Temps High on Custom Loop

  • Thread starter Thread starter mk1_salami
  • Start date Start date
mk1_salami said:
The 4 front fans are intake, and the rest are exhaust (2x 120mm case fans) and 2x Static Pressure 140mm radiator fans.

Since fixing the inlet / outlet on the 3090 I'm seeing max GPU temps around 64c with 3dmark on loop. The CPU still hits 90c almost immediately on Cinebench r23 but the boost speeds are strong (4.55ghz) and a multi-core score of 22500, which seems about right. I'm using PBO with -15 curve.

I guess I could drain the loop (COMPLETE PAIN IN THE ARSE) and re-seat the CPU block. Might even be worth checking contact on the GPU too. I just hate draining this loop and re-filling.
Click to expand...

Unlucky, that's one of the reasons I avoid hardline (plus it's more difficult to do of course).
 
Are you able to read coolant temperature?

And that's one drawback with hard Tubing I guess
With soft Tubing I have quick disconnects
So can isolate and remove parts of the loop if needed
So can reseat cpu or gpu etc without draining everything
 
Unfortunately that's the only way you're going to fix this (by draining and reseating the CPU cooler).

I almost ended up in a similar situation on my last build. 3x 360 rad, CPU and GPU all hardline tubing... went to fire the system up and the front panel connectors (inc. power) were not wired correctly because of the gigabyte "easy install" connector showing the markings in the wrong orientation for my motherboard :rolleyes:. With the GPU being vertically mounted I had to squeeze my hand between the bottom rad and the GPU and try to disconnect the power cable header or face a full drain just to reverse a tiny cable.

Thankfully I managed to get it done without removing everything but in this instance, you know the CPU isn't making full contact if it's jumping to 90C+ on load.
 
mk1_salami said:
The 4 front fans are intake, and the rest are exhaust (2x 120mm case fans) and 2x Static Pressure 140mm radiator fans.

Since fixing the inlet / outlet on the 3090 I'm seeing max GPU temps around 64c with 3dmark on loop. The CPU still hits 90c almost immediately on Cinebench r23 but the boost speeds are strong (4.55ghz) and a multi-core score of 22500, which seems about right. I'm using PBO with -15 curve.

I guess I could drain the loop (COMPLETE PAIN IN THE ARSE) and re-seat the CPU block. Might even be worth checking contact on the GPU too. I just hate draining this loop and re-filling.
Click to expand...

This thread has got me worrying about my own CPU cooler so I checked my notes. I remember that my Cr23 scores seemed a bit low and research suggested that I had made a bad choice of system memory with high latency. However with the stock settings the CPU temperature under Cr23 was around 72C which a web search suggested was typical. I was worried that in idle and low load the CPU varies around 40-45C but again this seems typical for a 5900X. When I tried the CPU undervolting I saw a significant variation in peak CPU temperature depending on my PPT/EDC/TDC limits. When I put them too high the temp under Cr23 rose to between 80C and 90C but I got better scores. I got focussed on just how far I could push the undervolt curve and maybe I should revisit it. I think I'm letting my CPU get hotter than it needs to for not much gain. Of course the 5900X only throttles above 90C so in theory it can run at that temperature just fine.

But my 3090 rarely exceeds 40C under load so yours doesn't look right. If I have understood the information provided you previously had no coolant flow through the waterblock and the GPU reached 68C, and now that you have flow it still reaches 64C?
 
You got me worried about the GPU. I just checked the latest installation instructions for the waterblock and they say to use ONLY 1mm pads. I could have sworn when I received that the instructions said to use a mix of pads on each component. It has me thinking I got one of the very early units with wrong install instructions (I waited months for it to be released and was one of the first to receive it). This would make sense because LTT had the same issue on an early EK block on a 3090.

Also worth noting the card has the active backplate too.

It seems to me the thing to do is:

Order 1mm thermal pads
Drain loop
Reinstall Waterblocks with 1mm pads on GPU
Re-seat CPU waterblock with new thermal paste.

Fill & cross fingers.
 
Last edited by a moderator:
Black Noir said:
If he has 2x exhaust then having all the radiators as intake would actually create positive pressure and be better for temps (and dust!).

You want more intake than exhaust with watercooling as a rule of thumb.

