Overclock.net banner
Status
Not open for further replies.

Serial VS Parallel 9.6LPM

30K views 243 replies 23 participants last post by  mega_option101 
#1 ·
a detailed parts list, of the two water loops used in this test setup.
2 MCR320-QP 360mm radiators
http://www.swiftech.com/MCRx20-QP.aspx
2 EVGA GeForce GTX 580 FTW Hydro Copper 2
https://www.evga.com/articles/00592/
1 XSPC raystorm CPU block
http://www.xs-pc.com/waterblocks-cpu/raystorm-cpu-water...
1 swiftech MPC655 pump
http://www.swiftech.com/mcp655.aspx
6 corsair SP fans
http://www.corsair.com/en-us/air-series-sp120-high-perf...
6 bits power 3 way 90's
http://www.frozencpu.com/products/12084/ex-tub-744/Bits...
1 Y block
http://bitspowerwork.com/html/product/pro_show.php?prod...
1 250mm reservoir
http://bitspowerwork.com/html/product/pro_show.php?prod...
8 3/8ID 1/2OD compression fittings
http://bitspowerwork.com/html/product/pro_show.php?prod...
2 T's
http://www.xs-pc.com/hose-fittings/g14-t-fitting-black-...
5 male to male fittings of varying sizes
http://www.xs-pc.com/hose-fittings/g14-5mm-male-to-male...
4 female to female fittings of varying sizes
http://www.xs-pc.com/hose-fittings/g14-18mm-female-to-f...
2 koolance flow meters with display
http://koolance.com/ins-fm19-coolant-flow-meter
http://koolance.com/dcb-fm01-flow-meter-adapter-with-di...
1 loop filter
http://koolance.com/ins-fltr03-inline-coolant-filter
2 temp sensors
http://www.frozencpu.com/products/10373/ex-tub-620/Bits...
1 alphacool bubble down
http://www.frozencpu.com/products/17333/ex-tub-1558/Alphacool_Bubble_Down_G14_Thread_Reservoir_Flow_Regulator_-_Deep_Black.html?tl=g30c97s169
1 bitspower aqua pipe
http://www.frozencpu.com/products/11863/ex-tub-713/Bitspower_G_14_Silver_Shining_Aqua-Pipe_I_BP-WTP-C17.html?tl=g30c97s169

This system is overclocked, on both the serial and parallel test setups.
CPU overclock @ 4Ghz
GPU1 overclock @ 850Mhz
GPU2 overclock @ 850Mhz
Memory overclock @ 1600Mhz
CPU stock @ 3.2Ghz
GPU1 stock @ 772Mhz
GPU2 stock @ 772Mhz
Memory stock @ 1333Mhz
this is the serial loop setup and what it looks like, and here are the hardware temps at idol, ambient temp of the room is at 28.6C
CPU 43C
GPU1 37C
GPU2 37C
the water temps at idol are
29C-30C
the hardware temps at load are
CPU 61C
GPU1 51C
GPU2 51C
the water temps at load are
37C-38C
the delta t of the system at load is 10C
water loop flow rate is at 2.4LPM
http://imgur.com/Gi4wFZO,G1Mte0g,Pb8ot4j,k8LlLK5


this is the parallel loop setup and what it looks like, and here are the hardware temps at idol, ambient temp of the room is at 28.6C
CPU 43C
GPU1 35C
GPU2 35C
the water temps at idol are
29C-30C
the hardware temps at load are
CPU 57C
GPU1 46C
GPU2 46C
the water temps at load are
34C-35C
the delta t of the system at load is 7C
water loop flow rate is 6.8LPM
http://imgur.com/pBZY5ne,yIT8rwA,CgNlfyH


this is the difference between the hardware used on the serial loop, and the hardware that needed to be added to the parallel loop.




