[SOLVED] 100C watercooled without overclock

[IndycentJester]

I built my first water cooled rig a frw months agio.
I9=9900k
Gigabyte Z390 Master
32 GB DDR4 3200
Gigabyte AORUS GeForce RTX 2080 Xtreme WATERFORCE WB 8G
Thermaltake Pacific M360 Plus D5 Res/Pump PETG Hard Tube Water Cooling
XSPC EC6 High performance coolant
Arctic Silver 5 thermal compound

Without any overclocking, running Prime95 my core temps reach 100C according to CPUID
HWmonitor. Any advice on what I did wrong?

 

[Phaaze88]

Try imgur.com

Also, when you ran Prime 95, did you disable the AVX options? That'll make just about any high power cpu throttle...

 

[IndycentJester]

Try imgur.com

Also, when you ran Prime 95, did you disable the AVX options? That'll make just about any high power cpu throttle...

All I did was select blend all test & ran,

 

[Phaaze88]

All versions after 26.6 have AVX enabled by default. So, if you're running the latest version, all you have to do - Run a Torture Test, but at the bottom, there are checkboxes next to the AVX options:

Although it says, 'not recommended', you really should. It doesn't affect AMD's Ryzen cpus, since they don't discriminate between SSE and AVX: they'll run the same frequency regardless.

 

[IndycentJester]

All versions after 26.6 have AVX enabled by default. So, if you're running the latest version, all you have to do - Run a Torture Test, but at the bottom, there are checkboxes next to the AVX options:

Although it says, 'not recommended', you really should. It doesn't affect AMD's Ryzen cpus, since they don't discriminate between SSE and AVX: they'll run the same frequency regardless.

Seems redundant to run weaker test on new build. Numbers look a little better. Up to 80C, but that stillseems high. Everyone says how easy it is to overclock this to 5GHz, but it will surely overheat.

 

[IndycentJester]

Actually, now it hit 98, so there must be s problem. Well, I knew going in water cooled would be much more maintenance,

 

[Phaaze88]

Seems redundant to run weaker test on new build.

Perhaps, but AVX instructions aren't widely used at the moment. At the very least, I'd suggest setting an offset of 2 or 3 in the bios.
While AVX is faster than the standard SSE instructions at the same frequencies, it does so with higher power and voltage draw, and higher heat output.

Actually, now it hit 98, so there must be s problem. Well, I knew going in water cooled would be much more maintenance,

You may have a problem with the loop itself.
I'm no liquid cooling expert - I only just started getting my feet wet with an AIO mounted on the gpu.

How's the picture coming along? Some visuals would help those who are better versed in this.

 

[IndycentJester]

Perhaps, but AVX instructions aren't widely used at the moment. At the very least, I'd suggest setting an offset of 2 or 3 in the bios.
While AVX is faster than the standard SSE instructions at the same frequencies, it does so with higher power and voltage draw, and higher heat output.


You may have a problem with the loop itself.
I'm no liquid cooling expert - I only just started getting my feet wet with an AIO mounted on the gpu.

How's the picture coming along? Some visuals would help those who are better versed in this.

Already edited original post with pic

 

[Phaaze88]

Ok, I'll hail a few - see what they say.
@rubix_1011 @Karadjgne @grimfox

 

[Karadjgne]

Ok, pump to rad, to ram, to cpu, to gpu, to res. That's fine, flow direction doesn't really make a difference after a few seconds, it all equalizes.

There's a few things you do have to look for though, and thats flow i/o. Check the manuals on cpu/gpu blocks, make absolutely certain they are correct as far as inputs and outputs go, run the coolant backwards or get the jetplates oriented wrong and that component gets next to useless.

Even at 64mm thick, a 2080 + 9900k is a lot of wattage, bordering 500w± and thats a lot for any single rad, especially with rgb fans that aren't exactly built for rads in the first place. You'll be needing to max out rpm pretty much most of the time.

Looks great though.

Games don't use much (if any) AVX technology, that's a set of more specialized instructions used mainly by game developers, CC etc. It's not a weaker test to run without it, it's a more realistic test, as AVX will drive a cpu upto @ 130% ish output in thermal watts, not good on a 9900k that can already hit 250w ish. For a more accurate test of 100% cpu load in a gaming environment, use the small fft torture test and disable all 3 AVX.

Apart from checking the i/o ports directions, I don't see anything really wrong, or might have done wrong, I'm thinking you just have more thermal output than a single 360 can handle at that load.

