XFX 7900XT merc 310 high junction temperature

nhtlquan

nhtlquan

New member
I also had the same problem. When playing games, the temperature will reach 90-105 degrees.
7900XTX Merc 310 + Alphacool Eisblock Aurora Water Block

Cooling system: EK D5 pump, 3 radiator 360.

I have tried removing and reinstalling more than 10 times. The hotspot temperature results are all very high.

Thermal Paste is spread evenly, the 1mm Thermal Pads used have matching ridges.

@Vanzin I might try new thermal pads. Please give me the right thickness so I can test.

1687232848950.png
 
DirtyAlpaca

DirtyAlpaca

Member
Dear Alphacool and readers

The first results are in with the larger custom pad on the backplate.
I have repasted the die with Kryonaut and applied new pads because the older ones were a bit torn apart from opening up a few times.
All the screws were thightened down whilst creating downward pressure on the center of the waterblock, right above the core.


The test ran on the same ambient temperature as my opening post.

Let's compare with my previous post.
Stock 350W board power
Core from 51 down to 44 degrees Celcius
Hotspot from 91 down to 83 degrees
Delta 40 to 39 degrees
350w stock side panel open.png

Using an infrared temperature meter here. Aiming on the backplate, right above the core. Measuring 3°C better conductivity with the newer pad.
Click here for the original pad picture. The secondary I received from Vanzin covers the entire core, with the idea it would create downward pressure and better heat transfer.
boschscan.jpg

Stock Power Limit 15% 400W board power
Core from 53 down to 47 degrees Celcius
Hotspot from 105 down to 96 degrees
Delta 52 to 49 degrees
400w stock clocks.png

Conclusion
Changing the backplate core thermalpad to a larger one did not change the final outcome.
Those 4 screws holding down the backplate do not offer enough pressure on the center, since they are located on the 4 edges.
This might have been different if there were extra backplate screws near the core.

The PCB remains bent and the delta between core and hotspot remains too high in my opinion with a wopping 39 °C.
But it was worth trying out whether the pad would increase central pressure.

I'm interested if the CORE block would keep the card within spec.
The good part
I didn't hit the thermal wall of 105/110°C during the 400W heat test with the power limit set to 15%.
 
Last edited:
direstorm

direstorm

New member
DirtyAlpaca said:
Dear Alphacool and readers

The first results are in with the larger custom pad on the backplate.
I have repasted the die with Kryonaut and applied new pads because the older ones were a bit torn apart from opening up a few times.
All the screws were thightened down whilst creating downward pressure on the center of the waterblock, right above the core.


The test ran on the same ambient temperature as my opening post.

Let's compare with my previous post.
Stock 350W board power
Core from 51 down to 44 degrees Celcius
Hotspot from 91 down to 83 degrees
Delta 40 to 39 degrees
View attachment 5243

Using an infrared temperature meter here. Aiming on the backplate, right above the core.
Measuring 3°C better conductivity with the newer pad.
View attachment 5244

Stock Power Limit 15% 400W board power
Core from 53 down to 47 degrees Celcius
Hotspot from 105 down to 96 degrees
Delta 52 to 49 degrees
View attachment 5239

Conclusion
Changing the backplate core thermalpad to a larger one did not change the final outcome.
Those 4 screws holding down the backplate do not offer enough pressure on the center, since they are located on the 4 edges.
This might have been different if there were extra backplate screws near the core.

The PCB remains bent and the delta between core and hotspot remains too high in my opinion with a wopping 39 °C.
But it was worth trying out whether the pad would increase central pressure.

I'm interested if the CORE block would keep the card within spec.
The good part
I didn't hit the thermal wall of 105/110°C during the 400W heat test with the power limit set to 15%.
Click to expand...
That is disappointing. I'm actually getting close to 105 degrees Celsius when playing games. I was thinking of using less thick pads on the front to see if the IHS made better contact but I'm just guessing at this point. I might even try liquid metal if I get desperate enough.
 
DirtyAlpaca

DirtyAlpaca

Member
direstorm said:
That is disappointing. I'm actually getting close to 105 degrees Celsius when playing games. I was thinking of using less thick pads on the front to see if the IHS made better contact but I'm just guessing at this point. I might even try liquid metal if I get desperate enough.
Click to expand...
Interesting idea, but I'm sorry to counter this idea.
Let me explain why.

Imprint 1mm dark grey pads original aircooler
Let's take a look to the 1mm thermal pads on the original XFX heatsink first.
In my opinion this imprint on the 1mm pads look exeptionally good. (dark grey ones)

xfx original heatsink.jpg

Imprint 1mm dark grey pads Alphacool waterblock
Hang on there, i'm giving my thoughts below :)
I'm seeing less pressure here on the pads. Nothing to panic about ofcourse. Just sharing info.
Though I notice less pressure on the thermalpaste in the middle. Where the GCD is located in the die.

imprint eisblock.jpg

Could these be the problem?
I'm only marking two of them to prevent clutter. Nothing wrong with these two in particular, but these offsets, standoffs, spacers, washers. However you want to call them. They might be marginally too high or maybe uneven. Even though this seems impossible since Alphacool probably uses top tier 3d scanners and ports the model over to their CNC machine. Try screwing them tight.

What I'm basically saying is that changing your pads to thinner ones will worsen your performance. Because these washers limit your PCB.
washers eisblock.jpg

Closer view
washer zoom eisblock.jpg

What's up next?
This depends on Alphacool's feedback. It's unfortunate, but for the time being I'm not going to modify this block. Would kill the 2y waranty.

Changing the washers after measuring with a lasermeter would be an option.
Modifying the PCB to release tension on the center so the bulge can disappear would be an option.

Ordered a Thermal Grizzly Kryosheet for testing purposes.
I highly recommend trying out AMD Chill to prevent unnecessary heat dispersion in games where you don't benefit the extra power until there's a proper solution to our junction temperature.

We'll see.

Enjoy the weekend all.
 
Last edited:
direstorm

direstorm

New member
DirtyAlpaca said:
Interesting idea, but I'm sorry to counter this idea.
Let me explain why.

Imprint 1mm dark grey pads original aircooler
Let's take a look to the 1mm thermal pads on the original XFX heatsink first.
In my opinion this imprint on the 1mm pads look exeptionally good. (dark grey ones)

View attachment 5247

Imprint 1mm dark grey pads Alphacool waterblock
Hang on there, i'm giving my thoughts below :)
I'm seeing less pressure here on the pads. Nothing to panic about ofcourse. Just sharing info.
Though I notice less pressure on the thermalpaste in the middle. Where the GCD is located in the die.

View attachment 5248

Could these be the problem?
I'm only marking two of them to prevent clutter. Nothing wrong with these two in particular, but hese offsets, standoffs, spacers, washers. However you want to call them. They might be marginally too high or maybe uneven. Even though this seems impossible since Alphacool probably uses top tier 3d scanners and ports the model over to their CNC machine. Try screwing them tight.

What I'm basically saying is that changing your pads to thinner ones will worsen your performance. Because these washers limit your PCB.
View attachment 5246

Closer view
View attachment 5245

What's up next?
This depends on Alphacool's feedback. It's unfortunate, but for the time being I'm not going to modify this block. Would kill the 2y waranty.

Changing the washers after measuring with a lasermeter would be an option.
Modifying the PCB to release tension on the center so the bulge can disappear would be an option.

Ordered a Thermal Grizzly Kryosheet for testing purposes.
I highly recommend trying out AMD Chill to prevent unnecessary heat dispersion in games where you don't benefit the extra power until there's a proper solution to our junction temperature.

We'll see.

Enjoy the weekend all.
Click to expand...
I wonder if those washers are removable. I will be experimenting myself including switching thermal paste and adding a thicker layer. I am suspecting that there is gap between the IHS and the waterblock.
 
direstorm

direstorm

New member
direstorm said:
I wonder if those washers are removable. I will be experimenting myself including switching thermal paste and adding a thicker layer. I am suspecting that there is gap between the IHS and the waterblock.
Click to expand...
After doing some more experimenting, I think my theory that the waterblock is not making solid contact with the IHS is turning out to be correct. I took it apart again and this time put a thick layer of thermal paste on the GCD and when I put it back together, I screwed on the bolts as tightly as possible without stripping the bolt head on both the front and the backplate. After it was assembled, I used my fist to pound that backplate that was bulging due to the thermal pads that were too thick. Believe it or not, it did help. For testing purposes, I was able to get Port Royal to go from 107Celcius to 100Celsius when overclocking and playing The Last of Us, I was nowhere near the 105Celsius it used to be. It was near 80Celsius or lower which was more to my liking. Hopefully, this helps everyone else who has this problem.
 
Last edited:
nhtlquan

nhtlquan

New member
direstorm said:
After doing some more experimenting, I think my theory that the waterblock is not making solid contact with the IHS is turning out to be correct. I took it apart again and this time put a thick layer of thermal paste on the GCD and when I put it back together, I screwed on the bolts as tightly as possible without stripping the bolt head on both the front and the backplate. After it was assembled, I used my fist to pound that backplate that was bulging due to the thermal pads that were too thick. Believe it or not, it did help. For testing purposes, I was able to get Port Royal to go from 107Celcius to 100Celsius when overclocking and playing The Last of Us, I was nowhere near the 105Celsius it used to be. It was near 80Celsius or lower which was more to my liking. Hopefully, this helps everyone else who has this problem.
Click to expand...
Can i see the gcd and pcb image you made?
 
nhtlquan

nhtlquan

New member
I screwed on the bolts as tightly as possible without stripping the bolt head on both the front and the backplate
I don’t understand?
direstorm said:
After doing some more experimenting, I think my theory that the waterblock is not making solid contact with the IHS is turning out to be correct. I took it apart again and this time put a thick layer of thermal paste on the GCD and when I put it back together, I screwed on the bolts as tightly as possible without stripping the bolt head on both the front and the backplate. After it was assembled, I used my fist to pound that backplate that was bulging due to the thermal pads that were too thick. Believe it or not, it did help. For testing purposes, I was able to get Port Royal to go from 107Celcius to 100Celsius when overclocking and playing The Last of Us, I was nowhere near the 105Celsius it used to be. It was near 80Celsius or lower which was more to my liking. Hopefully, this helps everyone else who has this problem.
Click to expand...
 
DirtyAlpaca

DirtyAlpaca

Member
direstorm said:
After doing some more experimenting, I think my theory that the waterblock is not making solid contact with the IHS is turning out to be correct. I took it apart again and this time put a thick layer of thermal paste on the GCD and when I put it back together, I screwed on the bolts as tightly as possible without stripping the bolt head on both the front and the backplate. After it was assembled, I used my fist to pound that backplate that was bulging due to the thermal pads that were too thick. Believe it or not, it did help. For testing purposes, I was able to get Port Royal to go from 107Celcius to 100Celsius when overclocking and playing The Last of Us, I was nowhere near the 105Celsius it used to be. It was near 80Celsius or lower which was more to my liking. Hopefully, this helps everyone else who has this problem.
Click to expand...


I think we can both agree on same thing leading us back to the first post of this thread. A GPU does not have an IHS, but I do know what you meant back there.(y)

It does help to create central pressure on the PCB before fixing the 4 screws around the core, but does not fix the high delta between Core and Junction. I could try a clamp before fixing the 4 center screws with some improvising...

I'm waiting comfortable for Alphacool's solution and I'm also quite sure we will settle this straight together thanks to the public support forum.
 
Last edited:
Vanzin

Vanzin

Support
Staff member
HI @DirtyAlpaca

Thank you for testing it and the detailed documentation. Our Develepment is now running out of ideas how to fix it, without a affected GPU and a water block to test it. I will write you a PN with a probosal.
 
D

dcox12

New member
Hi,

sorry for my interference but I'm interested in that water block. And I want to buy it when this issue is fixed. Thus, I want to present my theory:

I think those standoffs might too short. The GCD is surrounded by a metal frame (correct me if I am wrong but the metal frame should be level with the GCD as it's there to protect the GCD from unevenly mounted coolers). When you tighten the screws, the water block will push against this metal frame. The metal frame itself won't give way and thus the whole PCB is bending under the water block's pressure. The GCD is right in the middle of the bending in the PCB an thus will - ever so slightly maybe - lose contact with the cooler.

It might be counterintuitive but is it possible that you thighten the screws only so much that the PCB doesn't bend? Would you give it a shot?
 
Last edited:
nhtlquan

nhtlquan

New member
z4465965307607_c7360697473daabdc7086ffb871221bb.jpg


1687831279908.png


Tried dozens of ways but didn't work.
I ended up having to use Liquid Metal.
Room temperature: 28c
Fullload: 550W
GPU Heat: 38c
Hotspot: 69c
 
DirtyAlpaca

DirtyAlpaca

Member
direstorm said:
That's funny considering I just ordered liquid metal as well. Hope I know what I'm doing. Only thing I'll be doing differently is adding liquid tape around the GCD. Glad it worked out for you.
Click to expand...
Dear direstorm

Hope to see some pictures and results from your modification. I tried to interpret "tape around the gcd", but my imagination is limited to understand that. :)
I had amazing results with liquid metal on my intel 8700K delid back in the days.


My card and block was sent to Alphacool for a closer look.
 
You must log in or register to reply here.
Top