serial loop
http://imgur.com/RxZGdWG,KAPjdJ8,TFC62xb,6pKBBiq,wzGMlwW,7TWMe5C,6r6rhB3

parallel loop
http://imgur.com/uElbcxQ,Ir22WaC,rMsjSDu,yQZeo5t,ijliSzM,HoXHIMM


ok, I built the full parallel loop with two pumps, and in my opinion it is a little better. the total flow rate of the system after install is 9.6LPM.



ambient air temp of the room is at 28.6
idle temps of the hardware are
CPU - 41C
GPU 1 - 33C
GPU 2 - 33C

the water temps at idle are
29-30C
the load temps of the hardware are
CPU - 54C
GPU 1 - 45C
GPU 2 - 45C

the water temps at load are
33-34C
the delta t of the system at load is 4C
water loop flow rate is at 9.6LPM
Link to Gallery

ok so this is my theory on how I lowered the delta t of my system with each new configuration of my water loop.
fist thing is that I did need to have enough radiator space to remove most of the heat being generated, by the components on the water loop.
than it is my opinion that the delta t of my water loop could be directly related to the flow rate through the radiators on my system.

this is how I came to that opinion.
on my serial loop setup with one pump, and a lower flow rate through the radiators of 1.2LPM it will have a delta t of 13C
on my serial loop test setup with one pump, the total flow rate through the radiators was 2.4LPM and I had a delta t of 10C
on my parallel loop test setup with one pump, the total flow rate was 6.8LPM so half that is 3.4LPM and that is the flow rate through one of the radiators on that loop and I had a delta t of 7C
on my parallel loop final build with two pumps, the total flow rate was 9.6LPM so half that is 4.8LPM and that is the flow rate through one of the radiators on that loop and I had a delta t of 4C
well that is my theory anyway, it seems to work.

so for some practical application of this theory it seems that for about every 1LPM increase of flow rate through the radiators, the delta t of the loop goes down by 3C.
I would like to test this theory further by seeing if a flow rate of 5.8LPM through one of the radiators will continue to lower the delta t of my loop by 3C if so, it would bring the total delta t of my water loop down to 1C :lol: I don't think that is going to happen, but I would still like to see what does happen to the delta t of my system at the 5.8LPM flow rate through one of the radiators.

cost of the serial loop test setup
$1003
cost of the parallel loop test setup
$1168
difference in cost of the parts to convert the loop from serial to parallel
$165
IMO it is worth the extra $165 to lower my load temps.
 
See less See more
9
#2 ·
That's some good work doing the testing for your self.

I have some trouble with some of the numbers though. For the delta t of the loop at load to halve would mean that the rads are dissipating twice the wattage per C. Careful testing of rads just doesn't see changes anything like that with flow rate.

Some of it could be put down to where in the loop water temp is measured. In a low flow system the water temp after the heat sources will comparatively high as the water is exposed to the heat source for longer. Basically in a high flow system there is a low difference between the coldest point and warmest point in the loop. In a low flow system there will be a higher difference between coldest and warmest point.

The serial loop flow rate was very low though. I assume caused by some of the components used like the filter and so on. Even so I just can't reconcile some of these numbers. It would be really good to have water temps from multiple points along the length of the different loops, eg, post heat source and post radiators.
In my own testing of the effects of flow rate the difference between a flow of 1.5LPM and 4.6LPM only caused a 3C core temp change and very slight coolant temp changes.
 
#3 ·
well I do have 2 inline water temp sensors one before the radiators and one after the reservoir.
I also did all of this with my own hardware, I have three different water cooled systems and all of them are paralleled to some existent, I did notice a larger difference in the load temps with my clocked system, than what the difference in temps where with my stock systems.
https://imageshack.com/i/exwiFW7Vj
https://imageshack.com/i/0vfi2nJ
https://imageshack.com/i/f1alldone005j
https://imageshack.com/i/04alldone003j
and don't get me wrong I did not jest learn all of this over night it took me 3 years of trial and error.
http://www.tomshardware.com/forum/274855-29-experimental-radiator-build
and I am still not done I have a lot more to learn I have a total of 4 GTX580 video cards and I would like to get all of the in a system at some point, working and staying cool.
the first time I tried to do a 5-way parallel setup with the CPU and 4 GPU's the system got real hot real fast all the way to 100C that was 2 years ago I took it apart and started working on how to better the system so that maybe some day I can try it again. today I am working on getting the flow rate of the system up to 15LPM to 17.4LPM that is my target range for the 5-way setup I do have a sketch of the system that I am playing with trying to get all bugs worked out of it still.
this is a sketch of what kind of system I am working towered trying to build, and why I needed my system to have such a high flow rate.

