[SOLVED] Temperature spikes on Ryzen Custom Loop

Jayant Arora

Distinguished
Mar 2, 2014
130
4
18,615
Hey Guys!!

I have recently completed by custom loop with Ryzen 3700X. I have done some stress testing on AIDA x64 Extreme
sh-jfif.jpg
stabilitytest-graph-pbo-off.png



stabilitytestpbooff798.png


I'm, seeing wired momentarily spikes in temps in intervals of 3-4 mins. I know ryzen is quite bursty CPU in nature but still with custom loop this behavior is wonky, how much cooling it needs to settle.
For a test of about 10 mins you can see temps of 65 with momentary spikes in temps to 70-74'C for like 1-2 seconds after every 3-4 mins interval.
Shouldn't be hovering in b\w 64-66 or just flat 65'C ? [not considering the initial spike, i.e, counter reaction time by cooling setup to handle heat load]
I ran it multiple times to monitor the process manager and boost clocks . voltages but no programs seems to have appeared to add stress when spikes are there, similarly boost clock shows no erratic behavior when the spikes are there ?

So why?
  1. I might have used too much thermal paste ?
  2. May be Ethyl glycol i added [10%] might have not mixed well with DW and when comes in contact with CPU die/fins in the block it reduces heat transfer momentarily.
I'm doing custom loop 1st time, so i'm not a pro i might goofed up some where may be What you guys think ?

Just for info my loop setup Info:

30mm x 360mm Copper Rad
d5 pump
210mm Res
Soft Tubes and Brass fittings
Also i have setup a fun android app for system monitoring: [Stock settings, PBO off, No Manual OC, that is just a lucky shot[4400Mhz]
img-20200829-140745-1.jpg
 
Last edited:
Solution
I tested my setup with various speeds 10% 20% 40% 50% 70% and 100% it didn't make any difference in temps

Strange, at load, it should make a difference.

See i have only one copper/ rad only 30mm thick which is kind'a not very good and loop overall length is short so i think moving past 2000 rpm ain't making any difference.

It's a 360 rad, so there's plenty of area to be effective.
Loop 'length' really has nothing to do with it, unless you have a lot of angled fittings.
There is little reason to run a pump so low unless there's a specific reason. If that reason is 'noise', then the noise issue needs to be addressed. Pumps should be incredibly quiet. Vibration noise could be from the mount for the pump/reservoir, not...
temperature builds up in these chips faster than it can be dissipated. that's why you're seeing a 1-2 second spike.
also you don't have a water temps sensor, but judging from the temps I'd avise you to up the flow rate a bit, turn the pump to 2000+ rpm, will probably drop the heat a bit.
 
  • Like
Reactions: Phaaze88

Jayant Arora

Distinguished
Mar 2, 2014
130
4
18,615
temperature builds up in these chips faster than it can be dissipated. that's why you're seeing a 1-2 second spike.
also you don't have a water temps sensor, but judging from the temps I'd avise you to up the flow rate a bit, turn the pump to 2000+ rpm, will probably drop the heat a bit.
Ok i can up the rpm on pump a bit.
My mobo is x570 aorus elite wifi, as per my knowledge it has no 2 pin temp sensor so i cannot use a cheaper $5 sensor and as of now i can afford to buy a full display temp-pressure-flow sensor which will cost me $60, can you suggest a temp sensor for water temp that i can use with my mobo.
 

Jayant Arora

Distinguished
Mar 2, 2014
130
4
18,615
2-71-copy.png


I did set the pump speed to 2100 RPM [20%] speed, it ain't do nothing besides moving water around faster. I was able to take screen shot when temp spikes happens with ryzen master and AIDA x64 sided by side. If you see ryzen master shows no such spike it stays in b\w 62-65'C range in a test run of 15 mins while AIDA shows temp spikes. Can some one shed light on ryzen temp monitoring so different ? I'm unable to find a definnate answer for that, my guesses are it shows CCD Temp and not Total Die Temperature, I'm not sure what's happening here.
 

rubix_1011

Contributing Writer
Moderator
Temp spikes are on the CPU core, not the actual coolant temp. This is something which we have covered multiple times over the course of a decade or more in the forums.

I would be willing to guess that:

- nothing is wrong

But why only run your D5 pump at 20%? Why not full speed to allow it to always provide the best cooling possible for your loop?
 

Jayant Arora

Distinguished
Mar 2, 2014
130
4
18,615
Temp spikes are on the CPU core, not the actual coolant temp. This is something which we have covered multiple times over the course of a decade or more in the forums.

I would be willing to guess that:

- nothing is wrong

But why only run your D5 pump at 20%? Why not full speed to allow it to always provide the best cooling possible for your loop?
I tested my setup with various speeds 10% 20% 40% 50% 70% and 100% it didn't make any difference in temps, See i have only one copper/ rad only 30mm thick which is kind'a not very good and loop overall length is short so i think moving past 2000 rpm ain't making any difference. I will add more thicker rads later may be next year, as of now i'm over with setting up custom loop. Also my pump makes some vibrations moving past 3700 rpms which are annoying cause my table short in length and pc case is very near to my work area[Mouse and keyboard] so it kind'a comfortable with 10-15% speeds having no vibrations at all.
 

rubix_1011

Contributing Writer
Moderator
I tested my setup with various speeds 10% 20% 40% 50% 70% and 100% it didn't make any difference in temps

Strange, at load, it should make a difference.

See i have only one copper/ rad only 30mm thick which is kind'a not very good and loop overall length is short so i think moving past 2000 rpm ain't making any difference.

It's a 360 rad, so there's plenty of area to be effective.
Loop 'length' really has nothing to do with it, unless you have a lot of angled fittings.
There is little reason to run a pump so low unless there's a specific reason. If that reason is 'noise', then the noise issue needs to be addressed. Pumps should be incredibly quiet. Vibration noise could be from the mount for the pump/reservoir, not the pump itself.

Your loop is relatively simple, so it should have no issue with flow rate unless you do not see circulation/turbulence very visibly in the reservoir.

By decreasing the pump speed, you are lowering the overall cooling loop performance capability.
 
Solution