Parallel Loop Theory

[emosun]

1 minute ago, W-L said:

The factor that would make it see a difference in flow is the extra distance away from the inlet and outlet of the lowest block but since the losses with friction but is very limited. The pressure applied by the pump will easily force fluid through all three blocks, not 100% evenly but very close. You will not have any kind of dead zone. 

yes , thank you. all of which could be avoided or made better if card 3 was the outlet or if all 3 had their own outlets

[SquintyG33Rs]

1 minute ago, Napalm360 said:

@SquintyG33Rs so apply the same principle to a CPU and ram block, where there's a t fitting on each port of the ram block to allow the two tubes going to and from the CPU, would that work?

you can simply assume that 2 blocks that do not have the exact same shape will have different resistance... you could measure this? with a flow meter and running the pump a different speeds plotting the flow speed out compared to pump speed. then math out how much flow each will get with how much they need to be cooled and you can make it happen. power to you but honestly way too much work..... so on that note goodnight

[Napalm360]

1 minute ago, emosun said:

hmm ok

 

what happens when you split one line into 3 lines and then combine them again? where does the water or electricity go? the path of least resistance right?

if all line are the same length , then all get equal flow of electricity or water.

is they are different , lengths like card 2 and card 3 being FARTHER away from the inlet and outlet , then they will receive less current.

You realise you contradicted yourself by saying if all lines are same length then equal flow then saying that the cards are different lengths, it still works, if you still can't wrap your head around it I'll make some fancy paint diagrams in the morning

[SquintyG33Rs]

2 minutes ago, emosun said:

yes , thank you. all of which could be avoided or made better if card 3 was the outlet or if all 3 had their own outlets

unless they each have their own loop individual outlets = joined outlets because they come together inevitably in a single loop.

[emosun]

1 minute ago, Napalm360 said:

You realise you contradicted yourself by saying if all lines are same length then equal flow then saying that the cards are different lengths

the cards 2 and 3 are different lengths away are you nuts?

[SquintyG33Rs]

1 minute ago, Napalm360 said:

You realise you contradicted yourself by saying if all lines are same length then equal flow then saying that the cards are different lengths, it still works, if you still can't wrap your head around it I'll make some fancy paint diagrams in the morning

yes more paint plox.. plz make it more RGB this tiem + 3D, 4k and VR ready

[Napalm360]

5 minutes ago, SquintyG33Rs said:

yeah. but psychology also taught us that in a position of equals it is near impossible to convince somebody of their mistake because of confirmation bias and every attempt to prove them wrong will only reinforce their convictions because we seem desperate to bring them to our side instead of just agreeing with them... humans are so dumb sometimes i swear.

Not to mention the thread hijacking 

[Napalm360]

3 minutes ago, SquintyG33Rs said:

yes more paint plox.. plz make it more RGB this tiem + 3D, 4k and VR ready

Why do I even bother continuing this also you need some tempered glass aluminum buttons and drawpad with optane to help you draw faster

[W-L]

6 minutes ago, emosun said:

yes , thank you. all of which could be avoided or made better if card 3 was the outlet or if all 3 had their own outlets

Yes having it inelts at GPU 1 and outlet at GPU 3 would balance it out but the difference with that and both being up top really is not an amount that is worth worrying about. Whatever aesthetically looks best will work. 

 

4 minutes ago, Napalm360 said:

@SquintyG33Rs so apply the same principle to a CPU and ram block, where there's a t fitting on each port of the ram block to allow the two tubes going to and from the CPU, would that work?

You can parallel any block together but basically as mentioned with different blocks comes different restrictions and flow. The simplest is still to run it in series you see no appreciable benefit from running it in parallel assuming your pump is not the limiting factor. 

[Napalm360]

1 minute ago, W-L said:

Yes having it inelts at GPU 1 and outlet at GPU 3 would balance it out but the difference with that and both being up top really is not an amount that is worth worrying about. Whatever aesthetically looks best will work. 

 

You can parallel any block together but basically as mentioned with different blocks comes different restrictions and flow. The simplest is still to run it in series you see no appreciable benefit from running it in parallel assuming your pump is not the limiting factor. 

