RRTech...The SOURCE for Extreme Liquid Cooling
Sidewinders Bitspower Jab-Tech

Go Back   RRTech...The SOURCE for Extreme Liquid Cooling > Liquid Cooling > Water Cooling

Reply
 
Thread Tools Display Modes
Old 05-06-2009, 05:06 PM   #1
Raccio14
WeakSauce
 
Raccio14's Avatar
 
Join Date: May 2009
Location: Florida (hate it here)
Posts: 22
Send a message via AIM to Raccio14
Default SLI/Crossfire multi-GPU flow setups: Paralel vs. Serial?

No matter how hard I think about it, I can't see how the paralel setup would work, let alone better.

In the serial setup, the water is forced to go through the full path of each card individually-


but in the parallel setup it seems like the water would just go straight through, and not do the path of all the cards, or maybe the areas that the arrows aren't going would be stagnant. I just don't see how it would work




Care to explain the pros/cons of a Parallel setup to GPU watercooling newbie?
Raccio14 is offline   Reply With Quote
Old 05-06-2009, 05:22 PM   #2
ballz0r
Wet DownUnder
 
ballz0r's Avatar
 
Join Date: Nov 2008
Location: United States of Australia
Posts: 4,459
Default

Water will flow where there is a difference in pressure.

Pressure drop in a water cooling loop comes from restriction of tubes, fittings and blocks.

When you connect your gpu blocks in parallel there is pretty much the same pressure drop across all your cards so flow across all your blocks is fairly equal.


Think about garden hose with holes in it... water will leak out all the holes, not just the one at the end


The main difference between series and parallel is flow, in series all blocks get the full flow rate, in parallel the flow rate thru each block is roughly 1/3 of the system flow rate.
__________________

To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.

mcoffey's Breadboard - Baleful's lappy farm - Coolmiester's Ghetto Skully - Baleful's Rig-in-a-box - JTrain's Desktop - Coolfris's Compact Cruncher - Just A Noobs Classy Cruncher - Shadowhunters NanoCruncher 4.0 - Oli4v's SuperSupreme with Cheeze - Floyd's Bottom Drawer - Oli4v's Sprawling Server

life getting you down? cant understand why people keep swearing at you? got sand in your mangina?
Sounds like you need to...
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.

Last edited by ballz0r; 05-06-2009 at 05:26 PM.
ballz0r is offline   Reply With Quote
Old 05-06-2009, 05:26 PM   #3
ILikeCosmosS
Proud Poonjabi
 
ILikeCosmosS's Avatar
 
Join Date: Apr 2009
Posts: 780
Default

dont you get cooler temperatures with parrellel loops?

slighty off topic question but : in a parrellel multi gpu loop could you have 1 end plug at gpu 1 and gpu 3 and 1 inlet/outlet at gpu1 and 3?
ILikeCosmosS is offline   Reply With Quote
Old 05-06-2009, 05:26 PM   #4
Snyxxx
Pushing Water & Oil
 
Snyxxx's Avatar
 
Join Date: Dec 2008
Location: Houston, Texas
Posts: 2,136
Default

I understand your confusion. It is really hard to get my head around the whole parallel flow concept.

But, ballz0r, hondacity and other leaders are proving that it is a viable option. There were many doubters, but testing has shown otherwise.


I think it is the most interesting topic in watercooling in years.
__________________

To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Snyxxx is offline   Reply With Quote
Old 05-06-2009, 05:30 PM   #5
ILikeCosmosS
Proud Poonjabi
 
ILikeCosmosS's Avatar
 
Join Date: Apr 2009
Posts: 780
Default

snyxxx have you ever thought a skulltrail system with parrellel loops?
1 to each cpu?
ILikeCosmosS is offline   Reply With Quote
Old 05-06-2009, 05:36 PM   #6
Kayin
Ancalagon the Black
 
Kayin's Avatar
 
Join Date: Nov 2008
Location: Tuscaloosa, AL
Posts: 1,251
Default

Parallel on ST has been done. It works just fine.
__________________

To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Kayin is offline   Reply With Quote
Old 05-06-2009, 05:47 PM   #7
slpdLoad
I'd like to buy a vowel..
 
