Advice needed - using 2 pumps and 2 radiators
- Thread starter heman
- Start date
Soldato
- Joined
- 12 Sep 2012
- Posts
- 11,704
- Location
- Surrey
If you have 2 D5's use them in series.
Whether to go parallel or series with pumps was an issue when the pumps were much weaker and the flow rates dropped bellow a level where they could keep temperatures uniform. Now a single D5 has way way more flow rate than you need to keep temperature uniform even with several blocks.
Testing with a flow meter will not show you much in terms of performance. The best thing to do is to compare temperature.
Parallel pumps Give an increase flow rate but no added head. This was usefull when most pumps available were weak aquarium pumps, but now you can achieve a uniform temperature in your loop even if your D5 flow rate were cut down to a third of its original due to high restrictive block.
Pumps in series do not add more flow rate but instead add head. Head is often described as pressure because it behaves in the same way as in it is the pushing force of the coolant. This sounds very similarly connected to the flow rate when put this way, which often causes the confusion, and not without reason, since the two are closely connected when talking about a waterloop. Head becomes important in a loop with either high restriction blocks and radiators or a large loop (your case) where there are a number of blocks and a large radiator and a large amount of water needs to be moved.
Adding a block will reduce the flow rate of a loop, regardless of whether pumps are parallel or not. With this in mind, often people think more flow is better. However with higher head, the reduction in flow is lower compared to systems with lower head, per block added. For you who has a number of blocks and radiators and will need to push a lot of water, you will find a series pump set up benefit you more.
Essentially in series your reduction in flow would be cut down, where as in parallel your much higher flow will be reduced heavily with every added component and in the end you may end up with a much less uniform temperatures. In the end you will find that the D5s in series will more likely out perform the parallel in terms of temperatures.
Note: that a similar example of head vs flow would be to compare static pressure of fans and the CFM. Where the radiators (high restrictive is high fpi and thicker ones) would be comparable to low/high restrictive blocks.
Source: my years of experience of monotonous lectures involving hydrodynamics when i studied for my Mphys.
Whether to go parallel or series with pumps was an issue when the pumps were much weaker and the flow rates dropped bellow a level where they could keep temperatures uniform. Now a single D5 has way way more flow rate than you need to keep temperature uniform even with several blocks.
Testing with a flow meter will not show you much in terms of performance. The best thing to do is to compare temperature.
Parallel pumps Give an increase flow rate but no added head. This was usefull when most pumps available were weak aquarium pumps, but now you can achieve a uniform temperature in your loop even if your D5 flow rate were cut down to a third of its original due to high restrictive block.
Pumps in series do not add more flow rate but instead add head. Head is often described as pressure because it behaves in the same way as in it is the pushing force of the coolant. This sounds very similarly connected to the flow rate when put this way, which often causes the confusion, and not without reason, since the two are closely connected when talking about a waterloop. Head becomes important in a loop with either high restriction blocks and radiators or a large loop (your case) where there are a number of blocks and a large radiator and a large amount of water needs to be moved.
Adding a block will reduce the flow rate of a loop, regardless of whether pumps are parallel or not. With this in mind, often people think more flow is better. However with higher head, the reduction in flow is lower compared to systems with lower head, per block added. For you who has a number of blocks and radiators and will need to push a lot of water, you will find a series pump set up benefit you more.
Essentially in series your reduction in flow would be cut down, where as in parallel your much higher flow will be reduced heavily with every added component and in the end you may end up with a much less uniform temperatures. In the end you will find that the D5s in series will more likely out perform the parallel in terms of temperatures.
Note: that a similar example of head vs flow would be to compare static pressure of fans and the CFM. Where the radiators (high restrictive is high fpi and thicker ones) would be comparable to low/high restrictive blocks.
Source: my years of experience of monotonous lectures involving hydrodynamics when i studied for my Mphys.
high flow in block is preferential, and low flow in rad is preferential
as more flow thru the block will remove more heat
low flow in rad will disperse more heat
so a balance between those depending on your rad and block performance is required
best thing is as large a volume of coolant as possible so a high volume system will be better than a low volume system
especially if your rads are exposed to a really low ambient temp, and as much of the coolant volume is external from your main case.
keep your room or the area where your rads and most of your coolant is as cold as possible.
