Water cooling of 0.5-1 kW processors

[CyklonDX]

the best one will not work if the waterblock can't get the heat.
wrong pressure.
too much/few/wrong thermal paste
hot spot wrong place. remember at SP5 Turin the heat comes from outer sides not central.

(as a note this isn't my problem, but op is looking for solutions to cool down 500W-1kW cpu - actual workload is unknown)

sure to a point, one can can still use vapor chamber or peltier inbetween cpu ihs and heatsink if heatsink's internal structure doesn't line up with whats under ihs.
ex. VC214 | Dynatron

(in terms of paste i'd rather use honeywell ptm7950 or thermalright helios v2.)

 

[DanRR]

my CPU eats 350W and the Cooler FAN needs 12Watts to cool 350W away.

Not just yours which has efficiency or coefficient of performance (COP) 350/12 ~30 , generally all the fans remove heat more efficiently than any heat pumps which have efficiency not larger than 5 and mostly 2-3, rarely 4. All these SEER, EER, BTU show some large numbers and they looks like were specifically designed by the salespeople just to trick and confuse the buyer because they show 3.41 times larger numbers than the absolutely clear parameter COP typically used in all other countries

 

[DanRR]

Here is how fans performed when I substituted stock 25mm fans on Silverstone waterloop radiator to Arctic 38mm ones.

Changing all three fans to Arctic 4K dropped temperature by 5 C. If one of these 3 Arctic 4K fans substitute with Arctic 8K the temperature dropped already just by 1 C, so the total 6 C

Easier variant: substituting just one single Silverstone fan with Arctic 8K also gave the same 5-6C drop of temperature with approximately the same noise like with 3 Arctic 4K

TrashMaster's movie above was pretty scary respect to the level of noise from his 4K and 8K. But in reality the operating noise is much much smaller so I would call it nonexistent because many other fans also produce combined noise. Hence the voltage lowering is not needed.

For Turins needed drop by 10C. Silverstone would easily do that with 38mm 4K Arctic fans and the increase of thickness of radiators by a half of inch. But for now what I can do is to find the best way to drop the room temperature for these missing 4C. And during hot days of summer for additional 5-10 C. During winter it is easy just to remove extra heat with fans, but during summer additional cooing will be needed.

This is not the end of story. As we are moving to GPUs, NVIDIA already has chips dissipating 1.4 kW of power.

 

[RolloZ170]

Not just yours which has efficiency or coefficient of performance (COP) 350/12 ~30

but it don't work with too much/wrong thermal paste, U2 cooler (must be 3 or 4U)
that is what i want to say.
the good with Air cooling is the VRMs get some air too.

 

[RolloZ170]

For Turins needed drop by 10C. Silverstone would easily do

i guess the waterblock fins are centered to the IO-DIE. this makes the construction sensible to thermal paste goodness.
some overclockers get 4C only with shifting waterblock to hotspot on AM4/AM5 sockets.

 

[DanRR]

I will try to substitute one more 4K with 8K and will see if that will shed couple degrees more. Though I really doubt about that, and doubt that any miracle thermopaste will help, there actually exists some basis for that. Specifically that the temperature difference accurately measured is 13-14 C for the half of cores running to >20C when all cores are used. Here is confirmation. First picture showing 38.5 C at the point of entrance of water into radiator, the second is the temperature of air (or fan fins) 25.4 C on entrance of fans for the half load case.







It is also possible to direct the cold air from outside exactly on fans. Room and street temperature could sometimes reaching 40 C at summer but at winter it is 0 - 10 C and that will give 10-20 degrees drop in processor temperature. Minuscule potential drop by miracle thermopaste is fading compared to that. But it is necessary to find how to protect the system from water during rainy stormy days