Question 3900x idle temps

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
Hi all,

So ive finished my new build it has a custom cooling loops, and ive been monitoring temps on AIDA64 during idle and during load. I have the CPU at a -0.2 offset because the Asus x570 board decided 1.45v was a good idea from stock... Anyway im seeing sporatic temps on it varying from 36 to 48 degrees over the space of a few seconds, however under 100% load for 3 hours it settles to a constant 48 degrees. Whats happening here?
 

Lutfij

Titan
Moderator
Is it possible to list your full system's specs? List them like so:
CPU:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:

Which BIOS version are you working with?
List your watercooling hardware as well and include the orientation of the airflow in your chassis, alongside the ambient air temps.

Lastly, which app are you using to monitor your temps?

This would be an interesting read for you:
https://www.legitreviews.com/amd-ryzen-9-3900x-12-core-cpu-undervolted-to-1-00v_213060
 
Hi all,

So ive finished my new build it has a custom cooling loops, and ive been monitoring temps on AIDA64 during idle and during load. I have the CPU at a -0.2 offset because the Asus x570 board decided 1.45v was a good idea from stock... Anyway im seeing sporatic temps on it varying from 36 to 48 degrees over the space of a few seconds, however under 100% load for 3 hours it settles to a constant 48 degrees. Whats happening here?
even if the computer is "idle" it's really in idle state, the only time it's "idle" idle is when it's off even in sleep mode it's using the cpu waiting for the wake command. when the pc is just sitting on the desktop "it's not doing anything" but it's still doing something running all the background task and running calcu;ation to decode whatever is going through the ram whether it be anti virus, discord, steam, blizzard, one drive, and ext
 

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
even if the computer is "idle" it's really in idle state, the only time it's "idle" idle is when it's off even in sleep mode it's using the cpu waiting for the wake command. when the pc is just sitting on the desktop "it's not doing anything" but it's still doing something running all the background task and running calcu;ation to decode whatever is going through the ram whether it be anti virus, discord, steam, blizzard, one drive, and ext

I feel like thats overly pedantic, when CPU usage is at a constant state of 1% spiking to 2% occasionally id argue most people would refer to that as an idle PC. Not to mention that your comment doesnt really contribute anything.
 
I feel like thats overly pedantic, when CPU usage is at a constant state of 1% spiking to 2% occasionally id argue most people would refer to that as an idle PC. Not to mention that your comment doesnt really contribute anything.
well then im sorry i dont know how to help you? i just answered your question for you, you're using a asus board which likes to push power the moment the cpu is touched, my asus 990fx r2.0 does it, my asus ASUS X570 Crosshair VIII Hero does it, and my brothers asus board does it." especially" on a oc'd system. asus boards send current basically at the snap of a finger if it thinks the cpu is about to do priority tasking.
 
Last edited:

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
Is it possible to list your full system's specs? List them like so:
CPU:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:

Which BIOS version are you working with?
List your watercooling hardware as well and include the orientation of the airflow in your chassis, alongside the ambient air temps.

Lastly, which app are you using to monitor your temps?

This would be an interesting read for you:
https://www.legitreviews.com/amd-ryzen-9-3900x-12-core-cpu-undervolted-to-1-00v_213060

CPU: 3900X
MoBo: Asus CH8 x570
Ram: G.skill trident Z Royal 64GB running at 3000mhz at 1.35v
SSD: ADATA SX8200 2TB M.2 NVMe
GPU: MSI Gaming Trio x 2080ti with a small OC
PSU: Corsair RM1000x
Chassis: Lian Li O11 Dynamic XL
OC: Windows 10 PRO

Ive yet to flash the latest Bios

Cooling is a EKWB Monoblock, EKWB vector trio GPU block, EKWB Lian Li Distro block, cooper piping, running through 2 360mm rads with 7 120mm SP fans, bottom fans pull air up through Rad #2, at lower RPM that top Fans expelling through Rad #1, with the final fan being a rear intake fan to allow for greater delta t (and to help with inefficency of bottom mounted rad heating intake air. Airflow past bottom rad is 28 degrees, ambiant case temp is 30 degrees and exhaust is 34 degrees. Under full GPU and CPU load GPU stays at 52 degrees and CPU at 48. Fans are custom set based on spending a few hours testing the most efficent config whilst considering noise.