Tbh only reason I considered running CPU and ram was it looked tidier, suppose easiest method of testing would be CPU temps before and after ram block being added in parallel (can a mod just delete this topic already, waste of time and space)

[emosun]

6 minutes ago, Napalm360 said:

You realise you contradicted yourself by saying if all lines are same length then equal flow then saying that the cards are different lengths

different lengths = different resistances

 

meaning gpu 3 will get less than 2 which gets less than 1

[SquintyG33Rs]

2 minutes ago, emosun said:

different lengths = different resistances

 

meaning gpu 3 will get less than 2 which gets less than 1

makes less than 0.01% difference. because the resistance from the tube is exactly what you mark as negligible in your physics equations. you know when you do cinematics in highschool air resistance isn't accounted for in moving wagons and shit. well this is that

[W-L]

3 minutes ago, Napalm360 said:

Tbh only reason I considered running CPU and ram was it looked tidier, suppose easiest method of testing would be CPU temps before and after ram block being added in parallel (can a mod just delete this topic already, waste of time and space)

Yeah unless you want to break out a bunch of flow gauges and temp monitors to check everything the easiest would be to do a quick and dirty mock up and test in series and parallel to see what would be most heavily affected. My prediction would be the CPU temps probably be a little higher in parallel with a RAM block as most have wide channels and aren't very restrictive. 

 

 

Also this is a good discussion point, with great information. 

[Napalm360]

1 minute ago, emosun said:

different lengths = different resistances

 

meaning gpu 3 will get less than 2 which gets less than 1

Think of it this way, the 'restriction' in flow between the cards is like adding an item that weights 10kg one that weights 20kg and 30kg  to a rocket but the rocket has plenty of thrust to lift up the 3 items, basically yes there's a resistance, it's not much and the pump has plenty of power to push the fluid through meaning in real world it isn't noticeable at all and does NOT affect performance

[Napalm360]

2 minutes ago, W-L said:

Yeah unless you want to break out a bunch of flow gauges and temp monitors to check everything the easiest would be to do a quick and dirty mock up and test in series and parallel to see what would be most heavily affected. My prediction would be the CPU temps probably be a little higher in parallel with a RAM block as most have wide channels and aren't very restrictive. 

 

 

Also this is a good discussion point, with great information. 

Well good to hear this is informational, what about parallel 280mm radiators for example, would it be fine then?

[emosun]

5 minutes ago, Napalm360 said:

Think of it this way, the 'restriction' in flow between the cards is like adding an item that weights 10kg one that weights 20kg and 30kg  to a rocket but the rocket has plenty of thrust to lift up the 3 items, basically yes there's a resistance, it's not much and the pump has plenty of power to push the fluid through meaning in real world it isn't noticeable at all and does NOT affect performance

you would then have to run a pump that has 3 times the throughput of a normal pump to move the water through this section at the same speed it would run if there was a single card

meaning the water has to move at 3x the speed through the cpu and radiator as well in order for the 3 gpu blocks to perform the same as they would if there was a single gpu.... what a joke

[W-L]

3 minutes ago, Napalm360 said:

Well good to hear this is informational, what about parallel 280mm radiators for example, would it be fine then?

Paralleling rads don't have any gains at all but yes it can be done. Rads are already considered to be areas of low flow in a loop so while you can help with flowrates with a good pump like a D5 or DDC you have little worry of that ever really being an issue. 

 

Performance wise you should in theory see no difference. 

http://www.xtremesystems.org/forums/showthread.php?145954-Radiators-in-Parallel

[emosun]

what's even better is you guys got all smug before you even realized the gpus were in fact different resistances. and you both responded with , "uh it wont make a difference" even though you have no clue if it really would because you've never tried it lol.

that's why radiators have their inlet and outlets end to end or kiddy corner.
 

talk about having your foot in your mouth , or perhaps you said it better with 

 

29 minutes ago, SquintyG33Rs said:

humans are so dumb sometimes i swear.

