never see above the high 70's/low 80's
The real question is what temps do you see in p95 though? afaik those are both GPU benchmarks and don't really stress the CPU at all. In my case furmark for 10 minutes didn't even get my 3600 past 50c, which is nothing at all like p95 darn near instantly shooting my temps up to max.
 
  • Like
Reactions: Captante

Captante

Lifer
Oct 20, 2003
30,239
10,745
136
I wonder if the temperature difference is also due to a matter of bin. my 3600 wont crack 4.3ghz let alone 4.4 on any core.

The real question is what temps do you see in p95 though? afaik those are both GPU benchmarks and don't really stress the CPU at all. In my case furmark for 10 minutes didn't even get my 3600 past 50c, which is nothing at all like p95 darn near instantly shooting my temps up to max.

Furmark includes a built-in cpu-burner ... I ran both that and the GPU benchmark at 800x600 in a window. In 3d mark 2011 Pro I loop the "combined" test... it's far more stressful than P95 by itself when run all together. (Another useful stress-test is OCCT)

Unfortunately the Gen 2 Ryzens tend not to be great overclockers and will only do at best a few hundred mhz over defaults most of the time. You are also correct that "YMMV" very much applies for this sort of thing!

The only other things I can think to suggest are insufficient case airflow or a mounting problem. What happens if you run P95 with the side of the case removed and a big fan blowing directly down on the MB?
 
Last edited:

dr1337

Senior member
May 25, 2020
303
498
106
Furmark includes a built-in cpu-burner ... I ran both that and the GPU benchmark at 800x600 in a window. In 3d mark 2011 Pro I loop the "combined" test... it's far more stressful than P95 by itself when run all together. (Another useful stress-test is OCCT)

Unfortunately the Gen 2 Ryzens tend not to be great overclockers and will only do at best a few hundred mhz over defaults most of the time. You are also correct that "YMMV" very much applies for this sort of thing!

The only other things I can think to suggest are insufficient case airflow or a mounting problem. What happens if you run P95 with the side of the case removed and a big fan blowing directly down on the MB?
bro if you're not going to run p95 and post your results then everything you're saying is meaningless. Idk if you even read mine and OPs posts. Its only your opinion that they're "far more stressful" and furthermore you have zero empirical evidence backing up your point.

Its also not helpful for you to bring up OCCT and not mention any results either. Which FYI only brings my 3600 to 70c and not the near 90c that p95 does. Putting it bluntly but it really sounds like you're just talking out of your behind for the sake of it, and to me it almost sounds like you're trying to spread misinformation on purpose.
 

dr1337

Senior member
May 25, 2020
303
498
106
and just to prove it to you, here is 10 minutes of occt
occt temp.JPG

and just 1 minute of prime95

p95 temp.JPG

hopefully this helps you understand that actually taking measurements is necessary to validate opinions. You may think your arbitrary test is "more stressful", but unless you take the time to validate that opinion you may was well not be saying such things.

I am still convinced that in cavini's case, the high temperature they're seeing is entirely due to the nature of prime95 (especially at small fft), and not due to any sort of real world problem with their cooling solution.

And just to be thorough heres p95 at 10 minutes
p95 10 minutes.JPG
 

DrMrLordX

Lifer
Apr 27, 2000
21,570
10,762
136
Prime95 Small FFTs is pretty hard on Matisse CPUs. And I'm surprised by the vcore settings he's seeing during the test as well. Those are pretty high clocks and pretty high volts. My "default" 3900x is running 1.116V according to CPU-z. HWiFO64 reports a VID of ~1.244 and a core voltage of 1.094v. The CPU maintains a clockspeed of ~3900 MHz, and temps of ~58C. PPT is sitting at ~130W, 12W shy of its allowed sustained maximum of 142W.

@cavini

I think your CPU is pushing too-high volts. You might need to look into that, and we might be able to help you.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,224
9,987
126
@dr1337 , it's entirely possible that OCCT (which uses Intel's Linpack code, and their math libraries?), doesn't activate multi-threaded AVX2 code (on Ryzen), the same way that modern Prime95 and PrimeGrid apps do.

@Captante may have prior experience using those two tools on Intel rigs, where they probably have a more comparable CPU load.
 

Captante

Lifer
Oct 20, 2003
30,239
10,745
136
and just to prove it to you, here is 10 minutes of occt
View attachment 38945

and just 1 minute of prime95

View attachment 38946

hopefully this helps you understand that actually taking measurements is necessary to validate opinions. You may think your arbitrary test is "more stressful", but unless you take the time to validate that opinion you may was well not be saying such things.

