3950x high temperatures in IDLE (alround 50 celsius degrees)

Methox

Junior Member
Jan 22, 2020
6
0
6
SETUP
case : lian Li pc-011 dynamic
motherboar : asus prime pro x570
cpu : ryzen 9 3950x (STOCK)
rams : gskill 32gb 16x2 3200cl4 (b-die)
psu : evga supernova 1200 p2 80+ Platinum
ssd : adata xpg sx8200 pro (1tb)
aio : asus ryjuin 240mm (push and pull)

thermal paste usage : Thermal Grizzly Hydronaut applied with the pea technique


FANS CONFIGURATION

SIDE

(intake aio)
PUSH into radiator x2 Noctua NF-F12 IPPC-2000 IP67 PWM
PULL into case x2 Noctua NF-A12X25 PWM

SIDE
(alone fan)
Intake x1 NF-S12A PWM

BOTTOM
Intake x1 Noctua NF-A12X25 PWM

TOP
Exaust x2 Noctua NF-A14 IPPC-2000 IP67 PWM

DESCRIPTION

-After trying infinite settings and fans, with fans in a half-curve configuration of the fans and the pump at 80/100%, the cpu package is usually kept between 50-52 degrees,
I think it's a lot to be in idle, I have all the microcontrollers updated to the latest version as well as the motherboard bios, I tried to change
The box, although Lian Li is not exactly a closed box, just as I had planned to buy the 280 aio Alphacool EISBAER Extreme, but really the Ryjuin is not a bad aio either,
I would like to know your opinions if you think it may be a fault of any part that is defective or is that you simply buy a 16-core toaster, I have seen reviews with this cpu less than 40
degrees with an air cooler and I'm starting to not believe the things I see xD What do friends think? Exchange cpu? Box change? Of aio? , I've been testing for more than 1 week
that I have decided to comment on the forums, I thank you in advance for your help if it is necessary for you to give them the exact version of chipset microcontroller or bios do not hesitate to ask me, thank you very much again, I read you
 
Last edited:

UsandThem

Elite Member
Super Moderator
May 4, 2000
13,054
3,710
146
I'm not sure if the settings are exactly the same with your Asus X570 motherboard, but I know I had to change a few settings in the UEFI with my Asus X470 motherboard. Asus' default settings are pretty aggressive with things like PBO, and my idle temps were in the same ballpark as yours are (with an Noctua NH-U14S as the cooler), and my power consumption at idle were high with my 2700X at stock settings.

I think I set PBO to disabled I believe (there might have been another setting I don't remember at the moment), and my idle power dropped by 15w, and my temps were better (and didn't shoot up 20-25 degrees every 2-3 seconds).
 

Methox

Junior Member
Jan 22, 2020
6
0
6
I'm not sure if the settings are exactly the same with your Asus X570 motherboard, but I know I had to change a few settings in the UEFI with my Asus X470 motherboard. Asus' default settings are pretty aggressive with things like PBO, and my idle temps were in the same ballpark as yours are (with an Noctua NH-U14S as the cooler), and my power consumption at idle were high with my 2700X at stock settings.

I think I set PBO to disabled I believe (there might have been another setting I don't remember at the moment), and my idle power dropped by 15w, and my temps were better (and didn't shoot up 20-25 degrees every 2-3 seconds).
I am currently in the office, when I get home I will share my true configuration of bios because I do not remember the values, although I remember that I have many tests with bios configurations, do you think then are not component problem?
 

UsandThem

Elite Member
Super Moderator
May 4, 2000
13,054
3,710
146
do you think then are not component problem?
No, it doesn't seem that way based on what you've described since I had the same idle temperatures with an 8 core CPU. You have double that, and low 50s is not any cause for concern. It's your what your CPU hits under full load that you would want to see. You never said what you are using to see your temps, but I recommend the AMD Master Utility to see things like temperature.

One thing of note I also just noticed in your post, you say you used the "pea method" when you applied the thermal paste. I have used that method for the better part of 20 years on CPUs, and I found out when I swapped coolers that the thermal paste only covered about 3/4 of the heat spreader. I've really only built with Intel over many of that time, and using that method covered the Intel heat spreader easily. However, the Ryzen heat spreader is larger and when I installed the new cooler, I used a credit card to apply a very thin layer on the entire heat spreader to ensure complete coverage.
 

