CPU Instantly Overheats on Watercooling. GPU Stays Cool

[suprc4]

My cpu has instantly been hitting 95c as soon as i run prime95 for two seconds. I thought maybe something was wrong with my loop. So i took my loop apart and added in a gpu waterblock for my 1080ti in the meantime. I put everything back together. And my gpu idles at 36c and under load goes up to 38c. cpu is 38c idle and 95c after 2 seconds of prime95. I have a xeon x5675 at 4.7ghz on evga x58 classified 3 motherboard. Waterblock is an xspc rasa waterblock and I have two 360mm rads with 6 fans on each one, as well as a 120mm rad with one fan on it. Now obviously I know it's not something wrong with my loop because my gpu only goes up 2 degrees celcius at full load. However how could my cpu be hitting 95c in two seconds, and how is that possible? If it was a flow restriction in the cpu waterblock then the gpu would be hitting the same temps too right? Since water wouldn't be flowing. But it seems like everything is flowing just fine. I'm stumped here and I don't really want to pay for a new cpu waterblock if I don't have to because I can't really figure out what else it could be.

 

Thanks everyone

[Jurrunio]

14 minutes ago, suprc4 said:

and how is that possible?

1. Not putting thermal paste for the CPU

 

2. Not tightening the screws holding the waterblock tightly enough

[suprc4]

1 hour ago, Jurrunio said:

1. Not putting thermal paste for the CPU

 

2. Not tightening the screws holding the waterblock tightly enough

I've tried remounting the water block 3 times over the past few days. Unless theres something wrong with the mounting system of the waterblock where its not making good contact with the cpu. But theres thermal paste on both the cpu and waterblock every time I take the waterblock off so it is making contact with the cpu.

[Benjeh]

is the waterblock made for that cpu? is it making contact? is the flow of coolant going through the right way?

[suprc4]

Just now, Benjeh said:

is the waterblock made for that cpu? is it making contact? is the flow of coolant going through the right way?

Yup I've had this waterblock for 5 years and used in on this cpu and a 3770k and back to this cpu again. I'm using the lga 1366 backplate and everything. http://www.sidewindercomputers.com/xsracpuacwab.html this is the waterblock but mine is all black not chrome.

 

I'm just so stumped lol.

[Benjeh]

6 minutes ago, suprc4 said:

Yup I've had this waterblock for 5 years and used in on this cpu and a 3770k and back to this cpu again. I'm using the lga 1366 backplate and everything. http://www.sidewindercomputers.com/xsracpuacwab.html this is the waterblock but mine is all black not chrome.

 

I'm just so stumped lol.

 

When did you last clean the block and dont EK make different base plates for different chips?

[Milkcrate336613f]

Hmm this one is a challenge to fully answer without beingthere hands on. I suspected immediately what the others also said. If its one loop and gpu temped are fine it rules out any restrictive flows during stressing especislly if the gpu is being stress tested too. That tells me hoses, blocks and pump should be fine. Now to me ot leaves eitger the mpunting system is somehow not securing tightly to the faceplate of the cpu which can sometimes be checked with the wiggle test(just be gentle), I would say thermal paste but you've already tried that. I'd also think maybe something is drastically wrong in the overclock settings. Maybe the voltage is spiking or needs a higher or lower LLC setting depending. Another thing to test is if you still have the previous cpu you were using put it back in and run the same tests if it works than it must be cpu related vs contact. 

[airdeano]

kinda hard to decipher, but lop order is:

pump->CPU (wrong inlet port)->GPU->something up front (bayres?)->xflow radiator->something on the back i/o)->pump\from the photo, you prolly have a large air bubble in CPU block depending on your reply of loop order is.

[suprc4]

2 minutes ago, Benjeh said:

 

When did you last clean the block and dont EK make different base plates for different chips?

It's an xspc block, and clean as in what? I've cleaned the thermal paste off of it 3 times now since i remounted it 3 times.