and I have been doing this for some time as well almost 20 years a lot has changed about water cooling in that time.
ok so I went back and did a really in depth look at all the systems that I have ever water cooled, and this is a break down of them, in order from oldest to newest.
the first H2OQPower system was to cool a super socket 7 CPU only at 550Mhz.
the second H2OQPower system was a to cool K7 CPU at 1.7Ghz to start and later a Ti4600 video card with block, I also added a 240mm radiator to the system.
then I up graded the first system to a K8 CPU at 2.6Ghz, a 6800GT video card, with two HHD's blocks, and I added memory blocks to the loop, I also added a 240mm radiator, then I upgraded the CPU block to a MPW50 CPU block, for this build.
then I up graded the second system to a intel CPU 4.3Ghz extreme core, and a ATi all in wonder video card with block, and I upgraded the CPU block to a MPW50 CPU block for this system.
then I built a third system to cool a AM2 CPU at 3Ghz, a 8800GT ultra video card with block, and I added the memory blocks to that system and it has 2 X 240mm radiators, I added a reservoir, a MPW50 CPU block, and a pump for the loop.
then I up graded the first system to a AM3 CPU at 4Ghz, and a 480GTX video card with block, I changed the 2 X 240mm radiators to a monstra 360mm radiator, the MPW50 CPU block for a Raystorm CPU block and added a reservoir and a pump to the system.
then I up graded the second system to a intel 2011 CPU at 3.2Ghz, and two 580GTX video cards with blocks, I changed the 2 X 240mm radiators out for two swiftech 360mm radiators, the MPW50 CPU block for a Raystorm CPU block and I added a reservoir and a pump to the system. I am still working on this system, I want to have 3 radiators and 4 video cards with blocks in total for this system.
I don't know if any of this helps, but let me know if there is something that I can help to clear up for you.
 
#4 ·
#5 ·
at some point I would like to clock the GPU's further, to see how the increased amount of heat into the loop will effect it.
 
#6 ·
#7 ·
@Jakusonfire
In my own testing of the effects of flow rate the difference between a flow of 1.5LPM and 4.6LPM only caused a 3C core temp change and very slight coolant temp changes.

here are the numbers, on how I calculate my potential flow rates through my parallel setup.
radiator = 5.9LPM 5.9 + 5.9 = 11.8LPM is the total flow rate possible on this parallel configuration
radiator = 5.9LPM

CPU = 4.6LPM 4.6 + 3.4 + 3.4 = 11.4LPM is the total flow rate possible on this parallel configuration
GPU = 3.4LPM
GPU = 3.4LPM

I achieved 10.5LPM on that parallel setup, so that is .9LPM less than the total flow rate possible.

here are the numbers on how I calculate my potential flow rate through my serial setup.

RAD = 5.9LPM in a serial configuration 3.4LPM is the total flow rate possible on a serial configuration.
CPU = 4.6LPM in a serial configuration
GPU = 3.4LPM in a serial configuration

on the serial loop I achieved 2.4LPM, and that is 1LPM less than the total flow rate possible.
 
#8 ·
I don't doubt anything you have done. Its just some of the conclusions.

For example when radiators are tested in a range from 1.9LPM to 5.7LPM the wattage dissipated changes from about 41W/C to 47W/C so for a 300W load the water temp would drop from 7.3C to 6.3C

http://www.overclock.net/t/1501978/ocn-community-water-cooling-test-thread/470#post_24011550


To get a 4.6 LPM flow rate is pretty easy with a single D5 and modern high flow blocks and rads. Thats through 3 blocks, 2 rads and a flow meter. Things like filters and that "Bubble Down flow regulator" can really crush flow rates.



