Looping CPU and GPU

Rezengyal

Junior Member
Jan 12, 2012
10
0
0
Hello all!

I have yet to attempt a watercool, so I am presenting you all with a relatively simple question: one a single loop, is it best to cool the CPU before the GPU or vice versa?

Put another way, which device should come first in the cooling loop? The CPU or the GPU?

And honestly, I'm not 100% sure how the loop would go in the first place. Assuming just cooling a CPU, is it: Resevoir -> CPU ->Radiator -> back to Resevoir?

Again, just a few questions from a watercooling novice!
 

Rezengyal

Junior Member
Jan 12, 2012
10
0
0
blckgrffn,

Actually I was reading that as I posted this thread! My questions stem directly from having read that guide; maybe I haven't reached a section where these questions are answered in it.
 

Rezengyal

Junior Member
Jan 12, 2012
10
0
0
Alot of "views" but very little discussion. Anyone have any insight? Two relative novices here, looking for some knowledge! ^_^
 

chimaxi83

Diamond Member
May 18, 2003
5,457
63
101
I'm piecing together my own system now, and after reading all over the webz and looking at countless pictures, I almost always see CPU being fed first.

It's most likely because the CPU will be dumping less heat into the loop than the GPU will. That's my thought on it, anyway.
 

WoodButcher

Platinum Member
Mar 10, 2001
2,158
0
76
It really doesn't matter, the water temp will stabilize and at that point the difference before and after any one block is almost immeasurable. This has been tested and argued to death. What it comes down to is preference and aesthetics. The biggest reason to put one in front of the other would be to keep the loop short and kink free. Res to pump to rad to block (s) and back to the res is my happy loop.
 

Rezengyal

Junior Member
Jan 12, 2012
10
0
0
Res to pump to rad to block (s) and back to the res is my happy loop.
So does your setup accumulate a lot of heat in the reservoir? It seems your setup is sending "chilled" water through the blocks rather than drawing warm temps from the blocks and up to the radiator to be dissipated as a first step.

Interesting.
 

blckgrffn

Diamond Member
May 1, 2003
9,676
4,309
136
www.teamjuchems.com
So does your setup accumulate a lot of heat in the reservoir? It seems your setup is sending "chilled" water through the blocks rather than drawing warm temps from the blocks and up to the radiator to be dissipated as a first step.

Interesting.

I agree. I would have thought something like Res-pump-blocks-rad-res. Hah, I truly know nothing here :p
 

OVerLoRDI

Diamond Member
Jan 22, 2006
5,490
4
81
It really doesn't matter, the water temp will stabilize and at that point the difference before and after any one block is almost immeasurable. This has been tested and argued to death. What it comes down to is preference and aesthetics. The biggest reason to put one in front of the other would be to keep the loop short and kink free. Res to pump to rad to block (s) and back to the res is my happy loop.

I disagree with this. It does make a difference, just a small one. In general the component with the lowest max temp should be put first when possible.
 

WoodButcher

Platinum Member
Mar 10, 2001
2,158
0
76
As I mentioned about the water temps in the loop pertains to the rad as well. The temp before and after the rad are almost immeasurable. Each block adds to the overall temp, the water moves too fast to accumulate any noticable temp in any one part of the loop. If you have high speed Delta fans, you know, the ones that fly if not bolted in? put those on a Mora 360 or 480 in a push pull configuration you you will be able to measure a temp difference, not any rad you can fit into a case.
As for the resevior it is there mainly to supply the pump with water, always mounted above and before the pump. Bleeding while much easier with a res can be accomplised nicely with a Y and fill tube/ drain if done right.
 

WoodButcher

Platinum Member
Mar 10, 2001
2,158
0
76
I disagree
I've read of tests and arguments since I dove into water. All I have ever seen myself has shown such a small difference that the shorter neater loop will perform overall better.

Pssst,
hey,:cool:
don't tell anyone but if I run a single loop,,,,:sneaky:
,,,,,,,,,
I run the CPU first!:p
 

OVerLoRDI

Diamond Member
Jan 22, 2006
5,490
4
81
I've read of tests and arguments since I dove into water. All I have ever seen myself has shown such a small difference that the shorter neater loop will perform overall better.

Pssst,
hey,:cool:
don't tell anyone but if I run a single loop,,,,:sneaky:
,,,,,,,,,
I run the CPU first!:p

I just think of gpus in serial. The 2nd gpu is warmer than the 1st.

Also consider that water entering the radiator and leaving the radiator is at a different temperatures.

The difference decreases as flow rate increases, but it is still there. I'm not sure what flow rate is high enough though. I have not researched this in depth.

Edit: I just noticed your comment about prioritizing loop neatness/reducing tube length over order. Ill mostly agree. Just make sure your flow rate is high enough.
 
Last edited:

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,250
136
Res > Pump > Whatever works best for you after that.

There is no set in stone order for a loop. Too many variables such as case, rad placement, pump and res options to make a this loop fits all.

My loop is Res > Pump > Upper Rad > CPU > Rear Rad > GPU > Res

In my current case it work best for me. Debating if I wanna switch to the Corsair 500r at the moment and would require a redo on the order most likely.

Temp wise the GPU being after the CPU doesn't seem to effect the GPU's temp too much.

Doing a unrealistic senario as far as loads go using Furmark and Intel Burn Test at the same time my GPU hits the low 40's and the CPU hits mid 60's on the hottest core.

Not into gaming anymore but I did do a couple of runs with AVP benchmark and the Stalker benchmarks to see what temps would be like in a semi real world senario. Using the AVP with my CPU at 4.8ghz and GPU at 1000/1300 the CPU was 42*C on the hottest core the GPU peaked at 36*C
 

WoodButcher

Platinum Member
Mar 10, 2001
2,158
0
76
I just think of gpus in serial. The 2nd gpu is warmer than the 1st.


Reverse the flow one time w/o changing the loop and see if this still holds true, I'll bet the temps don't reverse..




Edit- I've never done SLI or crossfire, just a hunch,,,,,
 
Last edited: