Question Updating water cooled build.

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
My current setup has xspc d5 pump combo, 2 360 rads and 1 280 rad, 6 fans for push pull on 360 rads and pull on 280 rad below to bring in cold air, have a ek monoblock to cool cpu and vrm on asus rog vi apex mobo and bykski gpu block for zotac 2080ti and my flow rate is only showing as 130 lph and the pump specs are 1200 lph and i have long tubes and lots of 90 bends probably decreasing flow, so i want to remove all rads and put a 100 to 120 w peltier cooler between pump and cpu and after gpu and before reservoir so tubing would be much shorted with less bends and not many 90 bends, i have also purchased evga 1600w t2 psu to meet my new needs and with all the components my pc will be pulling around 1300w on full 5ghz OC load, is this a good idea, please guide me, thank you.
 

tennis2

Respectable
There's no way your PC is drawing 1300W with a single RTX2080Ti, even with a 5GHz CPU.

Focusing on the flow rate problem, since it seems to be the reason for all these changes. I can't fathom a pump only being able to output 10% of it's rated flow. What pump do you have? Are you sure that the pump is being fed the proper 12V voltage and not being run at say...25% duty cycle? How is the pump being powered?
 

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
There's no way your PC is drawing 1300W with a single RTX2080Ti, even with a 5GHz CPU.

Focusing on the flow rate problem, since it seems to be the reason for all these changes. I can't fathom a pump only being able to output 10% of it's rated flow. What pump do you have? Are you sure that the pump is being fed the proper 12V voltage and not being run at say...25% duty cycle? How is the pump being powered?
@tennis2 When i am stress testing cpu at 5ghz and i have meter in mobo to see power draw cpu alone draws between 500 to 550w and add gpu to it it totals to 800w constant, i have xspc d5 photon pump combo and its connected to sata power from psu and there are no speed settings so i believe its the loop that restricting flow, i have a total of 15ft to 17ft tubing with close to 20 to 24 90 degree bends and cpu and water block and 3 rads, i said 1300w adding the future peltier swtup which draws lot of wattage.
 

rubix_1011

Contributing Writer
Staff member
2080Ti is a 250watt TDP part.

As comparison, I ran our cooling workbench 2990WX, 32-core AMD Threadripper at 4.2Ghz and 1.412vcore and it was showing a draw of 855 watts at the wall outlet. So, somewhere your math is wrong.

And why so many radiators? I mean, nothing says you can't use them, but why spend the extra money on them when you probably need half of that?

Also, if you have so many 90-degree bends...remove them. This sounds like laziness rather than necessity. 'Long tubing' doesn't provide much restriction, it's tight bends and restrictive blocks.

And the 1 peltier likely isn't going to be enough to replace all your radiators, so you're going from overkill to anemic.
 
Mar 27, 2019
52
7
35
0
This is insanity.
The peltier elements are absolutely inefficient as far as I know. Your 1300 watt powerdraw is not accurate to say the least. On top of that we are talking exotic and horrendously expensive cooling here without a single mention of temperature. How hot is your system?

I would start with:
  • removing the 280 radiator
  • checking the pump and the flowmeter
  • put it back together and go get some temps
Cheers
 

rubix_1011

Contributing Writer
Staff member
Loop design and misunderstanding of wattage is the biggest issue here, so far.

As others have asked - what is the main driver for this elaborate setup, how watts are being calculated and what the purpose the TEC/peltier units intend to solve?
 
Mar 27, 2019
52
7
35
0
I just re-read the posts and stumbled over the op mentioning 24 90 degree turns on 17 ft of tubing...
That sounds all rather odd and leaves me with an urge to see pictures of that setup.
How on earth are you putting 5 m of tubing in your case?
 

rubix_1011

Contributing Writer
Staff member
Are we talking 24x 90-degree fittings? Or is this hard tubing that is bent? This isn't being communicated, either, which both react differently - fittings are a bit more restrictive where hardline tubing bends are generally smooth and less restrictive in a matter of speaking.

I think a photo might help address some of these as well as perhaps some additional detail in writing. A lot of detail is being left out.
 