2 minutes ago, Milkcrate336613f said:

Hmm this one is a challenge to fully answer without beingthere hands on. I suspected immediately what the others also said. If its one loop and gpu temped are fine it rules out any restrictive flows during stressing especislly if the gpu is being stress tested too. That tells me hoses, blocks and pump should be fine. Now to me ot leaves eitger the mpunting system is somehow not securing tightly to the faceplate of the cpu which can sometimes be checked with the wiggle test(just be gentle), I would say thermal paste but you've already tried that. I'd also think maybe something is drastically wrong in the overclock settings. Maybe the voltage is spiking or needs a higher or lower LLC setting depending. Another thing to test is if you still have the previous cpu you were using put it back in and run the same tests if it works than it must be cpu related vs contact. 

Yea it has to be the mounting as water is flowing through the cpu block right? I don't think it's the overclock because it shoots up to 95c so fast that it's not even funny. like less than 2 seconds. I've tested in game and it sits at 65c. But for the amount of radiators i have it shouldn't even get that hot.

[Milkcrate336613f]

1 minute ago, suprc4 said:

It's an xspc block, and clean as in what? I've cleaned the thermal paste off of it 3 times now since i remounted it 3 times.

Yea it has to be the mounting as water is flowing through the cpu block right? I don't think it's the overclock because it shoots up to 95c so fast that it's not even funny. like less than 2 seconds. I've tested in game and it sits at 65c. But for the amount of radiators i have it shouldn't even get that hot.

That's where my gut is sitting all I keep thinking is possibly a really small mounting problem maybe something as easy as the wrong screws for the backplate possibly or just something not making full contact. Definitely keep us updated if you find out whats happening and hope some of this gets you in the right direction! Good Luck either way. 

[suprc4]

1 minute ago, airdeano said:

kinda hard to decipher, but lop order is:

pump->CPU (wrong inlet port)->GPU->something up front (bayres?)->xflow radiator->something on the back i/o)->pump\from the photo, you prolly have a large air bubble in CPU block depending on your reply of loop order is.

res/pump combo->gpu->cpu->120mmRad->external360mmRad->internal360mmRad->res/pump combo

 

I did see a massive air bubble this morning in the tube coming from the external radiator which is weird because i made sure to get all the air bubbles out last night. But still I had this issue even before I took apart the loop and added in the gpu waterblock. And i had that loop together for 2 years.

 

Here's a bigger pic.

 

[Benjeh]

So you havent take n the block apart and inspected it to see if all the channels are clear of any gunk or blockages? I'd check there first...

[suprc4]

15 minutes ago, Milkcrate336613f said:

That's where my gut is sitting all I keep thinking is possibly a really small mounting problem maybe something as easy as the wrong screws for the backplate possibly or just something not making full contact. Definitely keep us updated if you find out whats happening and hope some of this gets you in the right direction! Good Luck either way. 

Yea i'll give it a check when I get home. Thanks!

11 minutes ago, Benjeh said:

So you havent take n the block apart and inspected it to see if all the channels are clear of any gunk or blockages? I'd check there first...

I didn't think of that honestly. Didn't realize I could take it apart, nor do I know how. Guess I'll look it up to see if anyone else has taken that waterblock apart.

[For Science!]

3 minutes ago, suprc4 said:

Yea i'll give it a check when I get home. Thanks!

I didn't think of that honestly. Didn't realize I could take it apart, nor do I know how. Guess I'll look it up to see if anyone else has taken that waterblock apart.

What are the temps like if you dont overclock? Stock temps will also be informative.

 

Also, are you using the AVX version of p95? What version number is it.

[Neo-revo]

taking block appart is usally as easy as taking out the visable screws, just remember how to put them back in. 

 

check that the in and out of the cpu block dont have designated in/outs ?  

 

i assume the cpu ran fine with this block before if that what you mean by all the swaps?

