High temps custom watercooling loop

[narrdarr]

⁰Everyone here here has made valid  points.

Inverted cases can be hard to work with too. 

I agree with most of the info with 1 exception. That being the airflow. But I'm going to consolidated a list of things for you to do.

Since every degree seems to matter atm.

 

I see your not worried about about hot coming out the front.

So

Front exhaust

Top intake

Rear exhaust 

Bottom intake

Or

Front intake 

Bottom intake 

Rear exhaust 

Top exhaust 

 

You need to try to establish  some type of flow with your air. Hence the term airflow.

 

Lapping the CPU would also help, maybe a little, maybe a lot depending  on how uneven the IHS is. 

 

Under volting can go along ways. And it pretty easy. What is your current voltage under load? 

 

There been a lot of talk about the pump. 

I'll leave that to them, but a low flow rate will hinder the performance. 

 

I believe your CPU block is directional.  Do you have you inlet\outlet in the correct spot ?

 

 

[dutchguy]

Well, i guess i pushed it a bit to far. To much tension on the cpu block and now its leaking at the fitting xD.... have to wait for new parts to come in. So will try in a few days. 

[dutchguy]

So back up to speed. Got the new CPU block installed and the new GPU. So spend some time last week to balance the new component combination. 

 

Still think this could/should run cooler given the amount of radiator real-estate i have in this case. 

 

atm the build is configured as push from the front, pull from the top and push from the bottom and pull from the backside.  

still see temp climbing the more programs i open. for example hwinfo shows 50 degrees on cpu. 

running; chrome, teams, ryzen master, aurorus SIV, in the background, nvidia broadcast, razer synapse, logitech g sync, corsair i cue, streamdeck, rgb fusion. 

 

so not really a high workload, but the temp keeps stacking even when i ramp up the fans high. 

when i open docker, intellij and postman i will be high in 50's 

 

one thing i do notice is the RM reports a much lower temp then hwinfo does. 

At this moment hwinfo reports: 52

RM reports: 38,xx 

 

aurorus SIV reports same temp as hwinfo though. 

 

maybe my input data for all this is just wrong, what do you think ?  

[dutchguy]

also radiators still don't feel hot at all

[dutchguy]

okey so i started looking into undervolting. and achieved a stable dynamic offset of -0.066 and pbo limit set to motherboard and curve optimizer to -20 on fast cores and -25 on other cores. 

 

temps go as low as 32c now (but thats short after boot)

most of the time is goes in the 40's sometimes lower sometimes higher.

 

while working it goes down to reasonable levels (40's) so im not distracted hole day by my cooling solution going ham.

 

now the temperature really drops, and it feels like the cooling solution is able to combat the heat output. still thinking of reapplying TIM and using very little.

since it still feels like the cooling solution is not effective at getting the heat away from the cpu. but since the heat output is now reduced it is able to keep up. 

while gaming i still see temp rise sometimes to over 70c (at that point the cooling solution is almost running at 100% pump and fans)

 

i dont know, this entire adventure still leaves a sour taste in my mouth. going from 360mm performance rad and 240 extreme rad to 2 x 420mm (ut60 and xt45) i thought this build is going to be so low temperature and silent. but it ended up being just frustrating and distracting even though tinkering with computers is a hobby...

 

@airborne spoon @Lord Nicoll @narrdarr thanks for the help, if u have some more tips regarding this last post please let me know. 

[narrdarr]

33 minutes ago, dutchguy said:

okey so i started looking into undervolting. and achieved a stable dynamic offset of -0.066 and pbo limit set to motherboard and curve optimizer to -20 on fast cores and -25 on other cores. 

 

temps go as low as 32c now (but thats short after boot)

most of the time is goes in the 40's sometimes lower sometimes higher.

 

while working it goes down to reasonable levels (40's) so im not distracted hole day by my cooling solution going ham.

 

now the temperature really drops, and it feels like the cooling solution is able to combat the heat output. still thinking of reapplying TIM and using very little.

since it still feels like the cooling solution is not effective at getting the heat away from the cpu. but since the heat output is now reduced it is able to keep up. 

while gaming i still see temp rise sometimes to over 70c (at that point the cooling solution is almost running at 100% pump and fans)

 

i dont know, this entire adventure still leaves a sour taste in my mouth. going from 360mm performance rad and 240 extreme rad to 2 x 420mm (ut60 and xt45) i thought this build is going to be so low temperature and silent. but it ended up being just frustrating and distracting even though tinkering with computers is a hobby...

 

@airborne spoon @Lord Nicoll @narrdarr thanks for the help, if u have some more tips regarding this last post please let me know. 

no problem

i would also like to see the spread of the paste. fyi using more TIM maybe more helpful than using less.

 

other ideas.

given how old that block is. i wouldn't be surprised if it has convex shaped cold plate. aka the center is higher than the outer edges. companies did this because the cpu die would normally be in center. however, with ryzen cpus they are offset, so the block doesn't make the best contact where it matters. if the cpu heat spreader isnt flat it can be worse.

     you can try upping the mounting pressure. they can handle quiet a bit. its just kinda hard as your mounting screws don't have a way to use a screw drive. better yet lapping the cpu heat spreader and the water block. Or at least the cpu.

