TIM Pumping out/Drying Quickly. - 7950X

E

equinox654

Gawd
Joined
Apr 7, 2005
Messages
852
I am having a strange problem with my AM5 build.
7950X / EK 360 AIO / Kryonaut Extreme TIM

On fresh application my idle is ~38C. Web browsing and general windows usage temps are 40-60 depending on boost.
I noticed my idle was in the 50s and my all core boost that was holding at 5.45-5.5GHZ dropped to 4.6-5GHZ. That was after the paste being on for little more than a weekend.
I repasted the Kryonaut Extreme and temps went back to normal. A day later and it is almost as bad as it was. Idle was in the low 50s again and fans ramping quick.

On fresh Application I get over 38k in Cinebench R23. On the first application it dropped to 30K, testing this morning it was 36k. I repasted this morning and its back up to over 38K with low idle.

Originally, I used the EK Ectotherm that came in with the cooler I am not sure if it had the same problem, but I remember the fans ramping similar to when the Kryonaut stopped working.

The tube of Kryonaut Extreme is pretty much empty now. I found a tube of mx-4 from a couple of years ago. I may try that next.

image_67215873.JPGimage_67200001.JPGimage_67184641.JPGIMG_0547.jpg

It isn't as clear in the images, but a lot of the IHS is bare and the bare spots match with the cold plate, so it was dry. I noticed that the IHS has gouged the cold plate. I know how to change a tire... I tightened the thumbscrews in a cross pattern while keeping tension balanced to prevent one corner having alot more pressure than the others.
The last picture shows my calipers being used as a straight edge you can see the gap at the top of the cold plate. I know these are machined convex. Not sure if that is too much. I didn't get a picture of it, but I checked the IHS with a credit card and it seems as flat as I can see with my eyeballs. No where near as extreme as the cold plate.

I wonder if the IHS being so thick is causing the cold plate to not be able to deform the IHS into being concave, causing extreme pressure in the center and pumping out all of the TIM.
Maybe lap it?

Have any of you ever ran into this before or have any advice?
 
Honestly all that looks fine including the curvature. Its kinda hard to tell how the tim is sitting as it kinda deforms and suck out of the tight spots when you pull the cooler off.

I would try it with the mx4 ive never had a issue with that paste degrading in performance over any period of time.

If it really is pumping out im wondering what type of deflection you are getting when this setup heats up
 
cdabc123 said:
Honestly all that looks fine including the curvature. Its kinda hard to tell how the tim is sitting as it kinda deforms and suck out of the tight spots when you pull the cooler off.

I would try it with the mx4 ive never had a issue with that paste degrading in performance over any period of time.

If it really is pumping out im wondering what type of deflection you are getting when this setup heats up
Click to expand...
Yeah it is hard to say.


The Kryonaut Extreme is supposed to be stable at 80c and under. With that super thick IHS you would think the temps the TIM are exposed to would be less than 80c, but I was running stress tests... not over night, but several minutes here and there to get RAM and curve optimizer dialed in and it was seeing 245 watts at times. You would think the thick IHS would keep the TIM from being exposed to temperatures over 80.
Maybe it is a bad batch. I did check the card that came with it. It isn't counter fit.
 
equinox654 said:
Yeah it is hard to say.


The Kryonaut Extreme is supposed to be stable at 80c and under. With that super thick IHS you would think the temps the TIM are exposed to would be less than 80c, but I was running stress tests... not over night, but several minutes here and there to get RAM and curve optimizer dialed in and it was seeing 245 watts at times. You would think the thick IHS would keep the TIM from being exposed to temperatures over 80.
Maybe it is a bad batch. I did check the card that came with it. It isn't counter fit.
Click to expand...
80 seems low. Perhaps as you approach that point the viscosity changes enough that the geometry of the ihs and heatsink cause a issue.

I don't think the ihs and heatsink are perfectly compatible as far as shape. Laping can be inconsistent as it doesn't take into account thermal deflection when cycling. I think a different perhaps less extreme tim is the solution for a decent, reliable thermal setup. Mx-4 is older but ive found its very robust and have ran it under pretty extreme thermal situations with success. If it works well for you I would just run that and leave 2c or so on the table.

Also probably won't be a issue but scoring on the mating surface could cause a problematic gap. Since its on the edge I would just continue mounting the heatsink in the same way and pressure you have been to attempt to keep it lined up with the chip edge.
 
cdabc123 said:
80 seems low. Perhaps as you approach that point the viscosity changes enough that the geometry of the ihs and heatsink cause a issue.

I don't think the ihs and heatsink are perfectly compatible as far as shape. Laping can be inconsistent as it doesn't take into account thermal deflection when cycling. I think a different perhaps less extreme tim is the solution for a decent, reliable thermal setup. Mx-4 is older but ive found its very robust and have ran it under pretty extreme thermal situations with success. If it works well for you I would just run that and leave 2c or so on the table.

