Cooling a 200W 5900X

[pdifolco]

I'm now proud owner of a 5900X , and I've played a bit with PBO2
Using increased mobo PPT TDC EDC limits (X570 board)
Undervolting to -30 ticks

Results are good, except on the heat side, as max power consumption is now 200W, and temps quickly rises to 85C+ under stress/benchmark (CBR23 for example)

I use an Eisbaer 280mm customized with Arctics P14s fans, and suspect it ain't enough for that kind of power... I went back to stock limits to avoid all that heat

Would adding a D5/VPP pump to the circuit gives good results ? Or is the rad just too small for 200W ?

 

[Eddy]

You will hardly achieve better temperatures with a larger radiator. The AMD CPUs of the 5000 series suffer extremely from the problem that heat is hardly transferred from the DIE via the heatspreader to the cooler. And especially the 5900 is even more problematic than a 5950 due to the chip layout.

 

[pdifolco]

You will hardly achieve better temperatures with a larger radiator. The AMD CPUs of the 5000 series suffer extremely from the problem that heat is hardly transferred from the DIE via the heatspreader to the cooler. And especially the 5900 is even more problematic than a 5950 due to the chip layout.

Interesting, but it's kinda what I thought because the heat spike occurs so fast (like 3 sec from 40C to 80C+ running CBR23) that I wondered how any cooling could manage this...
Well will keep stock limits then, it keeps temps under 75C which is reasonable

 

[Eddy]

An extremely rapid rise in temperatures is normal. This is simply because the temperatures are not measured but calculated. There is no temperature sensor in any CPU, everything is only calculated based on various factors. As soon as clock and voltage are applied, which happens almost immediately, the temperatures are calculated accordingly and set high.

 

[pdifolco]

An extremely rapid rise in temperatures is normal. This is simply because the temperatures are not measured but calculated. There is no temperature sensor in any CPU, everything is only calculated based on various factors. As soon as clock and voltage are applied, which happens almost immediately, the temperatures are calculated accordingly and set high.

Ok, didn't know that
So is there any cooling solution (short of LN !) to get better temps by like 4C (which is the difference in my usual gaming between stock and increased limits 75C vs 80C, in CBR23 it's 12C less but it's just a benchmark)?

 

[Eddy]

More radiator area would add a few degrees, but whether that equals 4° is hard for me to tell you. I would rather look at what you can still turn down the voltage of the CPU. That brings by far the most. You could also try a little more downforce, but that carries risks. You could experiment a bit more with the thermal paste. Use a little more or a little less. But that won't help that much either. With a DIY cooling I would try a little more flow, but that is not possible with the Eisbaer if you run the pump on max. Personally, I would play around with the voltages of the CPU. AMD and also Intel usually give more voltage on it than actually necessary.

 

[pdifolco]

More radiator area would add a few degrees, but whether that equals 4° is hard for me to tell you. I would rather look at what you can still turn down the voltage of the CPU. That brings by far the most. You could also try a little more downforce, but that carries risks. You could experiment a bit more with the thermal paste. Use a little more or a little less. But that won't help that much either. With a DIY cooling I would try a little more flow, but that is not possible with the Eisbaer if you run the pump on max. Personally, I would play around with the voltages of the CPU. AMD and also Intel usually give more voltage on it than actually necessary.

Already did that, using -20 ticks undervolt on standard limits and -30 with increased EDC, can't go any farther !
Will add a VPP755 pump, repaste and see what happens

 

[pdifolco]

Happy ending !

Adding a VPP755 pump+150mm res, and repasting using a larger quantity of paste (Arctic MX-5) did it, I'm now at 36C idle open case, 40C closed and under my desk, then gaming went from 80ish to 65C and full load from 85C+ to 78C...
Using board limits so up to 200W with high EDC 215A and TDC 130A, CO (voltage) at -15 to -20, still experimenting

It looks like sh#t in the case tho due to tubes being way too long, will have to fix that soon

Seems I was mislead by advices on paste using smaller Intel chips, using more of it and spreading it gives much better results.. Maybe the cooling circuit lacked coolant, now I'm sure it has enough!