slpdLoad's Avatar
 
Join Date: Mar 2009
Posts: 1,016
Default

Quote:
Originally Posted by ballz0r View Post
The main difference between series and parallel is flow, in series all blocks get the full flow rate, in parallel the flow rate thru each block is roughly 1/3 of the system flow rate.
So theoretically this would mean more time for the water to heat up, leading to an ever so slight increase in temps throughout the loop?
slpdLoad is offline   Reply With Quote
Old 05-06-2009, 05:56 PM   #8
Worthy
Uses a 40 dollar mouse ;)
 
Worthy's Avatar
 
Join Date: Mar 2009
Location: Detroit, MI
Posts: 2,312
Send a message via AIM to Worthy
Default

Quote:
Originally Posted by Snyxxx View Post
I understand your confusion. It is really hard to get my head around the whole parallel flow concept.

But, ballz0r, hondacity and other leaders are proving that it is a viable option. There were many doubters, but testing has shown otherwise.


I think it is the most interesting topic in watercooling in years.
All that and you have to admit, it sure looks cool too with either a nice EK setup, or a WaterCool setup. I so want to add a third card, but my dogs went to the vet and that ate 600 bucks today. I was Dogged!!!
__________________
One only starts taking flak when they are right over the target.
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Worthy is offline   Reply With Quote
Old 05-06-2009, 06:00 PM   #9
Worthy
Uses a 40 dollar mouse ;)
 
Worthy's Avatar
 
Join Date: Mar 2009
Location: Detroit, MI
Posts: 2,312
Send a message via AIM to Worthy
Default

Quote:
Originally Posted by slpdLoad View Post
So theoretically this would mean more time for the water to heat up, leading to an ever so slight increase in temps throughout the loop?
More contact time will pull away more heat, but I think there wouldn't really be a lot of difference. The emphasis with parallel is that the temps between each card are more equal since they are all being fed water of the same temp whereas in a series setup the second, or third card will always be hotter which may or may not affect overclocking a triple sli/xfire setup.

I guess since a lot of us are perfectionists, equal temps are a nice thing.

Just my thoughts.

And if you have a good rad, it doesn't matter what temp the water is when it goes through the loop. A good rad will address it.
__________________
One only starts taking flak when they are right over the target.
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.

Last edited by Worthy; 05-06-2009 at 06:02 PM.
Worthy is offline   Reply With Quote
Old 05-06-2009, 06:01 PM   #10
ballz0r
Wet DownUnder
 
ballz0r's Avatar
 
Join Date: Nov 2008
Location: United States of Australia
Posts: 4,459
Default

Quote:
Originally Posted by slpdLoad View Post
So theoretically this would mean more time for the water to heat up, leading to an ever so slight increase in temps throughout the loop?
decreasing the flow rate would result in less heat being carried away from the cards so would result in higher gpu temps but lower coolant temps as less heat is being transfered to the coolant.

When you look at plots of temps v flow rates the curve flattens out, so there comes a point where the effect of this is negligible.

__________________

To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.

mcoffey's Breadboard - Baleful's lappy farm - Coolmiester's Ghetto Skully - Baleful's Rig-in-a-box - JTrain's Desktop - Coolfris's Compact Cruncher - Just A Noobs Classy Cruncher - Shadowhunters NanoCruncher 4.0 - Oli4v's SuperSupreme with Cheeze - Floyd's Bottom Drawer - Oli4v's Sprawling Server

life getting you down? cant understand why people keep swearing at you? got sand in your mangina?
Sounds like you need to...
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
ballz0r is offline   Reply With Quote
Old 05-06-2009, 06:08 PM   #11
slpdLoad
I'd like to buy a vowel..
 
slpdLoad's Avatar
 
Join Date: Mar 2009
Posts: 1,016
Default

Quote:
Originally Posted by ballz0r View Post
decreasing the flow rate would result in less heat being carried away from the cards so would result in higher gpu temps but lower coolant temps as less heat is being transfered to the coolant.