I noticed a reasonable increase in flow with 2 pumps in series and as a result a slightly lower temp in case and over cpu etc
as more flow thru the block will remove more heat
low flow in rad will disperse more heat
so a balance between those depending on your rad and block performance is required
best thing is as large a volume of coolant as possible so a high volume system will be better than a low volume system
especially if your rads are exposed to a really low ambient temp, and as much of the coolant volume is external from your main case.
keep your room or the area where your rads and most of your coolant is as cold as possible.
I noticed a reasonable increase in flow with 2 pumps in series and as a result a slightly lower temp in case and over cpu etc
Last edited:
In the spirit of complicating matters, what do people think of parallel vs series when the loop is big?
Something along the lines of reservoir -> two D5s -> eight cpu blocks -> big radiator -> reservoir.
Putting the CPU blocks in parallel seems an inherently sensible thing to do given the pressure drop across a block, though some combination of series and parallel may be better. Under those circumstances, I think it is less obvious whether the D5's should be in series or parallel.
There is another argument against parallel D5's though. If one fails, the other has a new low-resistance path available, through the failed pump. So two pumps in parallel may be more akin to double the chance of failure than to redundancy. I think it follows that parallel pumps would benefit from valves.
Something along the lines of reservoir -> two D5s -> eight cpu blocks -> big radiator -> reservoir.
Putting the CPU blocks in parallel seems an inherently sensible thing to do given the pressure drop across a block, though some combination of series and parallel may be better. Under those circumstances, I think it is less obvious whether the D5's should be in series or parallel.
There is another argument against parallel D5's though. If one fails, the other has a new low-resistance path available, through the failed pump. So two pumps in parallel may be more akin to double the chance of failure than to redundancy. I think it follows that parallel pumps would benefit from valves.
Soldato
- Joined
- 12 Sep 2012
- Posts
- 11,704
- Location
- Surrey
Under those circumstances where a large amount of water needed to be moved and the water passes through a number of high flow blocks (which could be simplified by treating it as a very restrictive large block or stealth radiator) much more head is needed, so linking pumps in series would benefit more.
I'm fairly sure you need less head pressure to pump a given mass flow rate through two blocks in parallel than to pump through a single block, probably as 1/Rt = 1/R1 + 1/R2 since electrical/water analogies mostly work.
If your system initially pumps m kg/s through a single cpu block and you add a second in series, you'll get less than m flow rate. If you add a second cpu block in parallel instead, you'll get more than m flow rate in total, but still less than m through each individual block.
The answer to optimising n pumps through m blocks inevitably lies in performance curves and maths, but without running the numbers I don't think the everyhing-in-series rule of thumb can be trusted.
If your system initially pumps m kg/s through a single cpu block and you add a second in series, you'll get less than m flow rate. If you add a second cpu block in parallel instead, you'll get more than m flow rate in total, but still less than m through each individual block.
The answer to optimising n pumps through m blocks inevitably lies in performance curves and maths, but without running the numbers I don't think the everyhing-in-series rule of thumb can be trusted.
Soldato
- Joined
- 12 Sep 2012
- Posts
- 11,704
- Location
- Surrey
I thought the comparison was between just putting the pumps in parallel or series not the blocks aswell, since that is what most of this thread has been discussing.
Soldato
- Joined
- 12 Sep 2012
- Posts
- 11,704
- Location
- Surrey
double post
Last edited:
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,528
in theory a 2nd block in parallel would be 1/8th of the pressure drop of 2 blocks in series - in practice (because of fittings etc.) it is about a quarter
however the flow rate would be half in each block, so it's a balancing act based on the spec of the pump(s) you have... each time you add a block in series or parallel you reduce the flow rate by something... you can add a pump in series to overcome pressure drop, or you can parallel some components to decrease pressure drop at the expense of halving the flow rate in those components
you also need to bear in mind that if you parallel items, and you then at some point come back together to go through another component, the pressure drop of that 1 component will be dramatically increased as you are trying to force a very high flow rate through that one point
and then you need to stop and think, just how many pumps in series can I add before I'm pushing so much pressure that a fitting pops from the pressure
my advice would be not to run 8 CPU blocks in a single loop, as they benefit from high flow (parallel bad), but also tend to be the highest pressure drop items in your loop (would need very high head pressure in series = bad)
Last edited:
Aye, that's the sensible answer. It's not the cheap one though.