System Light Dark

xentr_theme_editor

xentr_page_setup xentr_appearance xentr_sidebar

  • xentr_page_width

    xentr_toggle_page_width

    xentr_page_width_description

    xentr_color_pickers

    xentr_toggle_color_picker

    xentr_toggle_color_picker_description

    xentr_typography

    xentr_select_the_font_you_want_to_use

    xentr_node_layout

    xentr_grid_layout

    xentr_grid_layout_description

  • Style variation

    xentr_select_the_style_variation_you_want_to_use
    System Light Dark

  • xentr_sidebar

    xentr_toggles_sidebar

    xentr_sidebar_description

    xentr_sticky_sidebar

    xentr_sticky_sidebar_description
    xentr_sidebar_position
    Left
    Right
  • Please do not post any links until you have 3 posts as they will automatically be rejected to prevent SPAM. Many words are also blocked due to being used in SPAM Messages. Thanks!

How do GPU Loops like this one work?

Ashihtaka

Ashihtaka

Well-known member
Joined
Apr 2, 2014
Messages
84
Reaction score
0
xentr_thread_starter
I've been seeing many parallel builds recently where the tubing goes through one side of GPU into CPU and out the other side.
For example, this build from JayzTwoCents here.

How does the GPU receive enough cooling this way? I would assume from the pump it goes straight through to CPU and then back out to the radiator with only a trickle of water going into the GPUs.
Anyone able to explain how this is cooling the GPUs sufficiently?
Or is the entire premise of this type of parallel loop to have better aesthetics for lower cooling efficiency in GPU?

jBSmmp4.png
 
Last edited:
Te flow of coolant thru the loop equalizes within the entire loop...regardless of whether the GPU or CPU come first or last in the sequence.....in other words if you measure the coolant temperature as it exits the GPU or exits any Radiator...it will be pretty much the same... :)
 
xentr_thread_starter
rjbarker said:
Te flow of coolant thru the loop equalizes within the entire loop...regardless of whether the GPU or CPU come first or last in the sequence.....in other words if you measure the coolant temperature as it exits the GPU or exits any Radiator...it will be pretty much the same... :)
Click to expand...

I understand that is the case over time, but what i'm wondering is how are the GPUs even receiving enough flow of the liquid to be a consistent part of the loop?
My initial thought is that most of the liquid is bypassing the GPU and going straight through to CPU and out to the radiator in this instance.
 
Avet said:
This is simplified fluid dynamics. Fluid will go everywhere it cans. If you apply constant pressure (water pump) in a loop, flow trough narrow tube will be faster, and trough wider tube it will be slower so that flow in the unit of time will stay the same. What it means is that in 1 sec the same amount of fluid must pass trough narrow tube (at higher speed), and trough wide tube (at slower speed). It will split itself to every branch that is available. In this scenario part will separate for first GPU, part for second, and part for CPU. Flow (speed) trough each component of the loop will be lower because we now have wider tube than the inlet one. Imagine if you have one tube that splits into 3 tubes, and then they connect again into one tube. Every branch will have some % of the initial flow through inlet tube (depending on restriction, and diameter, etc...). At the end it will add up again to the same speed it went in. You will get fluid trough every component it will just go through them slower.

Fluid won't skip any part of the loop just because it needs to turn. It will go everywhere just at different speeds.

Edit: And the water has very high thermal capacity. Metal pot will heat up much faster than the water in it, because it can't store much thermal energy. Water can soak up a lot of heat, and change it's temperature very little. You will have little difference in temperature before, and after the radiator, because temperature differences are not big. And the water will try to equalize it's temperature throughout the loop. So the position of radiators and components doesn't matter that much. Speed of the fluid doesn't matter that much. Yes it makes a difference at fastest, and slowest setting but around the middle... Not so much. Few degrees at best. The only thing in a loop that matters is res before the pump, so that the pump doesn't run dry. This config will have slightly higher temperatures than full serial loop, but nothing drastic. And nothing that needs to be considered as bad.
Click to expand...

As I said...equal temp throughout the loop :thumb:
 
Ashihtaka said:
I understand that is the case over time, but what i'm wondering is how are the GPUs even receiving enough flow of the liquid to be a consistent part of the loop?
My initial thought is that most of the liquid is bypassing the GPU and going straight through to CPU and out to the radiator in this instance.
Click to expand...

The short version is, the inlet side has pressure from the pump, the outlet side has less pressure as it goes to the reservoir. Fluid wants to go from the high pressure side to the low pressure side.

You're thinking on kind of the right path in the fact that there's no (easy) way to control how much each block gets. That will be dependent on how much restriction there is through each waterblock and by the differential pressure across it. More restriction = less flow. It just happens that many blocks have a relatively equivalent amount of resistance. If that weren't the case and you had a waterblock that let fluid straight through a 3/8" path and one that only let water through a pinhole sized path, the one with the pinhole would get very little flow. There is also the compounding factor like the guys said that fluid has some momentum and wants to go straight, but it is almost negligible in a low pressure, low flow system like this.

Think of it like hooking your garden hose to a 3 way manifold. When you turn on the tap, water will come out of all 3 outlets relatively equally. Now partially block one of them off with your finger and the other two will get more flow. If you close each of them half way (representing the equal resistance of the waterblocks) water will still come out of all 3 relatively equally. The hose being the inlet side and the open atmosphere being the return side. That's what's happening in the parallel loop.
 
Loop CPU side.webp

I did a parallel loop between a gpu and CPU in a Haf XB. Temps were fine. These loops work so long as there isn't a large difference in flow between the two (or 3) blocks.
 
what I see from OP(edit) pic that the CPU is dead ended for flow takes the shortest route would be the bottom GPU leaving hardly any flow to CPU

the proper way would have 2 "T's" in center of the pci-e slots between cards, making them parallel , the 2 GPU's become one block
 

Attachments

  • negate flow.webp
    negate flow.webp
    26.6 KB · Views: 525
Last edited:
yes SS as long the tubing are equal runs as in your picture and single GPU

the OP shows that CPU is acting as plug negating flow

on GPU to have series flow would to stagger the ports
 

Attachments

  • negate.webp
    negate.webp
    7.6 KB · Views: 524
Last edited:
Think of the loop in terms of a dorm heated with HW rads..... Each rad has an adjustable flow valve to control temperatures so they must be set up in parallel otherwise the rad first in line would be able to severely restrict flow to the rest of the rads.

Just accept the premise that (like in electrical ccts) flow will continue throughout all legs of a parallel loop, it just won't necessarily flow equally through blocks with differing flow resistance.
 
You must log in or register to reply here.

Latest posts

xentr_legal_notice_title

xentr_legal_notice_description

Back
Top