68C under load is high but doesn't suggest an issue with the GPU. That may well be the temp after hammering a synthetic benchmark for 2hrs in which case I'd say it's reasonable. I have that exact card under water and it peaks around 50C using triple 360 rads. Also, if the CPU block isn't mounted correctly then don't forget he's outputting 90C+ heat in to the loop which will be causing the whole loop temp to rise.

Check the CPU as that's the smoking gun here.
Click to expand...
Quite a lot of inaccuracy here so I'm gonna have to try and clear some of it up...

68C under load is high for a GPU but the real issue here is about delta between silicon and liquid. Typically a GPU under load is only going to see 10, maybe 15° difference between the liquid and the silicon. Whereas a CPU with the same liquid temp is likely to be anything from 15-30°. So really I'd expect to see 15° difference roughly between GPU and CPU, under load, for a given liquid temp. If the CPU is hitting 90+ immediately then that suggests the issue is with the mount. Whereas the GPU is high but not insanely high, so more likely a liquid temp or flow issue as there is some liquid absorbing heat and carrying it away by conduction (even if no flow).

Now, the temperature of the CPU doesn't matter much to the loop heat - it's the power output of the CPU under load, in watts. The CPU hitting 90 immediately is because of a poor mount and/or no flow to carry the heat away. But it's still only kicking out (effectively) the same power in watts as if it was running at 60° under load with good coolant flow etc. As soon as that mount and flow rate are corrected the temperature would drop immediately as the extra stored energy would dissipate in such a large volume of water. It's not going to heat the liquid up much more than if it were at 60.
 
mk1_salami said:
The CPU still hits 90c almost immediately on Cinebench r23 but the boost speeds are strong (4.55ghz) and a multi-core score of 22500, which seems about right. I'm using PBO with -15 curve.
Click to expand...
My settings

bm1xO7D.jpg

Rig had been idling 18 minutes, then I did a R23 multi run 22479, then R23 single run, then another R23 multi 22526 in the screenie below.

4CpPbln.jpg

WC System specs

Bykski A-Ryzen-ThV2-X CPU Block
Bykski A-SP6800-X GPU Block
Arctic Silver 5 TIM on both blocks

EK-XRES 140 Revo D5 PWM
EK-Tube ZMT (5/8, 3/8)

2x Magicool 360 Slim G2, top rad as exhaust push only 3x Arctic Cooling F12 PWM, front rad as intake push/pull 6x Arctic Cooling F12 PWM.

Coolant distilled water with 20% mix Mayhems XT-1 Clear Concentrate
 
Last edited:
Did you redo the cpu when you did the gpu so that the inlet to the cpu is the actual inlet (left side) and not reverse flow like it was in your original pictures? While you are draining it down add a inline water temp sensor. These should be a must have for every custom water loop so that you can keep a eye on coolant temps. There are various types available and if your motherboard has two pin temperature sensor headers you can pick up a sensor that screws into a unused port (or T-piece added to the loop) and then connect to the motherboard header for around a fiver or less. Another type is a inline sensor that typically fits between a pair of fittings or a fitting and a block/reservoir/radiator and connects to a motherboard 2 pin header or Phobya does a version that has a small LCD screen attached to the sensor by wires. These are typically around a tenner. I personally use a Alphacool temp sensor that screws into a spare port on my reservoir and the temperature of the coolant is read from a LCD screen I have mounted above the reservoir. They are well worth adding to a loop.
 
@pastymuncher
I asked about the loop order
But the op didn't confirm what it was
But in his picture he is using the left inlet on cpu block
If I followed it correctly
Since left is the outlet on the xd3

The xd3 also has a temperature sensor on it
But questions about coolant temperature also weren't answered
So no idea if they can monitor that or not
 
Ok, noticed the video now so either a bad mount or a pump problem then. Can he monitor the pump speed? How about setting the pump to full speed (presume it's on a pwm header) to see what the temps are like then?
 