Im using AIDA64 cross checking with AI suite 5.
 

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
well then im sorry i dont know how to help you? i just answered your question for you, you'r usinga asus board which likes to push power the moment the cpu is touched, my asus 990fx r2.0 does it, my asus ASUS X570 Crosshair VIII Hero does it, and my brothers asus board does it." especially" on a oc'd system. asus boards send current basically at the snap of a finger if it thinks the cpu is about to do priority tasking.

I understand that, but your original message contained no such metion to anything other than about what idle was in your opinion, im not trying to have a go at you or anything. I get what your saying, but im mapping CPU usage and Voltage over time against temperature and its not lining up. Voltage, clocks, and load are remaining constant, temps are way too sporatic for the rest of what im seeing.
 
I understand that, but your original message contained no such metion to anything other than about what idle was in your opinion, im not trying to have a go at you or anything. I get what your saying, but im mapping CPU usage and Voltage over time against temperature and its not lining up. Voltage, clocks, and load are remaining constant, temps are way too sporatic for the rest of what im seeing.
if that's the case you may need to under volt your cpu a little bit i would try to flash to the latest bios first, however asus boards are notorious for over volting the cpu in preparation for oc i had this problem with many o' asus boards they like to be geek ready
 

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
if that's the case you may need to under volt your cpu a little bjt i would try to flash to the latest bios first, however asus boards are notorious for over volting the cpu in preparation for oc i had this problem with many o' asus boards they like to be geek ready

Originally it was at 1.45 and i was actually at 45 degrees pretty well flat line for temp, as oyu can imagine at that volatge the load temps didnt shift more than 5 degrees. I noticed the voltage whilst mapping it with AIDA64 and went straight to a -0.2 offset, AIDA64 is telling me i have stable 1.24v (can drop lower but ill want to plan the correct under voltage to provide the best performance), originally i was stable arounf 36-38, Im assuming the memory controller on the CPU might be the cause of the issue as its running 4 16GB sticks daisy chained at 3000mhz at 1.35v so perphaps thats causing some additional issues.
 
Im assuming the memory controller on the CPU might be the cause of the issue as its running 4 16GB sticks daisy chained at 3000mhz at 1.35v so perphaps thats causing some additional issues.
This could be, again that was one
However CPUID HWmonitor is reading out voltages at 1.42, despite bios, AI suite 5, CPUID ROG and Aida64 all reading out 1.2, not idea whats going on with that

of the things I was stating before it could be that it's decoding the information getring pushed the through the ram modules

Hwmonitior saddly to say is crap... Every time I use it on my system it tells me my cpu is hitting 560ish Celsius. The my gpu is hitting 999c and I have tried several times on my new build since I built it 4 months ago
 
Im assuming the memory controller on the CPU might be the cause of the issue as its running 4 16GB sticks daisy chained at 3000mhz at 1.35v so perphaps thats causing some additional issues.
This could be, again that was one
However CPUID HWmonitor is reading out voltages at 1.42, despite bios, AI suite 5, CPUID ROG and Aida64 all reading out 1.2, not idea whats going on with that

of the things I was stating before it could be that it's decoding the information getring pushed the through the ram modules

Hwmonitior saddly to say is **** sorry used to love not anymore... Every time I use it on my system it tells me my cpu is hitting 560ish Celsius. The my gpu is hitting 999c and I have tried several times on my new build since I built it 4 months ago
 

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
of the things I was stating before it could be that it's decoding the information getring pushed the through the ram modules

Hwmonitior saddly to say is crap... Every time I use it on my system it tells me my coy is hitting 560ish Celsius. The my gpu is hitting 999c and I have tried several times on my new build since I built it 4 months ago

Perphaps, im not awfully concerned averaged out its sitting at around 38-40 idle and then around 45 under load. And i could of course see better performance and temps by spending sometime in the bios playing it all to find the sweet spot. Im probably just overthinking the read out from AIDA64

Ironic how HWMonitor is one of the most reccomended sensor softwares on most forums
 
I feel like thats overly pedantic, when CPU usage is at a constant state of 1% spiking to 2% occasionally id argue most people would refer to that as an idle PC. Not to mention that your comment doesnt really contribute anything.
but in the point of time well your are monitoring the volts and the temps and everything you have a program running in the background using your cpu causing random usage spikes whether the systems recording it or not you have something running causing usage through the cpu and ram.
 

