Water Loop - High CPU Temp

[mrmeeseeks84]

I recently finished my new build with a custom loop (I'll attach a pic so you can see the loop) and it runs Distro > GPU > CPU > RAD > RAD > Distro. My water temperature sensor is at the bottom of the distro-plate.

 

I've been watching temps and noticed that my CPU seems to run a bit hot. In gaming it's not too bad, so perhaps it doesn't matter but thought I'd explore a bit so I started running Aida64 stress test on CPU/GPU.

 

During the stress test, the temps seem to shake out like this:

 

CPU immediately spikes to ~70c then climbs gradually to ~80c

Water temp reported as 28c and seems to stay there

 

• GPU/CPU running 100%
• GPU steady around ~50c
• CPU steady around ~80c
• Water temp says 27c

 

Because of the way the CPU temp seems so much higher than the GPU in the loop (granted, the GPU is BEFORE the CPU and after two radiators in the loop so maybe it makes sense) that perhaps I need to re-seat the CPU block or something, although I feel pretty confident the paste and seating is fairly solid but we all make mistakes.

 

Very possible I could be nitpicking here too as again, during gaming it seems relatively fine with CPU temp closer to the 60c mark. Idle temps are typically under 40c

 

 

[RageTester]

I guess indeed GPU pre-warms water and CPU gets warm water to begin with so doesn't cool down that much. At what point of the loop is water temperature measured?

 

System looks cool!

[mrmeeseeks84]

Just now, RageTester said:

I guess indeed GPU pre-warms water and CPU gets warm water to begin with so doesn't cool down that much. At what point is water temperature meassured?

The water sensor is down at the bottom of the distro plate, you can kind of see it in the picture inbetween where the radiator goes back into the distro plate and then the inlet for the pump.

[RageTester]

I don't think you could even start a stress test If cooling wasn't working.

[For Science!]

8 minutes ago, mrmeeseeks84 said:

 

The loop is fine, I think the mounting is also fine. Probably more to do with voltages and frequencies. CPU temp is almost always much worse than GPU temps which is down to IHS vs direct die, and small die vs big die.

 

PS The order of components doesn't matter, the GPU is not sending warm water to the CPU. 

[FloRolf]

9 minutes ago, For Science! said:

 

PS The order of components doesn't matter, the GPU is not sending warm water to the CPU. 

Of course it is, lol. 

But you are right, it doesn't matter because the temperature in the loop levels out and the water flow is too fast to make a notice. 

 

 

What CPU are you running btw? Is it OC'ed? What settings? 

[Chen G]

You have poor thermal conductivity between the chip and the cold plate, for whatever reason.

[For Science!]

6 minutes ago, FloRolf said:

Of course it is, lol. 

Yes, definitely at a true molecular level. Though the more practically correct way of looking at it is that the pump is circulating water that is gradually getting heated up by the heat generating components as a whole.

 

1 minute ago, Chen G said:

Thermal barrier, probably a CPU with paste under the IHS

It's a 3700X

[mrmeeseeks84]

Correct, 3700x - just running it stock. No overclocking or messing with the voltages at this time.

[For Science!]

4 minutes ago, mrmeeseeks84 said:

Correct, 3700x - just running it stock. No overclocking or messing with the voltages at this time.

I think the recent X570 boards will push the voltages and frequencies quite hard, perhaps have a look using HWinfo64 during the stresstest to see whats being requested for Vcore? Either negative offset or manual voltages (effectively undervolting) may get you better temps with no expense of performance. Just watch out for Zen2 there is some strange behaviour when you undervolt too much.

[mrmeeseeks84]

9 hours ago, For Science! said:

I think the recent X570 boards will push the voltages and frequencies quite hard, perhaps have a look using HWinfo64 during the stresstest to see whats being requested for Vcore? Either negative offset or manual voltages (effectively undervolting) may get you better temps with no expense of performance. Just watch out for Zen2 there is some strange behaviour when you undervolt too much.

Attached screenshot is what I'm seeing for voltages during stress test. VCORE says it's hanging around 1.256 with a peak of 1.312 I've seen it hit 1.328 as well after taking the screenshot. Guess I'll try and see what others are running for VCORE and what their temps are for this processor.

Also, an observation: My office is a bit cooler now in terms of ambient temperature, and the water temp is now 24c vs 27c and the CPU temp is hanging out at a more reasonable ~73c - 75c.

Room temp is 64f degrees vs it must have been around 74f in here when I was testing earlier

Maybe this is all normal, I guess just the vast difference in temp between my GPU and CPU makes me wonder if something is off with the block or something. *shrug*