 

idk if you can but maybe make your thick rad a push/pull config.

 

 

anyways glad to of help.

[Lord Nicoll]

5 hours ago, dutchguy said:

snip

Could you post like an AIDA64 type stress test that plots CPU temps ona graph so we can see when the temps reach steady state and how the system behaves on a very long test (at least 2-3 hours). Also remember that Ryzen has 3 dies on the CPU and those can be funky to keep cool as they're not in the middle of the IHS, where the most fins are. Also what's the report RPM of the pump? At this point I think you've hit the limit of what watercooling can do on Ryzen, but there might be something really obviously wrong that I can't see because I'm not working on the PC first hand.

[airborne spoon]

So I don't know about the 5950x but I just installed a 5900x at stock speeds and volts. Idle is 46c and while running cinebench R23 I get to 72c.

If you're getting hotter than that I am going to say you need to lap the CPU and/or the block because you're prob not getting good contact.

I was looking at my temps while just messing around doing random crap. I jump an avg of 6c when going from idle to opening something like chrome which is expected and it drops almost immediately after it opens. Again this is a 5900 not a 5950 but they are damn close.

Maybe pull the waterblock off and post a pic of the contact surface. Just make sure when you're taking it off to keep even pressure and pull straight off. And take a clear closeup, not that potato camera looking pics you've been posting

Because to me it still seems like a issue with the mount in some way.

 

[dutchguy]

@narrdarr @airborne spoon took the cpu block off. And found that my tim application might have been a but on the generous side -.-. Took some closeups. Airborne, hope that these look beter. 

[narrdarr]

10 hours ago, dutchguy said:

@narrdarr @airborne spoon took the cpu block off. And found that my tim application might have been a but on the generous side -.-. Took some closeups. Airborne, hope that these look beter. 

not really sure if the heat spreader is really uneven on 1 side or if it was uneven mounting pressure that cause excess to get trapped.

but i'd start with your first idea of repasting with a bit less. i also recommend.

1.spreading the paste yourself instead letting the pressure doing it for you.

2. when you mount they block. with you hand push it down and wiggle it back and forth. helps make sure the any excess gets out.

3. increase your mounting pressure. they screw will only let the thumb nuts go so far. so just get them as tight as you can Or you can add washer between the block and the thumb screws which will keep them from bottoming out which will allow you to get more turns with a tool just be careful as a 1/4 turn extra maybe alot depending on tight it already is.

4. you probably already know, but tighten the thumb nuts in a x pattern.

[dutchguy]

6 hours ago, narrdarr said:

not really sure if the heat spreader is really uneven on 1 side or if it was uneven mounting pressure that cause excess to get trapped.

but i'd start with your first idea of repasting with a bit less. i also recommend.

1.spreading the paste yourself instead letting the pressure doing it for you.

2. when you mount they block. with you hand push it down and wiggle it back and forth. helps make sure the any excess gets out.

3. increase your mounting pressure. they screw will only let the thumb nuts go so far. so just get them as tight as you can Or you can add washer between the block and the thumb screws which will keep them from bottoming out which will allow you to get more turns with a tool just be careful as a 1/4 turn extra maybe alot depending on tight it already is.

4. you probably already know, but tighten the thumb nuts in a x pattern.

i think that the pressure on the bottom side of the cpublock is higher due to the tubing hanging off it. thats why we see almost no TIM on the bottom side. But it might indicate the block not being mounted with sufficient pressure. so will try again. what could i use to spread the TIM ?  

[narrdarr]

30 minutes ago, dutchguy said:

i think that the pressure on the bottom side of the cpublock is higher due to the tubing hanging off it. thats why we see almost no TIM on the bottom side. But it might indicate the block not being mounted with sufficient pressure. so will try again. what could i use to spread the TIM ?  

 

 

I just use my finger. sometime in a ziplock so i don't get it on my skin.

but you can use whatever. like a plastic spoon. you break it or cut it to shape that works best for you. 

 

[dutchguy]

On 3/8/2021 at 9:10 AM, narrdarr said:

 

 

I just use my finger. sometime in a ziplock so i don't get it on my skin.

but you can use whatever. like a plastic spoon. you break it or cut it to shape that works best for you. 

 

re applied TIM, temps still spike (from 32 to 66) when starting allot of applications. But it does seems to return to lower values and stay +- there.
Next up the long stress test with metrics.   

[dutchguy]

I was thinking, @narrdarr does the orientation of the CPU block matter ? since i use an inverted MB setup. 

[narrdarr]

9 hours ago, dutchguy said:

I was thinking, @narrdarr does the orientation of the CPU block matter ? since i use an inverted MB setup. 

it shouldn't matter, but some blocks need to have to coolant flow into a specific inlet and out the other for optimal performance. 

 

AMD spikes are not uncommon.

 

how did the stress test go?