Why do people directly connect CPU and GPU water blocks together?

[idwwwoqq808]

I've had this question for quite a long time. Why do most people connect CPU block directly to GPU blocks without radiators in between while building their water cooling system? It's also very common in manufacturers' example cooling systems. The coolant is heated by CPU block and then flows to GPU block without being cooled, and this is not good for cooling the GPU, isn't it?

 

For example, I have two water cooling systems for gaming PCs, both using the same cases and 360 radiators:

CPU block → GPU block → Radiator 1 → Radiator 2 → Distro Plate (with pump)

CPU block → Radiator 1 → GPU block → Radiator 2 → Distro Plate (with pump)

Do these two systems have same or very different thermal efficiency? If it's same or close, what is the reason? I personally prefer the second one, by the way.

[LeDoyen]

the difference between the water coming out of the radiators (cool) and the water coming out of CPU and GPU block (hot) is so small it doesn't matter. at worse you may have one or two degrees, maybe less depending on your pump speed.

You will not notice any meaningful difference at all.

 

Thermal efficiency only depends on your radiator surface anyway. no matter what your loop order is, the dissipation will be the same.

 

Loop order is more of an aesthetic, or personal preference.

If you like to feed both blocks with water coming from radiators, you can do a parallel loop and feed both blocks at the same time.. that's a third option.

 

So yea, the real reason is the temperature difference between hot and cool is too small to matter.

[c-attack]

As noted above, it really doesn't matter and the typical 360mm radiator is going to decrease coolant by about 1C from enter to exit at medium fan speed. So using a simple 2 radiator, 2 block example:

 

CPU->GPU->rad->rad->return might pass a 1-2C penalty onto the GPU (increase in coolant temp while in the CPU block) since it did not go through a radiator, but the CPU might gain a small advantage over the example below since it gets 2 passes in the radiator before return.

 

CPU->rad->GPU->rad->return is down the middle with no penalty to either block.

 

Most people are not in a position where +1C to either component makes a difference to their usage. The usual obstacles to running the tubes and making a somewhat logical pathway around the case are more likely to take priority. That said, for people with new gear that are really interested in the data, I generally recommend block->rad->block->rad if it works so they can take temperature data from each in/out port on the radiator and see the temperature changes at each point in the loop. If it you never even considered putting a T-sensor on every radiator port, then you are not in that group. Most people get by with one at the reservoir just fine.

 

I would say if you are within 1C of hitting some personal CPU limit, you are already too close. For water cooling some GPUs, that 1-2C might have more meaning. The Nvidia Turing (and presumable Ampere) GPUs have definite temp X = negative clock reduction points. My GPU has a large one at 49C and I was sitting right on in Summer months. Through a natural course of trying different configurations, I was able to stay just under by leading with the GPU -> CPU-> radiator series-> return. Fairly niche example, but one of the few.

 

What likely has more relevance to most people is the radiator location in relevance to loop order. Most people in a standard case with a two radiator layout will have one in the front and one up top. The top several inches of your case is always going to be warmer than the front case rail. My preference would be to go to the top radiator first after heat pick up, then to the front radiator 2nd (possibly with CPU block stop on the way). This is more about radiator efficiency and the temperature of the intake air on the entrance side of the radiator. If your exit from the front radiator is 27.0C and you then go to the top radiator where the ambient temp is 28C, you won't be reducing your coolant temp any further and the top radiator becomes a passenger. This like most things with water cooling, this is chasing every last degree. Where it is more essential is if you decide to run a two radiator set-up where the front is intake and top is exhaust. In that circumstance, you must go to the top radiator first and I would really want the GPU going straight to the top radiator. The front in/top out dual radiator combination is a tricky one and not always recommended, but it is a common one for first time builders who are following the standard advice for heat management in an air cooled case. You might also run into this issue in triple radiator layouts where inevitably some of your radiators must exhaust into the intake of another. There you want the maximum temperature point at the top/first radiator to try and keep a temperature difference between intake air and coolant temp for each radiator.

 

There is always an exception to the rule and your case and layout will determine what works best. Someone running a reverse flow from the back to top/front exhaust may not have a warmer temp at the top then in the front of the case.