Oh, and AS5 isn't all that great for thermal paste, it's pretty mediocre overall, just about in the middle of the bunch. I'm not a fan of it with liquid cooling because it will dry out after @ 200 heat cycles (which is not an issue unto itself) but can create issues during yearly maintenance if the dryseal is broken. Think silicon on your bathtub, sticks forever even when dry, but break that seal and it never sticks again. Same deal. So be prepared to reapply yearly.

 

[IndycentJester]

Ok, pump to rad, to ram, to cpu, to gpu, to res. That's fine, flow direction doesn't really make a difference after a few seconds, it all equalizes.

There's a few things you do have to look for though, and thats flow i/o. Check the manuals on cpu/gpu blocks, make absolutely certain they are correct as far as inputs and outputs go, run the coolant backwards or get the jetplates oriented wrong and that component gets next to useless.

Even at 64mm thick, a 2080 + 9900k is a lot of wattage, bordering 500w± and thats a lot for any single rad, especially with rgb fans that aren't exactly built for rads in the first place. You'll be needing to max out rpm pretty much most of the time.

Looks great though.

Games don't use much (if any) AVX technology, that's a set of more specialized instructions used mainly by game developers, CC etc. It's not a weaker test to run without it, it's a more realistic test, as AVX will drive a cpu upto @ 130% ish output in thermal watts, not good on a 9900k that can already hit 250w ish. For a more accurate test of 100% cpu load in a gaming environment, use the small fft torture test and disable all 3 AVX.

Apart from checking the i/o ports directions, I don't see anything really wrong, or might have done wrong, I'm thinking you just have more thermal output than a single 360 can handle at that load.

Oh, and AS5 isn't all that great for thermal paste, it's pretty mediocre overall, just about in the middle of the bunch. I'm not a fan of it with liquid cooling because it will dry out after @ 200 heat cycles (which is not an issue unto itself) but can create issues during yearly maintenance if the dryseal is broken. Think silicon on your bathtub, sticks forever even when dry, but break that seal and it never sticks again. Same deal. So be prepared to reapply yearly.

had planned on push/pull, but was having too much trouble & figured, what difference would it make anyway?
Looks like it is back to the drawing board, Will check I/O and resume my headache of mounting the fans on the bock, Thanks for the info,

 

[IndycentJester]

I may end up changing to CL480

 

[rubix_1011]

Is your D5 pump running at 100%?

The little red dial exists on most D5's, does yours have this? Usually a 1-5 setting, with 5 being 100%.

Otherwise, it might be PWM controlled (4-pin) or the ability to be 100% all the time (SATA or MOLEX power and no PWM).

 

[IndycentJester]

Is your D5 pump running at 100%?

The little red dial exists on most D5's, does yours have this? Usually a 1-5 setting, with 5 being 100%.

Otherwise, it might be PWM controlled (4-pin) or the ability to be 100% all the time (SATA or MOLEX power and no PWM).

I feel pretty dumb for not knowing there was a dial on the bottom. Thanks, I'll try turning it up, but still need to get my pull fans in.

 

[rubix_1011]

Pump speed can make a bigger difference on cooling than a second set of PUSH/PULL fans. If you have very low flow rate, you aren't getting enough coolant movement to make a difference. As long as you currently have flow, this could help make a lot of difference.

If you have this, set it to 5 and see if your temps change. I am hoping there isn't a larger issue of restriction with the many 90 degree angled bends and multiple blocks in the loop, but we'll see how it goes first.

 

[IndycentJester]

Pump speed can make a bigger difference on cooling than a second set of PUSH/PULL fans. If you have very low flow rate, you aren't getting enough coolant movement to make a difference. As long as you currently have flow, this could help make a lot of difference.

If you have this, set it to 5 and see if your temps change. I am hoping there isn't a larger issue of restriction with the many 90 degree angled bends and multiple blocks in the loop, but we'll see how it goes first.

The temps came down a bit, but Prime 95 still took a few cores to 100c. I really hope I do not have to re design my loop because I couldn't think of s better route. Even a larger radiator throws off my look.

 

[IndycentJester]

When my new PrimoChill Vue arrives.I will break down & check my thermal paste.

 

[rubix_1011]

So the pump wasn't running 100% before?

Can you tell how 'fast' the flow is running by looking at the reservoir return line? Looks like in this picture, it is below the reservoir level (which is what you want) but can anything be derived from the flow rate?