I'm not following what you are meaning with total flow rates possible?
 
#9 ·
@jakusonfire
To get a 4.6 LPM flow rate is pretty easy with a single D5 and modern high flow blocks and rads. Thats through 3 blocks, 2 rads and a flow meter. Things like filters and that "Bubble Down flow regulator" can really crush flow rates

I think your understanding of the bubble down is backwards.
a stock reservoir has a 5.7LPM flow rate

http://imgur.com/DXioCNL,r6BGba1,GNJhifi


without the bubble down I could not get a higher flow rate through my reservoir

here is how I came to those numbers for my potential flow rates, and i used a single D5 pump.

in my serial loop, the most restrictive component is the GPU block at 3.4LPM.

http://imgur.com/MOYTv6t,yXrwjmn,no3mq1w



than the CPU at 4.6LPM

http://imgur.com/tSk0vWC,kb6d9J8,HvYFZ5P#0



than the radiator at 5.9LPM

http://imgur.com/Xw2ZusJ,hlrk6LC,S4doxpc



this is why I asked how you got your system to flow so fast. i don't see the hardware your talking about that can flow that high on any of your builds, also older hardware, does have a high flow open block design, it is also not designed to remove as much heat, as todays restrictive slit or pin design blocks are.
 
#10 ·
A lot of careful and elaborate testing goes into those charts.

Also in the same thread is pressure drop testing results for blocks and rads that illustrate why a d5 can push can push well over 4lpm through 3 blocks.

Now, the flow a pump can push through a single block is not the same as it can push through multiple blocks either in series or parallel.
If your pump can only push under 4lpm through a single block, it's clearly not an open high flow design or something is wrong with it. That would indicate well over 5 times the restriction of a modern EK gpu or Gpu block.
 
#11 ·
I prefer a real world testing, not a aquarium heater, that thing will not get to 150C if it is not under water, on the other hand a CPU will get 150C without a fan or block to cool it.
they also use pumps that most people would never put on there systems.
other than giving you a idea about the product and how it may preform, that kind of testing is not going to tell you what temp your CPU will run at, or what temp your GPU's will run at.
so I use a real world test, to see how my water loop works, with my hardware.
 
#12 ·
@jakusonfire
Now, the flow a pump can push through a single block is not the same as it can push through multiple blocks either in series or parallel.
If your pump can only push under 4lpm through a single block, it's clearly not an open high flow design or something is wrong with it. That would indicate well over 5 times the restriction of a modern EK gpu or Gpu block.

if you do not mind, I would like to see a pic of your flow meter and the flow rate on your loop, if you can do that for me, I would appreciate it?
I have never seen a EK CPU water block, or a EK full cover GPU block, that could run at a flow rate of 4.6LPM so this will be a good learning experience for me.

thanks in advance for your time.
 
#14 ·
so what your telling me is you do not have a flow meter, neither do you really know what your loop flow rate really is, well you do have your opinions, and that is fine your are entitled to them.
have a good day sir.
 
#15 ·
Quote:
Originally Posted by toolmaker03 View Post

so what your telling me is you do not have a flow meter, neither do you really know what your loop flow rate really is, well you do have your opinions, and that is fine your are entitled to them.
have a good day sir.
OK, I was willing to give you the benefit of the doubt but you are clearly way out of your depth. The nonsence being posted does not befit these forums.






