Order of Parts in custom loop

[Aaralli]

If I were cooling my CPU and GPU with a custom watercooling loop, and i had two radiators, could I do Reservoir>Pump>CPU>

Radiator>GPU>second radiator>reservoir?

 

Or would it be better to do both heat sources then both radiators?

[Master Disaster]

It makes zero difference. Loop order does not matter.

[Mick Naughty]

You can do what ever order you want. Wont really matter. Would try to explain that the res doesn't always need to be directly before the pump but so many people would reiterate it so it would be mute.

 

Go for the cleanest or easiest.  Should be the only factor.

[straight_stewie]

The order of loop components does not matter at all, for steady state conditions.

Loop order and type can affect ΔT and ΔT/time (the speed and acceleration of temperature change) but this is a non-sequitur when designing a cooling system for a daily driver machine, and would only be applicable to designing for short running overclocking benchmarks (although I highly suspect that anyone going this far to set an overclocking record has already moved on to LNO₂). The only reason to care about this would be if you are specifically trying to set a liquid cooled overclocking benchmark record.

Of far more importance to the design of long running cooling systems is the cooling capacity of the system, as this is what dictates how effective a cooling solution is. If you already have the components, this can be easily optimized by carefully placing radiators and selecting fans and fan direction for them to assure that each one gets the cleanest air possible while still maintaining good enough airflow through the case to cool the non-liquid-cooled components.

As a highly related aside to this discussion, the volume of coolant in the system does not affect steady state cooling either. It too only affects ΔT and ΔT/time. This is why reservoir-less systems perform as well as a reservoir system that is otherwise identical. Reservoir-less systems are, however, much more difficult to design and maintain: The primary purpose a reservoir serves in a cooling system is to easily handle the consequences of the Ideal Gas Law. It is only a side effect that having a reservoir can make a system easier to fill and drain. Consequently, this allows very wide latitude when selecting a reservoir: Volume of a reservoir is of minimal concern, and the decision of which reservoir to use can be driven almost entirely by fitment, aesthetics, and ease of maintenance. (The assumption here is that computers generally operate in a relatively narrow temperature band, and therefore there is not much fluctuation in fluid volume. Trivially, the reservoir drain must never allow aeration through the drain and the reservoir must never be overfull)

[Aaralli]

5 hours ago, straight_stewie said:

The order of loop components does not matter at all, for steady state conditions.

Loop order and type can affect ΔT and ΔT/time (the speed and acceleration of temperature change) but this is a non-sequitur when designing a cooling system for a daily driver machine, and would only be applicable to designing for short running overclocking benchmarks (although I highly suspect that anyone going this far to set an overclocking record has already moved on to LNO₂). The only reason to care about this would be if you are specifically trying to set a liquid cooled overclocking benchmark record.

Of far more importance to the design of long running cooling systems is the cooling capacity of the system, as this is what dictates how effective a cooling solution is. If you already have the components, this can be easily optimized by carefully placing radiators and selecting fans and fan direction for them to assure that each one gets the cleanest air possible while still maintaining good enough airflow through the case to cool the non-liquid-cooled components.

As a highly related aside to this discussion, the volume of coolant in the system does not affect steady state cooling either. It too only affects ΔT and ΔT/time. This is why reservoir-less systems perform as well as a reservoir system that is otherwise identical. Reservoir-less systems are, however, much more difficult to design and maintain: The primary purpose a reservoir serves in a cooling system is to easily handle the consequences of the Ideal Gas Law. It is only a side effect that having a reservoir can make a system easier to fill and drain. Consequently, this allows very wide latitude when selecting a reservoir: Volume of a reservoir is of minimal concern, and the decision of which reservoir to use can be driven almost entirely by fitment, aesthetics, and ease of maintenance. (The assumption here is that computers generally operate in a relatively narrow temperature band, and therefore there is not much fluctuation in fluid volume. Trivially, the reservoir drain must never allow aeration through the drain and the reservoir must never be overfull)

This is the most well-thought-out response I e ever gotten on this forum and I REALLY appreciate it. Thank you very much.  I do have a couple things to say about it though. So I bought a Corsair Crystal 680x case, which I plan to insert 280mm radiators at the top and bottom, and have the three front fans and single rear fan cooling everything that isn't watercooled. And I do plan to overclock my Ryzen 7 3800x and RTX 2080 Super as much as I can, so cooling might become more difficult than just running at stock. I'm sure the two radiators is enough, but is it too much?

[straight_stewie]

11 minutes ago, Aaralli said:

I'm sure the two radiators is enough, but is it too much?

The general goals of moving to liquid cooling are to generate a higher cooling capacity and to reduce the noise generated by the cooling solution.

To affect these goals, the general approach is to add as much radiator fin area as possible within space, budget, and aesthetic constraints. We have to be careful here though, as fin geometry can affect the fan requirements, and thus affect the amount of noise generated. Given constant physical space, there is a trade-off between cooling performance and noise.

But, there are some things we can do to minimize this trade-off. Firstly is to use only as much radiator area as required to get the cooling capacity we need, as this allows for a lower amount of airflow per radiator for the same cooling capacity. The second thing we can do is to carefully select radiators that are highly efficient, that is, they cool well for a given amount of airflow.

Basically, just select your radiators from well known brands like EKWB, HardwareLabs (if you can find them), Alphacool... These companies generally categorize their radiators based on performance, and the higher the performance, the higher the noise output (usually, anyway, as this is also affected by fan selection).

What I do is to use EKWB's configurator to see what radiators they recommend, and then I can compare similar radiators from other manufacturers to their suggestions to get a good idea of what radiators will probably meet my goals.

[PowerBaller]

If your plan is a simple serial loop, letting the coolest part of your loop connect to the inlet of your GPU block optimizes the performances. 

However your best option, if your pump is D5, is a parallel loop. D5 is known for high flow rate but low hydraullic pressure, the parallel loop take full advantage of that.

 

This is my novice build (the parallel loop). It takes short for the liquid flowing through a cycle either in the CPU or the GPU section, compare to the serial counterpart.