When you originally filled the reservoir to fill the loop, did the pump reservoir empty very quickly and coolant move through the tubing very quickly? Or was it sort of lethargic?

If you run a graphics load test (Unigine or something similar) to only load the GPU, does it run very hot also, or does it run cool (since they are both watercooled by the same loop)?

When you start a Prime95 test run, how quickly do temperatures rise to 100C?
When you stop a Prime95 test run, how quickly do temperatures drop back down?
What do CPU temps look like at idle or normal utilization?

 

[IndycentJester]

So the pump wasn't running 100% before?

Can you tell how 'fast' the flow is running by looking at the reservoir return line? Looks like in this picture, it is below the reservoir level (which is what you want) but can anything be derived from the flow rate?

When you originally filled the reservoir to fill the loop, did the pump reservoir empty very quickly and coolant move through the tubing very quickly? Or was it sort of lethargic?

If you run a graphics load test (Unigine or something similar) to only load the GPU, does it run very hot also, or does it run cool (since they are both watercooled by the same loop)?

When you start a Prime95 test run, how quickly do temperatures rise to 100C?
When you stop a Prime95 test run, how quickly do temperatures drop back down?
What do CPU temps look like at idle or normal utilization?

When I originally filled, was fast. Prime 95 reached 100c in 10 min. Back to 30's in prob less than a min. Idle Temp is 32c. I have not done load test yet on video yet.

 

[IndycentJester]

I ran Heaven Benchmark 4.0, but could not get past 46c on video card.

 

[rubix_1011]

Both of your last 2 posts each provide an indication that is helpful.

GPU maintaining 46C at load is very good and indicates that the loop is flowing well. If it was flow rate and restriction, this would be much higher. I assume this benchmark was run for at least 10-15 minutes?

100C on Prime95 over the course of 10 minutes also tells a story, and one that likely might be AVX settings on the CPU both in the BIOS and Prime95 setup. Run Small or Smallest FFTs and check the box to disable AVX2. You might also check BIOS to see if there is an AVX offset config.

Back down to idle within a minute and idling around 32C is also another story. It means that your block is likely setup and doing its job correctly, although I would almost expect this to be faster - within 10-20 seconds, depending on ambient room temps. I would expect higher idle temps or higher temps being noticed during normal tasks...Prime95 really is a torture benchmark for CPUs.

Something which hasn't been mentioned:

Do you have any overclocking setup on this CPU? If so, can you confirm the vCore and other BIOS settings being used? Wondering if it isn't an issue of an OC setup with higher-than-needed voltage like you'd see with a BIOS % OC slider.

 

[IndycentJester]

Both of your last 2 posts each provide an indication that is helpful.

GPU maintaining 46C at load is very good and indicates that the loop is flowing well. If it was flow rate and restriction, this would be much higher. I assume this benchmark was run for at least 10-15 minutes?

100C on Prime95 over the course of 10 minutes also tells a story, and one that likely might be AVX settings on the CPU both in the BIOS and Prime95 setup. Run Small or Smallest FFTs and check the box to disable AVX2. You might also check BIOS to see if there is an AVX offset config.

Back down to idle within a minute and idling around 32C is also another story. It means that your block is likely setup and doing its job correctly, although I would almost expect this to be faster - within 10-20 seconds, depending on ambient room temps. I would expect higher idle temps or higher temps being noticed during normal tasks...Prime95 really is a torture benchmark for CPUs.

Something which hasn't been mentioned:

Do you have any overclocking setup on this CPU? If so, can you confirm the vCore and other BIOS settings being used? Wondering if it isn't an issue of an OC setup with higher-than-needed voltage like you'd see with a BIOS % OC slider.

The reason I even noticed my temps was because I was intending to overclock, but when I did, computer reset.
Was already running prime with AVX disabled. I ran benchmark until test ended. That was the only option free edition Heaven with a score of 4548, which I am sure is prob irrelevant.

 

[rubix_1011]

Overclocking failure is common and mostly expected. When an OC fails, you make BIOS adjustments and try again. It isn't a one-try-success kind of thing.

 

[IndycentJester]

I am getting conflicting temperatures from AORUS system viewer and CPUID. 10 degree difference. Which on should I trust?

 

[EndEffeKt_24]

Are you sure that your waterblock is making proper contact with the cpu? 100c even with prime95 sounds too much to believe that the thermal transfer is working properly.
Do you have a watertemp probe?