When you look at plots of temps v flow rates the curve flattens out, so there comes a point where the effect of this is negligible.
Thanks, very good explanation. I guess I had always just heard that temps equalize throughout the loop but hadn't thought of that distinction being related to flow rates.
slpdLoad is offline   Reply With Quote
Old 05-06-2009, 06:39 PM   #12
Snyxxx
Pushing Water & Oil
 
Snyxxx's Avatar
 
Join Date: Dec 2008
Location: Houston, Texas
Posts: 2,136
Default

Quote:
Originally Posted by Worthy View Post
All that and you have to admit, it sure looks cool too with either a nice EK setup, or a WaterCool setup. I so want to add a third card, but my dogs went to the vet and that ate 600 bucks today. I was Dogged!!!
+1 for looking cool while being functional. I love looking at ballz0r's avatar with the tri-sli and parallel flow.
__________________

To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Snyxxx is offline   Reply With Quote
Old 05-06-2009, 06:56 PM   #13
Raccio14
WeakSauce
 
Raccio14's Avatar
 
Join Date: May 2009
Location: Florida (hate it here)
Posts: 22
Send a message via AIM to Raccio14
Default

cool, I never really thought about it that way. Thanks guys.

In reality though, performance and flow effects look negligible though, so it seems like the choice comes down to more of a loop-setup/visuals and really depends on the case/cooling loop.

unfortunately, my cooling loop isn't set up in a configuration that would work for the parallel flow, so I guess I don't get those awesome looks
Raccio14 is offline   Reply With Quote
Old 05-06-2009, 07:13 PM   #14
ballz0r
Wet DownUnder
 
ballz0r's Avatar
 
Join Date: Nov 2008
Location: United States of Australia
Posts: 4,459
Default

another factor to consider is the difference in head loss between the two setups


basically, higher the flow rate, means more head loss



We can see that for the Komodo at 2 gpm you have a pressure drop of ~1.25, so in series this becomes 3.75 mh2o

In parallel, 1/3 of 2gpm is 0.66, at this flow rate we only have ~0.25 Mh2o pressure drop

so parallel should result in higher system flow, which results in better radiator performance and better cooling for other blocks in the loop. most likely your gpu temps will be slightly higher than a series setup, but I dont think +5oC would make any difference to graphics card overclocks
__________________

To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.

mcoffey's Breadboard - Baleful's lappy farm - Coolmiester's Ghetto Skully - Baleful's Rig-in-a-box - JTrain's Desktop - Coolfris's Compact Cruncher - Just A Noobs Classy Cruncher - Shadowhunters NanoCruncher 4.0 - Oli4v's SuperSupreme with Cheeze - Floyd's Bottom Drawer - Oli4v's Sprawling Server

life getting you down? cant understand why people keep swearing at you? got sand in your mangina?
Sounds like you need to...
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.

Last edited by ballz0r; 05-06-2009 at 07:15 PM.
ballz0r is offline   Reply With Quote
Old 05-06-2009, 07:29 PM   #15
Worthy
Uses a 40 dollar mouse ;)
 
Worthy's Avatar
 
Join Date: Mar 2009
Location: Detroit, MI
Posts: 2,312
Send a message via AIM to Worthy
Default

Quote:
Originally Posted by Snyxxx View Post
+1 for looking cool while being functional. I love looking at ballz0r's avatar with the tri-sli and parallel flow.

Funny you mention that.
I was just browsing one day looking for watercooling info and I stumbled across his machine in this forum through a random google search. It's what made me join here and sparked my curiosity of what else was here.

I was a member of XS at the time, but there was no one there I really cared to talk to. I mostly read there.

That's why I think we need to show mmore of what we do and contribute more reviews and stuff. It'll only attract more people.
__________________
One only starts taking flak when they are right over the target.
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Worthy is offline   Reply With Quote
Old 05-06-2009, 08:04 PM   #16
Hondacity
Banned by Ira-k
 
Hondacity's Avatar
 
Join Date: Jan 2009
Posts: 1,553
Default

great job you guys...

but anyways...here's my 2cents bout parallel loops: it works

stock speed : 675MHz
OC speed : 812MHz

http://www.hwbot.org/result.do?resultId=847066
Hondacity is offline   Reply With Quote
Old 05-06-2009, 08:37 PM   #17
LowBrowser
Learning Liquid
 