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
I just re-read the posts and stumbled over the op mentioning 24 90 degree turns on 17 ft of tubing...
That sounds all rather odd and leaves me with an urge to see pictures of that setup.
How on earth are you putting 5 m of tubing in your case?
@EndEffeKt_24 i have TT core x9 case which is huge, i have my pump in right back corner and from there tube goes up and turns 90 and then another 90 to get to cpu block and its about 2 to 3 ft long and then another ft long with 2 90 bends to go to rad above and then another ft long tube with 2 90 to gpu and another ft long with 1 90 bend to rad above and then 4ft long with 3 90 and a T fitting between and a thermal and wateflow sensor also in between to a rad below and then another 2 ft long tube with 2 90 to go to reservoir, i couldnt figure out another way as the case was large, pictures in the link. https://www.dropbox.com/sh/b2cybm6lq2b5zgh/AACe5xjrjA_2T7NCGFcvB9P-a?dl=0
 
Mar 27, 2019
52
7
35
0
Thank you very much for the pics. Now we all know a little better what we are talking about. That thing is a behemoth.

I think your flow of approx. 2l/min could be correct considering the height differences, water volume and radiator restriction of the setup. Your pump got a lot of work there.
Are your temps actually bad? I could imagine the huge radiator space is making up for the flow just well.
 

rubix_1011

Contributing Writer
Staff member
That's the second time in the past week I've seen that flow meter - so.......wondering if I need to pick it up and do some testing on it? It looks really nice, but I wonder how accurate it really is. Actually...this is the same build, that's why. So this is a duplicate thread around the same topic? https://forums.tomshardware.com/threads/flow-rate-severely-restricted.3462527/

It is a nice build and looks like the majority of your 90-degree bends are hardline tubing bends, which should not introduce much restriction, which is good.

The CPU watt curve might be close, depending on what your actual vCore is set at.

Stock TDP for a 2080Ti is 250watts, I believe, so really depends on how much voltage is applied and how much of an OC you are running. Might be a good estimate at 300w.

So, what is the real basis around the questions being asked? Is it that you are not as happy with the CPU temps at current overclocks? Or do you feel something is wrong? Or you feel the flow rate is bad and causing temps to suffer? I thought we had determined that GPU temps are fine and CPU at stock speeds is good, just seems higher when overclocked and running Prime95? (which isn't a very accurate way to determine real-world loads).

In short - what are we trying to troubleshoot and resolve?
 

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
i9-7900X @ 4.8GHz = 260W extrapolate that curve out to say ~300W @ 5GHz.
Best I can find is that an OC'd RTX2080Ti draws....350W(?)
That's the second time in the past week I've seen that flow meter - so.......wondering if I need to pick it up and do some testing on it? It looks really nice, but I wonder how accurate it really is. Actually...this is the same build, that's why. So this is a duplicate thread around the same topic? https://forums.tomshardware.com/threads/flow-rate-severely-restricted.3462527/

It is a nice build and looks like the majority of your 90-degree bends are hardline tubing bends, which should not introduce much restriction, which is good.

The CPU watt curve might be close, depending on what your actual vCore is set at.

Stock TDP for a 2080Ti is 250watts, I believe, so really depends on how much voltage is applied and how much of an OC you are running. Might be a good estimate at 300w.

So, what is the real basis around the questions being asked? Is it that you are not as happy with the CPU temps at current overclocks? Or do you feel something is wrong? Or you feel the flow rate is bad and causing temps to suffer? I thought we had determined that GPU temps are fine and CPU at stock speeds is good, just seems higher when overclocked and running Prime95? (which isn't a very accurate way to determine real-world loads).

In short - what are we trying to troubleshoot and resolve?
Yes i have created thread a week ago about flow restriction but since then i thought i might need to update the loop with less tubing and maybe 2 150w peltier coolers for more silent operation, i dont use wall meter but a meter hooked to my mobo that shows the draw and i believe i had few pictures of that some place that i can find but dont have the meter now, i had the 3rd rad in the bottom because i thought it might help to bring in cool air so rads above are fed cool air and dissipate heat better, at stock 7900x temps are around 55 to 60 and this time i used intel burntest to stress cpu reasonably but at 5ghz oc temps spiked to 93 max and i cant figure out why my loop is not dissipating heat and the flow meter i recently added answered my question, i am just doing this for a hobby and don't mind messing around with the setup, i just want to increase flow rate, when i first installed the meter it was showing around 122lph and i use only distilled water so took the loop apart cleaned all the blocks, rads and fittings to make that is not causing it and now i get 132 to 136lph flow rate and i guess i might have done better job at bleeding air this time and this is only the 2nd build i ever did, just want to get to get to bottom of flow rate issue, i saw lot of videos on youtube from jayz2cents and other builders and they all mentioned that tube length, T fittings and 90 bends add to restrictions along with the blocks so i was stressing about them here, my 7900x is stable with 1.27v for 8 cores at 4800mhz and 1.31 for 2 cores at 5ghz and my main concern is why is not heat dissipating as i believed it would with this, thats all, i am not trying to spam here.
 

rubix_1011

Contributing Writer
Staff member
Don't worry, I know you aren't trying to spam us or cause any riff-raff...just want to make sure we're all understanding the issues so we can do our best to help.