A simple test. Two blocks, a d5 pump.
1.6 gpm or around 6lpm

The mps flow meter calibrated to same
 
#16 ·
well that is what I asked you for.
so I do have a few other questions for you.
1) how many pumps are on that loop?
2)what voltage is the pump running at?
I have a 24V step up for my pumps but they are not hooked to the pumps on my system, I am experimenting with the 24V step up, to get my loop to run at 15LPM to 17.4LPM, I still have a few bugs to get out of the system before it will run that fast.
@jakusonfire
OK, I was willing to give you the benefit of the doubt but you are clearly way out of your depth. The nonsence being posted does not befit these forums.

maybe I am out of my depth, as you put it, but it does not change the fact of how well my loop runs, or how well it works.

I don't understand why you did not simply post what I asked for to begin with, but I think you had some reason.

so your loop is running really fast have you done any calibration test on that flow meter to see how accurate it is.
mine is about a .5LPM slow from 1LPM to 4LPM at 5LPM to 10LPM it is really close to what it should be, at 10LPM the loop is running a little faster than what the flow meter reads.

I did take pics of all the stages if you need to see them.

http://imgur.com/qar2Cqa,48J6cqW,02n6g2r,H8OlBbr,bTTJR6o,OrXrIna,PPjHhO5,OwmM1ju

here are a few to get the idea of what I did.
 
#17 ·
also I have a aquaero 400 flow meter, with the Aqua Computer Aquaero 5 XT controller, it is really easy for me to change what is being displayed, by altering the programing, I can and have had it display that my loop is running at 1000LPM, even though my loop is clearly not running that fast, but I could have it display what ever I want it to.
 
#18 ·
As I said mate, there is a chart showing a 4.7lpm flow rate already posted. On top of that there are multiple sources of far higher quality testing that also show the sort of flow rates possible in modern blocks.

Also as already posted the mps flow meter is calibrated to the king instruments meter.

I suggest you read some of the testing work others have done, that you discount as not real world. Instead of the excuses. Because you seem confused about a lot of the things you are talking about.
 
#19 ·
well if I keep my house at 60F or 16C, I would have really good temps, but I keep my house a little hotter then that, at 24C to 30C, I live in a desert where the temps out side are at 110F to 120F or 43C to 49C most days out of year.

I am still trying to wrap my head around how you got your loop to flow at a rate of almost twice that of others with the same hardware you have, maybe it is the hard tubing that you have built your loop out of. I don't know, but good job either way.

I hope it all works well for you.
 
#20 ·
I don't know where you are getting this higher flow than other similar loops from.

It's fairly straightforward. The pressure drop of all the blocks and rads is available.

It adds up to a bit over 3 psi at 1GPM. A D5 pump makes around 4.8 PSI at 1GPM. So any loop with the same blocks will have a flow rate of over 1GPM
 
#21 ·
ok fine I will explain it to you.
fist read the instruction for that mechanical flow meter that you used for testing and calibration of your other flow meter.
it states clearly in the instructions that the size of tubing piped to it should be from 3/4in ID to 1in ID tubing, if you had done that the flow rate reported by that mechanical flow meter would be half of what it is now.
I work on million dollar machines with mechanical flow meters like the one that you have, I have to look at them every day, and If I had done what you did, by piping 3/8ID to 1/2ID tubing to that kind of flow meter, I would be fired, or at the very least written up for it.
does that clear every thing up for you, the fact that you calibrated you other flow meters to that skewed build not designed for the flow meter you are using, this is your issue, and why your reported flow rates are twice as fast as they should be..
I would suggest testing your flow meter with a real world test of how much water your pump really puts out, compared to what your flow meter is reading, and you will see what I mean.


rd1ozYT.jpg
 
#22 ·
You are just talking more and more nonsense now. The fittings on the flow meter simply won't allow tubing that large. It has 1/4" fittings built into it, so how exactly would 1 inch tubing help and where is this instruction to use tube that large ... Cause my manual doesn't say that.
Even if it could fit, it wouldn't change anything, certainly not by half. You are just flat making things up now and the fact you are criticizing an expensive King Instruments flow meter, (the key word there is instrument) while using a crappy koolance paddle flow meter is hilarious.

It's time to take this back to whatever noob forum it came from.
 