 

[SquintyG33Rs]

32 minutes ago, emosun said:

what's even better is you guys got all smug before you even realized the gpus were in fact different resistances. and you both responded with , "uh it wont make a difference" even though you have no clue if it really would because you've never tried it lol.

that's why radiators have their inlet and outlets end to end or kiddy corner.
 

talk about having your foot in your mouth , or perhaps you said it better with 

 

 

you don't know what you're talking about. there isn't a single radiator design. alpha-cool radiators have their ports at either end of the thing. some radiator have ports on both sides that you can use in multiple configs. even if there was the only reason it would be that way would be because it's easier to manufacture not because of you're completely flawed thought process. so your quote of me is just you pointing at your own reflection while being too dumb to know what you're seeing.

[SquintyG33Rs]

58 minutes ago, emosun said:

you would then have to run a pump that has 3 times the throughput of a normal pump to move the water through this section at the same speed it would run if there was a single card

meaning the water has to move at 3x the speed through the cpu and radiator as well in order for the 3 gpu blocks to perform the same as they would if there was a single gpu.... what a joke

yes but in reality flow speed isn't very important to cooling performance as long as it's enough. you don't see much temp drop from running pump at 100% vs running it at 25 in a straight series loop. the heat travels through the liquid itself through convection and induction and that matters more than the flow. the movement promotes the convection heat transfer and makes it something like 10x more fast/efficient, and it only takes a little of help it doesn't need to be going at the speed of light (theoretically at some point the speed causes more heat buildup from friction than it would ever help any cooling). you just don't want 90% of the flow on one thing and the rest getting 10% because then it's essentially not flowing since pumps aren't pushing very fast to begin with. the thermal mass of the total water volume has a bigger impact and then obviously the transfer surface water->air to keep the water temperature under threshold to stay within most optimal thermal transfer rates away from the heat generators (since this depends on the temperature delta)

 

what i'm trying to say is the division by 3 of the flow in the GPU doesn't mean you're cooling 3x worse. you're only losing a couple degrees. about as much as you gain/lose from changing the order of components in the loop.

[SquintyG33Rs]

1 hour ago, Napalm360 said:

Well good to hear this is informational, what about parallel 280mm radiators for example, would it be fine then?

the only annoying thing here is you'll need T splitter fittings. but same rads mean same resistance to flow therefore 100% feasible. you won't lose performance. if you prefer the look or it's easier for you to make the runs then go ahead.

[For Science!]

Here is a graphic illustration of parallel flow, and why it works in GPU parallel setups. The outputs then meet up on the other end and keep going in the case of a PC. The front hole receives more or less the same amount of water compared to the hole in the end.

 

 

I think its intuitive to grab that if the other end of the hose was open (i.e. a path of lower resistance) then most of the water would flow out the other end. Worst case scenario, water may not come out of the smaller holes. This is why you should only parallize things close in resistance.

[The Benjamins]

10 hours ago, emosun said:

you would then have to run a pump that has 3 times the throughput of a normal pump to move the water through this section at the same speed it would run if there was a single card

meaning the water has to move at 3x the speed through the cpu and radiator as well in order for the 3 gpu blocks to perform the same as they would if there was a single gpu.... what a joke

You do know Jayztwocents use to run 3 way SLI in parallel for years and at most he got a 1c difference between blocks.

[Napalm360]

20 minutes ago, The Benjamins said:

You do know Jayztwocents use to run 3 way SLI in parallel for years and at most he got a 1c difference between blocks.

I honestly think emosum is trolling

[Napalm360]

9 hours ago, SquintyG33Rs said:

the only annoying thing here is you'll need T splitter fittings. but same rads mean same resistance to flow therefore 100% feasible. you won't lose performance. if you prefer the look or it's easier for you to make the runs then go ahead.

Indeed. Not that Id be doing a parallel Loop for a LONG time (due to cost) but knowing the feasibility of certain things/

 

Kinda Off-topic but thoughts on this config?

Spoiler

Orange tubing is In front Of white Tubing (helps visualise depth?)

 

Edit: Except the run from cpu > bottom rad > gpu (the cpu > bottom rad is behind the tube that goes from bottom rad to gpu