90 degree fittings can cause a lot of restriction, but the gentle bends of hardline tubing don't have nearly as much restriction by comparison. Your CPU core temps are being reported at the individual cores themselves and represent the instant in time where that is recorded...but does not necessarily mean that is what your coolant temperature represents. In theory, if those temps remain at load and constant, your coolant temp should rise, but only by the difference of the delta-T of the cooling loop's capacity.

Also, you mention using a D5 pump - have you checked if it is running at full speed? Most D5's are variable speed and newer ones are even PWM controlled. So, if you have the pump set too low (manually) or if you have it set to be PWM controlled, are you sure it is running as fast as it can? I run a D5 from EK and I run it at full speed with the PWM control header disconnected. Can you confirm how you have your pump setup?
 

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
@rubix_1011 i have xspc d5 photon 170 pump/res combo and i confirm its powered by sata power from psu, it has no speed switch on the bottom and there is no pwm option for that model. And my coolant temps are always below 30c and never went past 30c as my PC is in basement with heat from house heater far away, once i stop the test the heat comes back down to around 25 to 27 on each core and about 30c for the package and gpu is never an issue as it always below 50c at any level of usage. And my bad i made big mistake here, i always thought i had d5 when i purchased but looking at pics on xspc website i might instead have x4 pump thats rated at 600lph.
 
Last edited:

rubix_1011

Contributing Writer
Staff member
x4 vs D5 is a big difference overall for flow rate per hour. x4 is rated at 600 liters per hour (as you mentioned) and most D5's are rated between 1200-1500 liters per hour, or roughly twice the rated rate.

Also, head pressure is important in a restrictive loop:

x4 is rated at 3 meters of head, D5 is 3.8-3.9 meters.

You might have just discovered the source of truth for your questions.
 

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
x4 vs D5 is a big difference overall for flow rate per hour. x4 is rated at 600 liters per hour (as you mentioned) and most D5's are rated between 1200-1500 liters per hour, or roughly twice the rated rate.

Also, head pressure is important in a restrictive loop:

x4 is rated at 3 meters of head, D5 is 3.8-3.9 meters.

You might have just discovered the source of truth for your questions.
@rubix_1011 and @EndEffeKt_24 so i updated my loop to include swiftech maelstorm d5 x100 pump/ res combo and 2 alphacool 360 monsta rads and removed bottom rad and the pump is at full speed and my meter still shows only 185lph at pump full speed, not sure if the sensor is showing it wrong or whatever and i removed it as it started to leak a little, any ideas appreciated.
 

Gmoney06ss

Reputable
Jul 3, 2015
595
84
5,090
13
Was there any temp improvements with the new set up?

Is there any type of calibration required for the flow meter? Can you move the flow meter closer to the outlet of the pump to measure flow before any restriction in the system?
 

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
@EndEffeKt_24 Ok with intel burn test for 10 minutes with 49ghz oc temps reached max of 75 to 80, with prime95 it reaches 90 to 95 easily, @Gmoney06ss there was no change in temps significantly with new build, gpu cools great never goes past 50c on full load its the cpu and the sensor was leaking and so i removed it and have a spinner instead.
 

Gmoney06ss

Reputable
Jul 3, 2015
595
84
5,090
13
So, nothing you've changed has really netted you any positive results? Other then turning down the clock?

Like I mentioned in the other thread, adding a second pump may help. But after reading all of this, I don't really think it will.

Have you tried to reseat the block with fresh paste?

It seems like your system is working, and you may just need to settle for a less extreme clock on the cpu.
 

lingarajug

Honorable
Dec 21, 2013
88
1
10,635
0
Yes i have reseated my cpu with clu liquid tim applied between dye and ihs and ihs and block, no problem there, i might have installed a ddc pump that has 7m head pressure like ek ddc 3.25, idk and not sure if that might have helped either and i might just not have good binned chip with less than 1.3v to stabilise oc, only 1.315 to 1.320 stabilises 5.0ghz oc.
 

ASK THE COMMUNITY

TRENDING THREADS