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
but in the point of time well your are monitoring the volts and the temps and everything you have a program running in the background using your cpu causing random usage spikes whether the systems recording it or not you have something running causing usage through the cpu and ram.
True, but interestingly Ryzen master is reading 1.42v to all cores... Hmmm im going to go and have a play with it and see what happens, test some OC's and some under voltages and see where i land
 
  • Like
Reactions: white.a.drew

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
Well i have what so far appears to be a stable 4.2 to all cores at 1.25v, seems promising based on what ive read online. Temps are much more stable now, low 40s across the board on around 10% load (makes me fairly sure the ofset to 1.2 was working), i think the AMD Boost settings could have been an issue... curious to see what the temps do under full load now.
 
Hi all,

So ive finished my new build it has a custom cooling loops, and ive been monitoring temps on AIDA64 during idle and during load. I have the CPU at a -0.2 offset because the Asus x570 board decided 1.45v was a good idea from stock... Anyway im seeing sporatic temps on it varying from 36 to 48 degrees over the space of a few seconds, however under 100% load for 3 hours it settles to a constant 48 degrees. Whats happening here?
Ryzen CPU's spike temperature during boosts at idle so that sounds normal. The spikes aren't significant, thermally speaking, and basically impossible to eliminate with anything short of LN2.

When you put the system under 100% load it will maintain the highest boosts it can given sufficient thermal headroom. Your custom cooling loop is working to keep temperatures fairly low, which is good, so it should maintain higher boosts.

But then dialing in a -200 mV offset could very well be gimping the processor as it won't be able to maintain performance. It's called 'clock compression' where you think the clock speed is still high but actual performance can fall off. If your system isn't hitting CB20 MT benchmark scores that could be why.

1.45 V voltage peaks at idle really isn't bad at all. In fact, AMD has repeatedly stated it's perfectly normal for Ryzen to hit up to 1.5V in light bursty loads which is exactly what it sees when 'idle'. If your processor isn't hitting it's max rated boost clocks (single core) at idle or in a 'light bursty load', e.g., Defender quick scan, that's probably why.

But knowing full system spec's would be helpful too.
 
Last edited:

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
Ryzen CPU's spike temperature during boosts at idle so that sounds normal. The spikes aren't significant, thermally speaking, and basically impossible to eliminate with anything short of LN2.

When you put the system under 100% load it will maintain the highest boosts it can given sufficient thermal headroom. Your custom cooling loop is working to keep temperatures fairly low, which is good, so it should maintain higher boosts.

But then dialing in a -200 mV offset could very well be gimping the processor as it won't be able to maintain performance. It's called 'clock compression' where you think the clock speed is still high but actual performance can fall off. If your system isn't hitting CB20 MT benchmark scores that could be why.

1.45 V voltage peaks at idle really isn't bad at all. In fact, AMD has repeatedly stated it's perfectly normal for Ryzen to hit up to 1.5V in light bursty loads which is exactly what it sees when 'idle'. If your processor isn't hitting it's max rated boost clocks (single core) at idle or in a 'light bursty load', e.g., Defender quick scan, that's probably why.

But knowing full system spec's would be helpful too.
Full system specs are already listed.

Ive moved it to a 1.225v 4.2 all core OC and I'm out performing full stock R20, time spy, and fire strike performance, temps are even better now as well.

And i know they have said it can go that high, but it was running 1.4v at lowest points before the offset. That'll significantly decrease CPU lifespan, since most tests suggest 1.25v is the safest voltage.

Considering alot of users are only getting 4.2 as their max boost while pulling 1.45v I'm.pretty happy with my manual OC particularly with stock scores at 7000-7200 and im getting 7400-7600
 

Karadjgne

Titan
Ambassador
Windows is what's happening. That and a minor misunderstanding on how liquid cooling works.