Methox

Junior Member
Jan 22, 2020
6
0
6
I did all that and I'm still having problems, I've observed this and I think it's the problem


although I have no idea how to fix it, there is a single core shot at 4.3+ while the others are calm
 

UsandThem

Elite Member
Super Moderator
May 4, 2000
13,054
3,710
146
although I have no idea how to fix it, there is a single core shot at 4.3+ while the others are calm
That is normal operation for Ryzen (and Intel) CPUs. Ryzen CPU's best cores boost higher than the others.
 

UsandThem

Elite Member
Super Moderator
May 4, 2000
13,054
3,710
146
Is there no way to distribute work more evenly?
No.

Ryzen automatically uses what cores are the best of what you are doing. You can always try to manually overclock the CPU, but it sounds like you might not want to do anything like that until you do a little further reading on it. That said, most people leave the CPU alone and let it do its thing at stock settings.

Not trying to sound mean or rude, but did you read a review at places like Anandtech or Tom's on how Ryzen CPUs work/perform before spending all of that money?

https://www.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40/2

 

Methox

Junior Member
Jan 22, 2020
6
0
6
no, I just needed cores and bought the one that best suited the budget, when I started trying it, Ryzen hit me in the snout, thanks for all I feel the inconvenience
 

lobz

Golden Member
Feb 10, 2017
1,249
1,237
106
While it shouldn't idle at 4.3 GHz, when it's boosted at that, the temp is not that bad at all.
 

Zstream

Diamond Member
Oct 24, 2005
3,264
180
116
Why is 50C bad? What's your ambient temp. Why are you using a push pull method? You should have three intake fans up front, and one fan pulling heat out the back, along with the water cooler pushing the heat out as well.
 

UsandThem

Elite Member
Super Moderator
May 4, 2000
13,054
3,710
146
no, I just needed cores and bought the one that best suited the budget, when I started trying it, Ryzen hit me in the snout, thanks for all I feel the inconvenience
A mistake I see among newer users of Ryzen CPUs, is they tend to focus on the various boost rates of the cores. Ryzen was engineered and designed to use the best cores, and then work it's way down to the "worst" core. My 2700X's two "best cores" are #1 and #5, and they boost the highest on 1-2 core workloads (which is most desktop use).

There was another user on here who returned several Ryzen CPUs (I think they returned a 3900X and 3950X before "settling"on a 3800X) over what the cores boosted to. They said they were "disappointed" and thought this boost behavior was "defective". :oops:

Instead of focusing on core speeds so much, a much, much better way to test out Ryzen CPUs is to actually just use them for whatever you do with your PC, and benchmark the performance it provides. I personally could really care less what my cores boost to (I looked once just out of curiosity), but all I care about is the actual performance it provides.
 

Topweasel

Diamond Member
Oct 19, 2000
5,326
1,524
136
Did you try repasting the cpu?
Wanted to note a few things. The CPU in the image is not super active but all cores are being pinged in that 50c picture. That isn't idle. The CCD's spike up in temperature pretty quickly. If all cores are active even well underclocked I can see the temps being a little high (50c isn't by any means high on these).

But this brings up a good point. The IO die is in the center where die usually is, but the CCD's are off to the right a bit. A tip that I saw early when I built my 3900x was to do a normal pea in the center then smaller pea's about where the other dies are on the heatspeader. This makes absolutely sure there is full and comfortable thermal compound contact where the CCD's are.
 

lobz

Golden Member
Feb 10, 2017
1,249
1,237
106
Wanted to note a few things. The CPU in the image is not super active but all cores are being pinged in that 50c picture. That isn't idle. The CCD's spike up in temperature pretty quickly. If all cores are active even well underclocked I can see the temps being a little high (50c isn't by any means high on these).

But this brings up a good point. The IO die is in the center where die usually is, but the CCD's are off to the right a bit. A tip that I saw early when I built my 3900x was to do a normal pea in the center then smaller pea's about where the other dies are on the heatspeader. This makes absolutely sure there is full and comfortable thermal compound contact where the CCD's are.
a simple LN2 pot solves all this... this thread is pointless







:laughing:
 

AnandThenMan

Diamond Member
Nov 11, 2004
3,880
204
106
I don't think I've ever seen my 3700x pin a specific core for any length of time here's a few seconds of my system with a few programs open but mostly idle. Sorry for the bad quality hope you get the idea. (click to go to the video right click>view video to make sure it's full size)

 

DrMrLordX

Lifer
Apr 27, 2000
16,009
4,962
136
@Methox

Temp isn't the only thing you should be looking at here, especially not in low activity/idle situations. Temp/voltage on Matisse (desktop Zen2) needs to be viewed in light of current draw. If current draw (and by logical extension, power draw) are low during these voltage/temperature spikes during low periods of activity (idle, or close to it), then the spikes are largely irrelevant. Matisse can sustain voltages of up to 1.47v (or so, depending on the FIT tables of the CPU) in ST workloads without a hitch.