LowBrowser's Avatar
 
Join Date: Feb 2009
Location: New Zealand
Posts: 471
Default

It's quite easy to think about if you liken the water flowing to electricity (analogy puts the electrical terms in brackets). Each waterblock has a certain pressure drop (resistance). There's a certain head pressure (voltage) that the pump can provide. Now water is going to flow through every waterblock, but at a proportion related to the pressure drop. You have to ask, why would the water take the path of highest resistance [it doesn't]? The shortest and least resistant to flow path is through the first waterblock and through none of the other ones. But that doesn't happen. Most of the flow does go through the first waterblock though [an insignificant difference however], then slightly less through the second block and again slightly less through the third. The flow sort of ends up being the current if we continue the analogy, except that pumps are more complex, but it's still fairly valid. There are ways to make the flow be absolutely even between the waterblocks, but as far as temperatures, it won't make a difference.
LowBrowser is offline   Reply With Quote
Old 05-07-2009, 03:45 AM   #18
> Evil E. <
WeakSauce
 
> Evil E. <'s Avatar
 
Join Date: Apr 2009
Location: Belgium
Posts: 26
Default

Interesting theory, LowBrowser. This is a fairly complex concept... I do have another question regarding this, maybe you can answer that one for me as well :
In the parallel Tri-VGA setup example from the first post (2nd picture), is the coolant being pushed out of the 3rd (lowest) block partially preventing the coolant coming out of the 2nd block (middle) and the coolant coming out of the 3rd + 2nd block partially preventing the coolant coming out of the 1st (top) block, and thus (proportionally ?) decreasing the flowrate in the upper blocks ?
In other words : does this type of (parallel) cooling setup add an extra hindrance for the coolant in the upper blocks because of the outcoming coolant from the (in the example from the 1st post) lower blocks partially pushing back the coolant in the upper blocks ?

And here's another one, also in the line of ILikeCosmosS' question regarding inlet/outlet positioning query : would there be a difference in coolant flowrate and/or effective heat transfer (temperature) if you put the outlet of the parallel setup for triple VGA's at the opposite end (in this case the lowest block), compared to a setup where inlet/outlet are placed on the same block ?

Is someone in here perhaps capable of testing this ?
> Evil E. < is offline   Reply With Quote
Old 05-07-2009, 04:19 AM   #19
LowBrowser
Learning Liquid
 
LowBrowser's Avatar
 
Join Date: Feb 2009
Location: New Zealand
Posts: 471
Default

Quote:
Originally Posted by > Evil E. < View Post
Interesting theory, LowBrowser. This is a fairly complex concept... I do have another question regarding this, maybe you can answer that one for me as well :
In the parallel Tri-VGA setup example from the first post (2nd picture), is the coolant being pushed out of the 3rd (lowest) block partially preventing the coolant coming out of the 2nd block (middle) and the coolant coming out of the 3rd + 2nd block partially preventing the coolant coming out of the 1st (top) block, and thus (proportionally ?) decreasing the flowrate in the upper blocks ?
In other words : does this type of (parallel) cooling setup add an extra hindrance for the coolant in the upper blocks because of the outcoming coolant from the (in the example from the 1st post) lower blocks partially pushing back the coolant in the upper blocks ?

And here's another one, also in the line of ILikeCosmosS' question regarding inlet/outlet positioning query : would there be a difference in coolant flowrate and/or effective heat transfer (temperature) if you put the outlet of the parallel setup for triple VGA's at the opposite end (in this case the lowest block), compared to a setup where inlet/outlet are placed on the same block ?

Is someone in here perhaps capable of testing this ?
There shouldn't be a difference if you put the inlet/outlet at the other end as far as I can tell (intuitively, I haven't checked it at all) but if anything, it should work better.

Anyway, maybe I didn't explain it that well. If you have a pump pushing water through a hose (sorry to be a jerk and simplifying, but it should become obvious), and you split it into three smaller pipes for a short length, then rejoin into another larger hose where the water is flowing to the right. Would the water that rejoins after the short piece of pipe flow back the wrong way, or would it continue in the same direction? Effectively a pipe with the same pressure drop could be substituted for a GPU block to show that it works.

