Water Cooled 5800X3D Strange Temperatures

Having some interesting CPU temperatures after water cooling my system. It has a 5800X3D, a RTX 4090, and 360mm or RAD volume. The 4090 temps never break 70C. The part that has me puzzled is the CPU cores never break 87C and the CPU does not throttle at all according to HWinfo. But the average CPU temp is often pegged at 90C under a gaming load…

Do I have a bad mount??? Is this just kinda normal with the 3D cache thermal properties???

I really don’t want to have to drain and dismantle the loop… it’s an ITX build that look great just a bit concerned about this one thermal anomaly.

Whats temp doing CPU benchmark? Guessing 4090 gets the cool water first before the cpu as well?

1 Like

The exact same, and yes the 4090 does go before the CPU, however when doing a CPU only workflow such as CPU encoding the numbers are also the same. The 4090 can be chilling at 45C and the CPU under load will have those numbers.

Also from what I can perceive it’s not running any slower also supporting the it’s not throttling bit.

1 Like

I’d say try mounting it again

If nothing changes then change the loop order
What’s the current entire loop order including pump and rads?

The loop order is pump/res, GPU, CPU, 120mm rad, 240mm rad, pump/res.

That’s kinda what I have been wondering I just REALLY don’t want to if I don’t have to try for a remount… as I said it’s a small ITX system and to get to the CPU block and backplane I will have to fully dismantle the system and take the MB out…

How long have you had the system together, might be gunk in the CPU block

Not long, but I ran cleaner through it and flow is very good, I am leaning towards its being a bad paste… just ordered some more thermal paste and I guess I’ll try taking it apart without draining… not looking forward to it though

Well now I’m a bit annoyed, the paste does not look like it was the issue,


And just confirmed, temp number are the same, Core temps are OK but for some reason the average is high… did use a slightly higher quality paste this time (kryonout extreme) and that has made the thermal response after load ever so slightly faster… but not much to really be of major note.

Before you put it back together run it with a air cooler to rest if it’s the block or the CPU

It’s back together already, that and the only air cooler I have will not fit with the other water cooling stuff… I wish I could remember what the Aircooling numbers where before the upgrade…

With that said I’m now wondering if it’s my first theory that it’s just the way the 5800X3D is, I know it has some strange thermal properties. So perhaps on the air cooling the core temps would have gotten high enough to case whatever it is to make the now average high throttle…

And as I said the thermal recovery is very good. If I do a cinebench run after the test within 1 second the average drops down to like 68 along with the cores.

Would love to see some thermal numbers from someone else with a water cooled 5800X3D to confirm, but I can not seem to find any… I know it’s a bit silly to water cool the 5800X3D as it’s not overclockable. The reason I’m water cooling is to fit the 4090 in my case…

What radiator do you have?
I had a similar system, with 4080 and 7950X3D and my temps were relatively high.
I had EK 360x120x48 radiator and it wasn’t enough to dissipate heat.
Tank > Pump > Radiator > CPU > GPU

I added one more of the same radiator like this:
Tank > Pump > Radiator > CPU > Radiator > GPU

Now my temps are about 20C lower for the CPU and 10C for GPU

1 Like

One 120mm and one 240mm. With that said I’m 100% sure it’s not a rad volume issue. I can push a full synthetic load on both the CPU and GPU, the CPU temperature are exactly the same as above, and the GPU never breaks 70c. Water temps are also fine at just 44c without the fans even spinning up all the way that I have based on Water temperature. (100% fan speed at 50c) and from what I’ve read you only really need to be worried about water temps around 55-60c.

But the big outlier is that fact that the CPU behavior is exactly the same even when the GPU is doing literally nothing…. And apparently the paste was not the issue… and as is shown in the HWinfo shot it does not report as throttling either… that’s why it’s odd.

Have you double checked the in/outlets is correct on CPU?

Water going in where it should be going out could cause it not to flow right

The manual for the block states that there is no in and out, the flow can be in either direction with no difference in performance.

I am a huge fan of air cooled CPUs. Too much work for water and potential for issues

My temps were ~40C before adding the 2nd rad. Now, under full load (both CPU & GPU) I am at 32-35C after an hour. Keep in mind, water does not get warmed up immediately, it needs time, but the processors, do.

Yes I know, that water temp is stable after 5 hours of synthetic load. And I’m not really going for quite, the only reason the system is water cooled is to fit a 4090 into an ITX build.

What make mobo?

Where / how are the rad fans plugged in? Are you using a fan controller or anything similar? Seen issues with those not spinning up the fans in relation to the CPU temps. If direct into the mobo, check the profiles to make sure they are aggressive enough.

It’s an ROG Crosshair VIII Impact. The pump is controlled by the CPU header on the MB, set to go to 100% if the CPU gets to 70c. The pump definitely responds to the the CPU load as the water is more flustered in the top of the rad and such.

The fans are controlled by a Corsair Commander Core XT and all 5 fans (3x 120mm attached to the RADs, 2x 140mm for general airflow) are governed by the water temperature.

Again I do not think it’s a RAD volume issue. If I simply are set the fans to 100% under a full load the water temps are 38c.

Everyone seems to think it’s something with the RAD and cooling configuration. And I’m very sure it’s not. I’m almost sure that the issue (if there even is one…) presides between the CPU and the water…. And I have already confirmed it’s not the paste… so it has to be the CPU or the block…

Again really with there was someone else with water cooled 5800x3D numbers that I can compare with…