The pressure thing is not important, except for the pump inlet, so the argument that the rad should go after the block is invalid. The flow rate is the same at every point in a loop: those that don't believe that don't know what they're talking about.
The exception, of course, is parallel setups.
Blocks in series will each see all of the flow, but the total pressure drop will reduce the pump output.
Blocks in parallel will see part of the total flow, but they'll allow the pump to provide more flow.
So it's down to 2 choices:
-all of less
or
-half of more
In your case, I think Airspirit and I agree that "half of more" (aka parallel blocks) would be best.
This is where it gets tricky.
If the blocks have exactly the same pressure drop for a given flow rate, then your total flow will be split equally.
If that's not the case (and it applies to you), then one block will get more flow than the other. You can solve this by throttling the flow to the GPU block, with a ball valve.
As for the order, it should be:
res-->pumps(in series)-->splitter-->blocks-->splitter-->rad-->back to res , as you pointed out last. this will (as Airspirit wrote) reduce the temp going into the pump(s).
|