Last edited by LowBrowser; 05-07-2009 at 04:38 AM.
LowBrowser is offline   Reply With Quote
Old 05-07-2009, 08:25 AM   #20
Afterburner
411 Cruncher
 
Afterburner's Avatar
 
Join Date: Mar 2009
Location: Puget Sound
Posts: 155
Default

Very interesting read..... I had DD Blocks and this was not an idea I found to be possible with those, do to my own lack of knowledge it seems....

Now I have a question though. Can we do this with the HK's when we finally get them?


__________________

To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Afterburner is offline   Reply With Quote
Old 05-07-2009, 09:04 AM   #21
Hondacity
Banned by Ira-k
 
Hondacity's Avatar
 
Join Date: Jan 2009
Posts: 1,553
Default

Afterburner

here's my setup

Hondacity is offline   Reply With Quote
Old 05-07-2009, 09:24 AM   #22
LowBrowser
Learning Liquid
 
LowBrowser's Avatar
 
Join Date: Feb 2009
Location: New Zealand
Posts: 471
Default

One thing, if your card fries from bad information from me, well sorry (but it won't happen as long as you think about it for a second). As long as you have your outlet and inlet on different sides (ie inlet runs parallel, but isn't directly in line with the outlet).
LowBrowser is offline   Reply With Quote
Old 05-07-2009, 09:26 AM   #23
Worthy
Uses a 40 dollar mouse ;)
 
Worthy's Avatar
 
Join Date: Mar 2009
Location: Detroit, MI
Posts: 2,312
Send a message via AIM to Worthy
Default

Quote:
Originally Posted by LowBrowser View Post
There shouldn't be a difference if you put the inlet/outlet at the other end as far as I can tell (intuitively, I haven't checked it at all) but if anything, it should work better.

Anyway, maybe I didn't explain it that well. If you have a pump pushing water through a hose (sorry to be a jerk and simplifying, but it should become obvious), and you split it into three smaller pipes for a short length, then rejoin into another larger hose where the water is flowing to the right. Would the water that rejoins after the short piece of pipe flow back the wrong way, or would it continue in the same direction? Effectively a pipe with the same pressure drop could be substituted for a GPU block to show that it works.
With the inlet/outlet on opposite ends you have the possibility of accumulating an air bubble in the top port that is plugged to the point where it could actually kill flow in the top gpu and cause the temp of that gpu to be hotter than the others. This was largely my problem and why I put the inlet and outlet at the top.

Don't ask how I got a bubble in there. it just happened and it happened a few times. I guess with the typical rads you never really get ALL of the air out of them and there is the possibility as this air tried to purge that it will sit in any spot where it can rise through the water. A plugged top port on a gpu block just as much as a fillport.
__________________
One only starts taking flak when they are right over the target.
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Worthy is offline   Reply With Quote
Old 05-07-2009, 10:20 AM   #24
LowBrowser
Learning Liquid
 
LowBrowser's Avatar
 
Join Date: Feb 2009
Location: New Zealand
Posts: 471
Default

Ouch, that's shitty luck. Well that's why you need a little practical experience to iron out some kinks. I'd just hate to have to be the one to pay the bill.
LowBrowser is offline   Reply With Quote
Old 05-07-2009, 11:36 AM   #25
Worthy
Uses a 40 dollar mouse ;)
 
Worthy's Avatar
 
Join Date: Mar 2009
Location: Detroit, MI
Posts: 2,312
Send a message via AIM to Worthy
Default

Quote:
Originally Posted by LowBrowser View Post
Ouch, that's shitty luck. Well that's why you need a little practical experience to iron out some kinks. I'd just hate to have to be the one to pay the bill.
No detriment to the hardware. it just got uneven temps, and one time the top card reached around 67c idle which pointed it out to me.

The whole emphasis is that anywhere an air pocket can form, it should be eliminated. I do not believe this would happen at all or near as much with a series cooling setup, but I will have to experiment there to make a solid conclusion.

I have used parallel mode from the getgo.
__________________
One only starts taking flak when they are right over the target.
To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
Worthy is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 11:36 PM.


Powered by vBulletin® Version 3.7.6
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Ad Management by RedTyger