I am still convinced that in cavini's case, the high temperature they're seeing is entirely due to the nature of prime95 (especially at small fft), and not due to any sort of real world problem with their cooling solution.

And just to be thorough heres p95 at 10 minutes
View attachment 38947

I'm not sure where you got the idea I didn't believe you? :confused_old:

The last time I ran those tests was shortly after building this system over the summer and it was only for my own information not to prove some point to you. Sorry but I'm not re-running them so I can show you screenshots. If I get bored later maybe I'll do P95. (which has been around for over 20 years btw ... it's NOT all that demanding or frankly even very useful)

Difficult as it may be to accept I have a life that involves things other than posting here and was only trying to be helpful. (kinda made me sorry for doing so too was that your intent?)

Good luck. ;)




EDIT: The only significant differences between our setups I see are the three intake-fans in my case blowing over the MB/GPU and mesh vs glass front. My case interior is usually right at room temp with strong airflow over hot components.
 
Last edited:

Iron Woode

Elite Member
Super Moderator
Oct 10, 1999
30,851
12,350
136
I don't have manual OC. The bios is on its default settings (PBO and Core Performance Boost on auto). During Prime95 Small FFTs, the clocks are oscilating from 3967MHz to 4067MHz on all cores, and VCore varies from 1.237V to 1.287V.

The peak wattage on CPU reported by HwInfo64, CPU Package Power (SMU) was 87.859W under full load on prime95.

I don't use two app at same time to check temperatures, but already compared the temperatures on HwInfo64 and Ryzen Master software, which reports the same temperatures.
assuming I did it the same as you, here is a pic of my system which is in my sig:

1612136678363.jpeg

After an hour of this my system readings stayed the same as in this pic.
 

Furious_Styles

Senior member
Jan 17, 2019
492
228
116
I'm not sure where you got the idea I didn't believe you? :confused_old:

The last time I ran those tests was shortly after building this system over the summer and it was only for my own information not to prove some point to you. Sorry but I'm not re-running them so I can show you screenshots. If I get bored later maybe I'll do P95. (which has been around for over 20 years btw ... it's NOT all that demanding or frankly even very useful)

Difficult as it may be to accept I have a life that involves things other than posting here and was only trying to be helpful. (kinda made me sorry for doing so too was that your intent?)

Good luck. ;)




EDIT: The only significant differences between our setups I see are the three intake-fans in my case blowing over the MB/GPU and mesh vs glass front. My case interior is usually right at room temp with strong airflow over hot components.

Uh, what? P95 small FFT is very demanding and will put a serious load on your vcore. It is NOT a very useful test for gaming stability but if you want to slam your CPU hard it certainly will.
 

DrMrLordX

Lifer
Apr 27, 2000
21,570
10,762
136
Uh, what? P95 small FFT is very demanding and will put a serious load on your vcore. It is NOT a very useful test for gaming stability but if you want to slam your CPU hard it certainly will.

I agree, especially on AVX2 and AVX512 systems where Prime95 will load the CPU up more than a game engine ever could. Prime95 was more relevant as a general-purpose stability tester on older CPUs with lower-throughput SIMD extensions.
 
  • Like
Reactions: Furious_Styles

Captante

Lifer
Oct 20, 2003
30,239
10,745
136
Uh, what? P95 small FFT is very demanding and will put a serious load on your vcore. It is NOT a very useful test for gaming stability but if you want to slam your CPU hard it certainly will.


I still DO use P95 sometimes for testing memory overclock stability or for trouble-shooting RAM errors although the built-in Win10 utility is actually pretty good these days. Been using it for a long time. In fact I can recall running it on my Dell Pentium Pro 200n in 1996.

;)



I agree, especially on AVX2 and AVX512 systems where Prime95 will load the CPU up more than a game engine ever could. Prime95 was more relevant as a general-purpose stability tester on older CPUs with lower-throughput SIMD extensions.


I prefer OCCT when it comes to deciding if my system/overclock is stable. I find if a "mixed/auto" run with (for a 3600) 18-24 threads will go 1-2 hours looping and doesn't crash or get uncomfortably warm you usually are in good shape. (no promises!)


For me personally gaming is the most demanding thing I do on a regular basis and most of it at 1440p as well so I'm absurdly GPU-limited with a GTX-980 anyway.

:confused_old:

(getting a 3070 soon hopefully)
 
Last edited: