Eisbaer Aurora HPE 360 x 9800X3D poor performance

skankeh

New member
I have recently switched from i7 12700k (lapped heat spreader and a contact frame) to 9800X3D and I don't think the AiO is performing on the new CPU as it should. The Aurora fans have been replaced with Lian Li TL fans that have higher static pressure and airflow than the Aurora fans.

I was running the i7 with 250w of power draw in OCCT, staying below Tmax (87-90 celcius). When at this level of load the air coming out of the rad was rather warm.

I have now switched to 9800X3D, using AMD AM4 mounting (no AM5 specific mounting hardware was included but it should be the same). Running OCCT with no PBO hits Tmax after a few seconds, the air out of the radiator is not warm. Applying PBO and under-volt (I forgot the exact under-volt) keeps just below or on Tmax but with only 5Ghz clock speed despite +100Mhz on PBO, again the air is not getting very warm. I checked the mounting of the cold plate and it looks like there is not enough pressure on the heat spreader, the layer of TIM (Thermal Grizzly Kryonaut) is quite thick still compared to on the i7.

Has anyone any ideas on this, or is there now a specific AM5 mounting solution?
 

Eddy

Iceman
Staff member
Mounting on AM5 is identical to AM4 as nothing has changed in the socket holder. Do you have a picture of the imprint of the thermal paste please?
Have you checked whether the pump is still running properly?
Are you still using the nuts to pre-assemble the retaining screws on the mounting frame? If so, please remove them.

Basically, you can't compare the cooling performance on an AMD with an Intel. These are completely different architectures that dissipate their waste heat differently at certain points. If the AMD generated just as much waste heat as the Intel, it would only be coolable with dry ice. Intel dissipates its heat over a wider area and is therefore easier to cool than an AMD. Added to this is the X3D design, which makes cooling even more difficult.

Nevertheless, at room temperatures of around 25° you should be able to achieve a core temperature below 90°.
 

Vanzin

Support
Staff member
HI,

Your CPU can reach under full load nearly 90°C. To check if the cooler has any problems, we will need a photo from the imprint of the thermal paste and did you check if the pump is running with 12V and shows a RPM of 2600 +/-10%?
 

skankeh

New member
I will take a picture of the imprint, though I won't be able to today. The pump is running at just over 2500 rpm according to tachometer readings in HWinfo and in Fan Control v209. I am using the nuts, but they are not "getting in the way" of anything. I understand there are differences to Intel, however the new Ryzen 9 series X3D has the 3D cache mounted under the CPU die so the CPU die has direct contact with the heat spreader, this should improve the cooling efficiency compared to Ryzen 7 X3D chips. I provide the Intel result as an indication the AiO rad and pump are performing well and able to dissipate a high thermal load.
 

skankeh

New member
HI,

Your CPU can reach under full load nearly 90°C. To check if the cooler has any problems, we will need a photo from the imprint of the thermal paste and did you check if the pump is running with 12V and shows a RPM of 2600 +/-10%?
Hi Vanzin,

Yes, I am expecting nearly 90 celcius, I am hitting Tmax though and have lower clock rates with PBO enabled.
 

Eddy

Iceman
Staff member
Leave the nuts off. It is not so much that they are “in the way”, but rather that they can reduce the contact pressure if they are not screwed on tightly enough. They also have no purpose other than to hold the mounting bolts in position to make it easier to mount. But we have already had some customers who couldn't cope with this. They are therefore no longer included with the newer models.

The new Ryzen are just as difficult to cool as the old ones. Or we currently cool almost all CPUs. Yes, the relocation or reversed structure of the X3D cache makes it a little easier, but it doesn't make a big difference.
 

skankeh

New member
Leave the nuts off. It is not so much that they are “in the way”, but rather that they can reduce the contact pressure if they are not screwed on tightly enough. They also have no purpose other than to hold the mounting bolts in position to make it easier to mount. But we have already had some customers who couldn't cope with this. They are therefore no longer included with the newer models.

The new Ryzen are just as difficult to cool as the old ones. Or we currently cool almost all CPUs. Yes, the relocation or reversed structure of the X3D cache makes it a little easier, but it doesn't make a big difference.
I mean the nuts are not preventing mounting pressure as they are not making contact with the cooler mount screw holes, however I will remount with them off. Do you think the offset mount, potentially, can make more difference with the Ryzen 9 X3D?
 

Eddy

Iceman
Staff member
Do you think the offset mount, potentially, can make more difference with the Ryzen 9 X3D?
Not really in this case. The offset kit only brings ~2°, which won't help you here. There is a fundamental problem here.
 

skankeh

New member
HI,

Your CPU can reach under full load nearly 90°C. To check if the cooler has any problems, we will need a photo from the imprint of the thermal paste and did you check if the pump is running with 12V and shows a RPM of 2600 +/-10%?
I think this shows I may have an issue, they are cooling with a Noctua NH-D15 (a very strong air cooler, I know) and getting 86.7 celcius. I am guessing that the Blender CPU test is not as stressful as OCCT though, but on a 360 AiO with high CFM and static pressure fans I expected better.
 

skankeh

New member
Not really in this case. The offset kit only brings ~2°, which won't help you here. There is a fundamental problem here.
Let's see what remounting without the nuts brings, I'll also take a photo of the TIM when I do.
 

Eddy

Iceman
Staff member
Blender CPU test is not as stressful as OCCT though
Very similar.
I think this shows I may have an issue, they are cooling with a Noctua NH-D15 (a very strong air cooler, I know) and getting 86.7 celcius.
They test on an open system. This is not comparable with a closed housing. You always have higher internal temperatures in the housing and therefore always poorer cooling performance than on an open benchtable.

The advantages of AIOs are that they blow the waste heat directly out of the case, whereas with air coolers, the waste heat goes round in circles and cannot be sucked out directly by the rear fan. Thermaltake once supplied an exhaust duct with their Macho cooler to minimize the problem. But it looked terrible, didn't always fit into the case and then disappeared from the market very quickly.
 

skankeh

New member
20241211_154206894_iOS.jpg20241211_154235346_iOS.jpg


I noticed something on the pump, may be nothing. There is a small bubble, approximately 7mm across in the pump window, it doesn't seem to be moved by the flow of water, I remember when I first installed the AiO a few months ago on the Intel system that there was a small, 2mm bubble and it was agitated by the water flow. I can feel the vibrations from the pump through the hoses and Fan Control is reporting 2550rpm approx. I tried using a straight 12v SATA power adaptor to see if there is a difference and there doesn't seem to be. Perhaps as the bubble is much larger than earlier it isn't moved easily by the water flow. I have also removed the retaining nuts from the installation bolts and remounted the cooler.

OCCT temperature is now higher than the last mounting by about 3-4 Celcius. 1.05 Vcore 147w CPU Package power.
 

Vanzin

Support
Staff member
HI Thanks for the photos. THe imprints looks okay. Did you check the clock of your CPU, because AMD´s CPU´s are working different von Intel. THey are boosting until they are reaching the maximal allowed temperature of 94°C and from this point one, they will lower the clock to hold the temperature. Because of this most review sides don´t like to make reviews with AMD CPU´s.
 

skankeh

New member
Running OCCT for 45 minutes:

1734000654999.png

I have PBO set to 5200 Mhz max. With 155w CPU package power I was hoping to hit 5200Mhz under load.

I'm a bit surprised by the amount of TIM myself. It's my first Ryzen CPU (last AMD CPU was Athlon 64!) and my Intel ones would always have a noticeably lesser amount (though I nearly always lapped the heat spreader on those). The idea is to just fill in the tiny gaps between heat spreader and cold plate after all, but if you think that's okay I trust your judgement.
 

Vanzin

Support
Staff member
Hi,

Thanks for the information but depending from the information I can´t see any problems. The imprints is okay and the CPU is running within AMD specifications. As long as your CPU is not below the normal clock of 4,7Ghz, everything is okay. The boost clock will be hold till the temperature limit is reached and from there on the clock will be automatically lowered until it reach standard clock to hold the temperature. Did you check if the CPU has reached the Boost clock of 5200MHz and how long did it hold it?
 

Eddy

Iceman
Staff member
150W? Are you sure? That is much more then i see on different sides from different people. They are all below 100W with temperatures between 85 and 95°. Which tool are you using? Please try HWInfo. That is the only real proper tool for meassurings.
 

skankeh

New member
1734018621575.png

CPU doesn't hit 5200Mhz, only 5050Mhz

Edit: It does hit 5200Mhz but not with all cores loaded. With all cores it hits 5050Mhz
 

Eddy

Iceman
Staff member
To be honest, I can't think of anything else. I would play around with the voltage a bit more. The impression is good, but I can't understand why the cooling doesn't work as it should.
I can only think of things like an incorrect air flow in the housing and therefore excessively high internal temperatures in the housing and similar. Sorry, but at this point I have no idea why you have such problems.
 

skankeh

New member
It is a bit of a mystery, I do appreciate all the time you all spend on helping me here.

The airflow in my case is good, 6 intake fans and 3 exhaust through the radiator. GPU runs cool and motherboard temps are good.

Is it possible that the flow of water is restricted somehow despite the pump running 100%? I keeps seeing this bubble sat against the window not moving, I would expect it to move around in water flow.
 
Top