Windows constantly does stuff in the background at idle, you leave the pc alone for a few minutes and windows gets busy. That's where the temp jumps are originated. They happen because cores apply an almost instant load, temps jump up, they are read ever 254 milliseconds.

Now picture heat, it doesn't travel instantly, just quite fast on metal to metal contact. So the cores transfer heat through the Tim, to the IHS, through the paste, through the pretty thick block base, into the coolant. That's gonna take a few seconds. Far longer than the cores actual work, far longer than the dozen or so times the cores temps are read.

Think of it this way. Take a pan and stick it on the burner on high. Burner takes 1-3 seconds to get cherry red, the pan takes a few seconds after that to start getting warm, a few more seconds to get very warm and a few more seconds to get hot. Put water in the pan. 5 minutes later and it's still not boiling.

Heat transfers take time. Temp readings don't. Core temps change almost instantly.

So at idle, or any variable load condition where core usage changes, you'll see temp variations. It's only on a static load that the temp remains static.
 
  • Like
Reactions: RodroX and wehler53

wehler53

Distinguished
Dec 30, 2013
492
5
18,865
Windows is what's happening. That and a minor misunderstanding on how liquid cooling works.

Windows constantly does stuff in the background at idle, you leave the pc alone for a few minutes and windows gets busy. That's where the temp jumps are originated. They happen because cores apply an almost instant load, temps jump up, they are read ever 254 milliseconds.

Now picture heat, it doesn't travel instantly, just quite fast on metal to metal contact. So the cores transfer heat through the Tim, to the IHS, through the paste, through the pretty thick block base, into the coolant. That's gonna take a few seconds. Far longer than the cores actual work, far longer than the dozen or so times the cores temps are read.

Think of it this way. Take a pan and stick it on the burner on high. Burner takes 1-3 seconds to get cherry red, the pan takes a few seconds after that to start getting warm, a few more seconds to get very warm and a few more seconds to get hot. Put water in the pan. 5 minutes later and it's still not boiling.

Heat transfers take time. Temp readings don't. Core temps change almost instantly.

So at idle, or any variable load condition where core usage changes, you'll see temp variations. It's only on a static load that the temp remains static.

Yeah i figured that out along the way, and i understand thats how it works, just the variance was much higher than ive seen in multiple other custom water cooled builds and was quite different from what i was seeing on the same system only a couple days earlier. And higher than what i saw on comparable systems. Anyway i appreciate you taking the time to provide that answer, and im sure that will be useful to more people in the future as well.
 

Karadjgne

Titan
Ambassador
Windows just had an update. There's going to be some residual processing, security updates, bug checks, fixes, driver updates etc that are going to drive temps crazy for a bit. Should settle back down in a week or so.

While other systems can be comp'd, you'll need to take results with a grain of salt. They might have a different bios or slower ram or different voltage settings etc, so temps will always be slightly different than yours. You may have just been the unlucky one to have higher than theirs, but ppl in higher ambient areas will be higher than yours, or in cases with worse airflow, or different fans or lower fan curves etc.
 
Last edited:
...
And i know they have said it can go that high, but it was running 1.4v at lowest points before the offset. That'll significantly decrease CPU lifespan, since most tests suggest 1.25v is the safest voltage.
...

Do you have any authoritative links to support that?

I've seen a 1.325 V claim...which has been regretted by the person where it originated (The Stilt) because it was taken out of context and so significantly misunderstood. And at any rate was not supported by any type of testing. This is the first I've seen a 1.25 V claim.

But what's much more interesting would be long-term reliability testing to demonstrate any voltage is going to result in 'longer' life than letting it boost to, and above, 1.4 V as designed.
 
Last edited:

Karadjgne

Titan
Ambassador
There's 2 definitions for cpu lifespan.
  1. Physical lifespan, which for a cpu is not really measured in years, but in decades.
  2. Usable lifespan, which is measured in software development.

Our of the 2, the second is of far more importance, because it's by far the shortest. Anyone with a dual core cpu or some ppl with quad cores have reached software death already, moving on to bigger and better. While lga1155 is still very marketable in Indo-Asiatic markets, that's a price thing.