#23 ·
Quote:
Originally Posted by toolmaker03 View Post

Now, the flow a pump can push through a single block is not the same as it can push through multiple blocks either in series or parallel.
If your pump can only push under 4lpm through a single block, it's clearly not an open high flow design or something is wrong with it. That would indicate well over 5 times the restriction of a modern EK gpu or Gpu block.

if you do not mind, I would like to see a pic of your flow meter and the flow rate on your loop, if you can do that for me, I would appreciate it?
I have never seen a EK CPU water block, or a EK full cover GPU block, that could run at a flow rate of 4.6LPM so this will be a good learning experience for me.

thanks in advance for your time.
I have 2 290xs in parallel with EK water blocks and a bunch of rads, as well as flowmeters on EACH block on the outlet, and then a ball valve after each flowmeter before the res.

I'm able to adjust the flow through each gpu individually, closing or limiting one gpu forces the water through the other. My loop order is

reservoir > mcp655b > koolance pmp500 > rad > rad > rad > rad > rad> gpu 1, with parallel tube straigt down to gpu 2, then each gpu has it's own seperate out line. Attached in this line is a 90* elbow, followed by a flow meter, folloewd by 6" tube, followed by ball valve and res.

These are both acetal / nickel blocks. With just the mcp655 running at setting 4, my flow is as follow:

gpu1 1.2lpm
gpu2 1.2lpm
total flow: 2.4lpm, due to the parallel nature of the block setup


If i close the valve on EITHER of the GPU's, the flow becomes 2.4LPM, sometimes bouncing down to 2.1LPM on my koolance flowmeter display adapters.

Temp wise, under load, both cards are in the high 40s. This is 24/7 use daily and is silent with just the D5 running, a combined flow of 2.4LPM total or 1.2LPM individually through the blocks.

But when performance is needed, i turn on the PMP500 to 50% speed, which is 3400rpm - temps IMMEDIATELY drop 4-5* celsius, and the flow rate spikes to 2.1, and 1.8 (as adjusted to by the valves), equaling a total of 3.9liters per minute, and this is repeated again by closing either of the ball valves on one of the gpus it forces the other one to go to 3.9LPM flow rate. So the blocks are pretty high flowing i'd say..

If I put the PMP500 to 100% pump speed, the flow rate becomes 3.1 and 2.7LPM respectively, which is effectively 5.8 liters per minute total flow in parallel, .

I dialed down the flow in GPU 1 to reduce it to 2.9LPM, the flow rate on the other increases to 2.9LPM, allowing 5.4 liters per minute PEAK in parallel flow,

and by completely closing the gpu 2 valve all flow is forced through gpu 1, which becomes essentially a serial loop by definition, the flow rate through the single EK full cover water block becomes a pretty high 5.1 liters per minute, a loss of 0.3LPM in flow by forcing the water through just one of the blocks.

These numbers correlate with the readings I'm getting from the MPS400 at the bottom of the res about 6 inches before pump inlet, and the FM-16 U shaped koolance flow meter after the pump in the pedestal.
 
#24 ·
cool, so you get what I have done here, I have a 3.6LPM going through my CPU, all the time, and a 3.0LPM through each of my GPU's, for a total flow rate of 9.6LPM for my loop, I do have two flow meters on my loop, one right after the CPU, and the other is right after the reservoir/pump, it took me 3 years of playing with the loop configuration, to get it to this point, but I finally got all of the restrictive parts out of my loop, allowing for my loop to run at almost full capacity for the components attached to it.

looks like you are well on your way to getting higher flow rates through that loop of yours yourself, if you have questions please ask, I would be glade to help, or even just explain some of the issues that I ran into along the way.

I do have a lots of pics of stuff that I have done, and never posted, because no one has asked for that info yet, so let me know.
 
#25 ·

here is a look, at what the serial loop that ran at 1.2LPM looked like.
 
#26 ·
running cool this morning



even cooler

hhSNRjb.jpg
 
Status
Not open for further replies.
You have insufficient privileges to reply here.
Top