CPU + GPU in parallel loops?

[VanGoghComplex]

I most commonly see loops like this in show-quality builds where function seems to follow form. I know in theory that if the CPU and GPU blocks have similar levels of restriction, both should see similar levels of coolant flow. In practice though, does it work?

 

[Pandur]

As long as the theory is good, you should see similar results in practice. Ergo if your two blocks have the same level of restriction you'll see a 50/50 split of flow between the two. The problem is that in practice I'd be willing to bet they don't have identical tech specs and you'll end up with something like 70% flow through one block and 30% through the other. Or a lot worse even.

 

[thesmokingman]

I most commonly see loops like this in show-quality builds where function seems to follow form. I know in theory that if the CPU and GPU blocks have similar levels of restriction, both should see similar levels of coolant flow. In practice though, does it work?

No, it does not work er it works but not well. Differences in restriction will mean the water will flow thru the path of least resistance. Much like the rule of never getting into a land war in Asia, you never plumb dissimilar blocks in parallel.

 

[Brian_B]

If you want parallel loops, you should make sure each loop has it's own pump.

Path of least resistance will see more flow. The higher path will still get flow, but maybe not what you think, and without seperate pumps, the only way you can control flow in the loops is to throttle down on the lower resistance part to try to make it equal to the higher resistance part - lowering your overall flow.

Even in heavy industry pumping applications - it's the same thing: use separate pumps when possible, when it's not, you have to use flow balancing valves. See this a lot in like HVAC.

Will the picture above work? Yeah, probably. But if the CPU starts running hot - which it very well could because most of that flow will get going through the GPU - then you have no way to adjust it to fix the problem - apart from splitting them back out to series run or entirely separate loops.

 

[VanGoghComplex]

Well, applying what I know from electrical engineering, there should be more at play here, theoretically. What you're all saying is correct; I know about the path of least resistance.

However, whereas in series, the resistance seen by the pump is the sum of the resistance created by both blocks, the resistance seen by the pump in a parallel loop would be less than the least resistive block. In theory that means overall flow rate would be higher in a parallel loop, right? That would offset the lower flow through the more resistive block, at least to a degree.

I'm curious if anyone's ever done a write-up on this using flow meters to test just how much difference there is. Anyone seen anything like that?

 

[Nimisys]

I have been running a full parallel setup for 6 years, with 3 blocks total. Pump to rad to 3-way splitter to feed each block, back into a 3-way splitter to another rad and then res. Raystorm CPU block, and universal GPU blocks for about 4 years, then same CPU block and EK heatkiller full coverage GPU blocks.

Guess what? The difference in blocks and supposedly flow rates meant that jack shit for temps. Never had an issue with GPUs or CPU overheating in comparison with each other. Kept an oc'd 2600k and later 3770k in check during Intel burn test without issue, would hold oc'd 7970s below 50c during gaming.

 

[thesmokingman]

Well, applying what I know from electrical engineering, there should be more at play here, theoretically. What you're all saying is correct; I know about the path of least resistance.

However, whereas in series, the resistance seen by the pump is the sum of the resistance created by both blocks, the resistance seen by the pump in a parallel loop would be less than the least resistive block. In theory that means overall flow rate would be higher in a parallel loop, right? That would offset the lower flow through the more resistive block, at least to a degree.

I'm curious if anyone's ever done a write-up on this using flow meters to test just how much difference there is. Anyone seen anything like that?

There are a few articles but mostly on pumps these days. Here's a guide from the guide from back in the day. It's old but the fundamentals are still true. Btw, check the write up on hard lines.

http://wc101.com/articles/vs