Voltage and temperature are of greater concern in heavy MT loads that cause the CPU to pull a lot of current.

During these temp spikes, you need to use a tool like HWiNFO64 to monitor CPU current draw. If it's 10a or lower, big deal, no harm done. I'd be more interested in seeing what temps you get running Blender Benchmark, Handbrake, Prime95 Small FFTs, and Cinebench R20. Those workloads will push your CPU near the ~145W package power limit.
 
  • Like
Reactions: Methox

Kenmitch

Diamond Member
Oct 10, 1999
7,740
854
126
I don't think I've ever seen my 3700x pin a specific core for any length of time here's a few seconds of my system with a few programs open but mostly idle. Sorry for the bad quality hope you get the idea. (click to go to the video right click>view video to make sure it's full size)

Either he's trolling us or he doesn't realize that his core is actually under load for some reason or another. It's not normal behavior at all for the new AMD offerings.

Using his image it looks like normal behavior if one's running a single core benchmark, temps included.

Here's a snap of a 3600 that was taken at a oddball time which shows just how good the power management works.

1579786893905.png
 
  • Like
Reactions: lightmanek

Topweasel

Diamond Member
Oct 19, 2000
5,326
1,524
136
Either he's trolling us or he doesn't realize that his core is actually under load for some reason or another. It's not normal behavior at all for the new AMD offerings.

Using his image it looks like normal behavior if one's running a single core benchmark, temps included.

Here's a snap of a 3600 that was taken at a oddball time which shows just how good the power management works.
You know what I missed the fact in OG shot that one of the cores is at 4300, the CPU graph doesn't show it or show it well. Just with light usage with several of my cores looking like the rest of his ill go to 40-45c (with none running above 2k). He has a core pinged, of course the CPU is going to flash up into the 50's.
 
  • Like
Reactions: lightmanek

Methox

Junior Member
Jan 22, 2020
6
0
6
@Methox

Temp isn't the only thing you should be looking at here, especially not in low activity/idle situations. Temp/voltage on Matisse (desktop Zen2) needs to be viewed in light of current draw. If current draw (and by logical extension, power draw) are low during these voltage/temperature spikes during low periods of activity (idle, or close to it), then the spikes are largely irrelevant. Matisse can sustain voltages of up to 1.47v (or so, depending on the FIT tables of the CPU) in ST workloads without a hitch.

Voltage and temperature are of greater concern in heavy MT loads that cause the CPU to pull a lot of current.

During these temp spikes, you need to use a tool like HWiNFO64 to monitor CPU current draw. If it's 10a or lower, big deal, no harm done. I'd be more interested in seeing what temps you get running Blender Benchmark, Handbrake, Prime95 Small FFTs, and Cinebench R20. Those workloads will push your CPU near the ~145W package power limit.
I have alround 19A wtf men. Now I will test cinebench and others,
currently in stock only with DOCP / XMP.








I will report the results
 

Attachments

Kenmitch

Diamond Member
Oct 10, 1999
7,740
854
126
I have alround 19A wtf men. Now I will test cinebench and others,
currently in stock only with DOCP / XMP.








I will report the results
Not sure what's up with your rig but in the photo your core 0 is still pegged. What were you running during that shot?

Have you tried turning off DOCP and setting your memory up manually?
 

Topweasel

Diamond Member
Oct 19, 2000
5,326
1,524
136
Yeah 13-15w on Core 0 means its running at or close to full speed. You have something keeping Core 0 active. Whatever it is, is preventing CCD 0 from ever getting close to idle and that is a normal temp for a CCD that has any sustained activity. Notice that CCD 1 is running at 39c. That's within a degree or 2 of mine CPU when actually idle.
 

DrMrLordX

Lifer
Apr 27, 2000
16,009
4,962
136
Not only is Core 0 pegged, but package power was 86W at the time of the screenshot. Something is running on that CPU.
 

ASK THE COMMUNITY