I actually missed the video too
So apologies for that
So that answered my question about the loop route
Which was same as I asked in my post

At the beginning
He said both the fans and the pump
We're at full blast so i assume yes he's able to monitor
The pump speed
 
Mcnumpty2323 said:
@pastymuncher
I asked about the loop order
But the op didn't confirm what it was
But in his picture he is using the left inlet on cpu block
If I followed it correctly
Since left is the outlet on the xd3

The xd3 also has a temperature sensor on it
But questions about coolant temperature also weren't answered
So no idea if they can monitor that or not
Click to expand...
Sorry the loop order is:

Xd3 left outlet > front rad > gpu > too rad > cpu inlet left side with arrow > xd3

Water temps don’t exceed 34c during tests when I see GPU at 64c and CPU at 90c
 
Last edited by a moderator:
I missed your video somehow before
But yeah looks like you're using the correct inlet/outlet on it all

And was just curious about coolant temperature
In case it gave any hints
Like not enough flow
Or a bad cpu mount

I did wonder about that pumps abilities
Guessing it's a ddc in there
So won't have the same head as a d5
But I haven't used a ddc for a loop that size so may be it's no issue anyway
Just trying to cover all possibilities really
 
If your water temp is only 34 degrees (which is excellent) then there has to be a problem with your cpu block mounting surely? From my testing in the past, water flow doesn't make much difference to temps, maybe a degree or two at the very most so even if the pump is weakish it shouldn't hurt the temps that much. Could you have used too much thermal paste and is the block tightened fully down? I know it's a pain with hard tubing but I would be doing a remount of both the gpu and cpu blocks to see if it makes a difference.

***Edit***
I just looked up some reviews of the pump and it is a DDC with performance of 380 LPH at 2.1 m head pressure which does seem rather low. There is also a person that left a review on Amazon saying that his 5800x was running hotter with this pump even after splashing out on a pair of high end rads to attempt to get the temps down (it sounds like he bought the pump for the looks as it looked nicer than his old one). If I remember right from the days that everyone was buying DDC's, didn't these pumps run very hot and many had to buy heatsinks to cool them? I wonder if the pump is dumping heat to the coolant? In any case, it seems to be a poor pump that is all looks and no go which is not good seeing as the cheapest I have seen it is £135.
 
Last edited:
After more testing … the gpu temps are actually higher now than before! Just found out the inlet and outlet is supposed to be like that on the GPU when you have the ek active backplate. D’oh!!!

Reckon I’m gonna drain the whole thing. Re do the thermal pads on gpu and new thermal grizzly on the cpu block.
 
Though its a PITA to do
That may be the best bet

Maybe fit a flow meter at same time?
That dx3 with ddc pump is kind of low head compared to
A d5 anyway
And it's quite a big loop
So being able to see flow rate could be useful

It's probably not the source of the problem
As even some flow wouldn't likely immediately shoot the cpu
To 90c it would take time to saturate the loop
I am just a belt and braces guy
So tend to try to cover most possibilities
 
I bought some Gelid Extreme pads for the GPU memory and Thermal Grizzly Kryonaut for the CPU and GPU core. Should arrive Monday so will see how I get on next week. Not cheap though! Including new fluid around £85 for all the necessary pads. Urgh.
 
Thermal pads are quite expensive for some time now. At least with the Gelid ones you sure won’t be limited by them. The actual EK is okay, but nowhere as good. In the past, when the prices were more reasonable, I used to replace all the thermal pads on my new motherboards. Night and day difference. But luckily, the actual MSI I’m using already have good ones, so I’m not tempted in doing it now.
 
pastymuncher said:
***Edit***
I just looked up some reviews of the pump and it is a DDC with performance of 380 LPH at 2.1 m head pressure which does seem rather low. There is also a person that left a review on Amazon saying that his 5800x was running hotter with this pump even after splashing out on a pair of high end rads to attempt to get the temps down (it sounds like he bought the pump for the looks as it looked nicer than his old one). If I remember right from the days that everyone was buying DDC's, didn't these pumps run very hot and many had to buy heatsinks to cool them? I wonder if the pump is dumping heat to the coolant? In any case, it seems to be a poor pump that is all looks and no go which is not good seeing as the cheapest I have seen it is £135.
Click to expand...

Not sure what the deal is with the DDC in the XD3. Its supposedly a DDC 3.2 but specs says it only does 380 LPH and 2.1m head pressure. But a DDC 3.2 should have head pressure around 5.5m which is quite a bit more than a D5, but a lower flow rate of about 900LPH if I remember correctly.
 
You must log in or register to reply here.
Back
Top Bottom