[suprc4]

57 minutes ago, For Science! said:

What are the temps like if you dont overclock? Stock temps will also be informative.

 

Also, are you using the AVX version of p95? What version number is it.

36c idle. I always do the maximum heat test on prime95. It's version 29.4 I actually just tried using OCCT after reading that prime95 overheats cpus. Well my temp doesnt go above 66c when occt is putting 100% load on the cpu. but prime goes straight to 95c. So i wonder is it just the prime program overheating my cpu? Because otherwise I don't think theres anything wrong with the waterblock or loop.

55 minutes ago, Neo-revo said:

taking block appart is usally as easy as taking out the visable screws, just remember how to put them back in. 

 

check that the in and out of the cpu block dont have designated in/outs ?  

 

i assume the cpu ran fine with this block before if that what you mean by all the swaps?

Ah ok. In and out is fine, it's routed properly. Yea i've had this block for 5 years and always ran fine until i checked the temps about a week ago, but the last time i checked temps was two years ago.

 

I included an image of the temps while running OCCT

[suprc4]

lowered my overclock to 4.45ghz and reduced the voltage from 1.59 to 1.475 and now temps dont go over 71c with prime96 and they dont go over 54c with occt. So i guess the issue was that i was using prime95 and also my high voltage on my overclock, extremely high voltage i should say. Didn't really care about running super high voltage since ive always had 3 rads to keep everything cool. But lowered it anyway and i don't think i'll see much of a different dropping from 4.6 to 4.45 ghz.

[Mick Naughty]

My 4790k could do the same thing. I just don’t care as real world temps are all that matter. 

[PrimeSonic]

On 10/28/2018 at 1:24 AM, suprc4 said:

lowered my overclock to 4.45ghz and reduced the voltage from 1.59 to 1.475 and now temps dont go over 71c with prime96 and they dont go over 54c with occt. So i guess the issue was that i was using prime95 and also my high voltage on my overclock, extremely high voltage i should say. Didn't really care about running super high voltage since ive always had 3 rads to keep everything cool. But lowered it anyway and i don't think i'll see much of a different dropping from 4.6 to 4.45 ghz.

That's some crazy voltage. No wonder you were seeing those temps.

 

Maybe it's time to pop the lid and get liquid metal in there.

My old custom loop was pretty big for a CPU+GPU loop, having 120+240+360 radiators in there.

The GPU was fine, but the CPU could always jump over 85C on stress tests.

 

I knew the "thermal bottleneck" with the thermal compound between the CPU die and the IHS, but I never realized I would be getting over a 20 degree delta until I did it for myself.

 

If you're going to push that much voltage in your CPU just to get to 4.6Ghz, sounds like you need all the cooling help you can get.

[suprc4]

On 10/29/2018 at 12:56 PM, PrimeSonic said:

That's some crazy voltage. No wonder you were seeing those temps.

 

Maybe it's time to pop the lid and get liquid metal in there.

My old custom loop was pretty big for a CPU+GPU loop, having 120+240+360 radiators in there.

The GPU was fine, but the CPU could always jump over 85C on stress tests.

 

I knew the "thermal bottleneck" with the thermal compound between the CPU die and the IHS, but I never realized I would be getting over a 20 degree delta until I did it for myself.

 

If you're going to push that much voltage in your CPU just to get to 4.6Ghz, sounds like you need all the cooling help you can get.

Yea i know it was crazy voltage for an extra 150mhz. I dropped it down to 4.45 and 1.475v and temps dont really go over 65. However I wanted to try something else. So i cleared the cmos and left everything at stock. Now the cpu idle's at 22c instead of 34c and it doesn't go over 30c in occt. So this whole time it was just my overclock lol. I thought xeons are soldered, so would it even make sense to delid it? I mean at this point i don't really care, 65c is low enough for 1.475v i guess. I just wanted to make sure my loop and cpu waterblock were working properly lmao.

[Benjeh]

good temps at those volts tbh