Also probably won't be a issue but scoring on the mating surface could cause a problematic gap. Since its on the edge I would just continue mounting the heatsink in the same way and pressure you have been to attempt to keep it lined up with the chip edge.
Click to expand...
I could go over it with some fine grit sandpaper I suppose. Make sure a burr that I can't feel isn't lifting it.
 
equinox654 said:
I could go over it with some fine grit sandpaper I suppose. Make sure a burr that I can't feel isn't lifting it.
Click to expand...
At that point I would just lap the CPU and heatspreader. I dont think there is much of a reliable way to knock down exclusively that edge. And if there was that Ridge was mating with a elevated or pressure point of the current chip setup. May be better to just leave it and hope it mates in the same way that created the grove
 
I'd RMA/return it. Seems like they installed the heatspreader and glued it high on the corner. Can you lay the cpu flat to see if you see it raised on the bottom left?
 
hititnquitit said:
Do you have any other heatsinks to test with?
Click to expand...
Not really. I have a Corsair H150 that I am handing down to my son with my 9900k, but lost the amd bracket a couple of years ago.. Guess I could order one.

Rvenger said:
I'd RMA/return it. Seems like they installed the heatspreader and glued it high on the corner. Can you lay the cpu flat to see if you see it raised on the bottom left?
Click to expand...
I’ll check that out. Hopefully no returns.
 
Pretty sure I solved the problem. It did it again, so I lapped the bitch took about .3mm off. Dropped temps 5c in Firestrike.
Cinebench r23 still pegs out 95, but takes a few seconds to get there instead of instantly now. That is using my old MX-4 that was performing pretty meh before the lapping.
Have some more pastes coming in tomorrow.
 
equinox654 said:
Pretty sure I solved the problem. It did it again, so I lapped the bitch took about .3mm off. Dropped temps 5c in Firestrike.
Cinebench r23 still pegs out 95, but takes a few seconds to get there instead of instantly now. That is using my old MX-4 that was performing pretty meh before the lapping.
Have some more pastes coming in tomorrow.
Click to expand...
Wow, that makes me wonder wth is going on with the stock ihs. Glad you're closer to a solution I would be curious what the pattern of the mx4 looks pike when ya pull the block off and how much you can gain with a better paste
 
cdabc123 said:
Wow, that makes me wonder wth is going on with the stock ihs. Glad you're closer to a solution I would be curious what the pattern of the mx4 looks pike when ya pull the block off and how much you can gain with a better paste
Click to expand...
1666372746165.jpeg
The notches are chamfered, but it looks like a channel runs in a bit. It was hard to capture with a camera.
I don't know if it came that way, or if the AIO caused it. I think that would be enough for it to pump out with the heat load of a 7950x.
If you look back at my original post. I think it was flowing into the bottom and right channels.

Pretty excited for new paste. The MX-4 I have is from a build 2 or 3 years ago and isn't performing well. The kryonaut exteme idled in the 30s and I could sustain loads of 220+watts.
Before lapping the MX-4 would idle in the 40s and sustain loads around 178watts. After lapping it idles at 38 and loads around 218watts. Wish I had a microcenter closer than an hour and a half away.

Aso need to lap the AIO a little better, I rushed that to make sure I didn't kill the chip. Need to measure how much cooler contact I still have. Maybe take a bit more off the cpu and see if I can get it off of 95c in all core loads.
 
equinox654 said:
View attachment 520302
The notches are chamfered, but it looks like a channel runs in a bit. It was hard to capture with a camera.
I don't know if it came that way, or if the AIO caused it. I think that would be enough for it to pump out with the heat load of a 7950x.
If you look back at my original post. I think it was flowing into the bottom and right channels.

Pretty excited for new paste. The MX-4 I have is from a build 2 or 3 years ago and isn't performing well. The kryonaut exteme idled in the 30s and I could sustain loads of 220+watts.
Before lapping the MX-4 would idle in the 40s and sustain loads around 178watts. After lapping it idles at 38 and loads around 218watts. Wish I had a microcenter closer than an hour and a half away.

Aso need to lap the AIO a little better, I rushed that to make sure I didn't kill the chip. Need to measure how much cooler contact I still have. Maybe take a bit more off the cpu and see if I can get it off of 95c in all core loads.
Click to expand...
so what paste did you get and how well did it resist pump-out?
 
With that particular AIO every TIM I tried pumped out between a day and a week. I ended up building a custom loop and applied the same kryonaut and it’s been fine for months.
 
Pumping out most of the paste is good. You only need a tiny haze like micron level of paste to fill the micro gaps. The amount of paste actually needed if you could apply it equally across the surface to the required thickness would probably equate to 15% of what you put on.
 
Yeah, but in this case… I had good scores and thermals and after a day I would have full speed fans and hard cpu throttling. Hell opening an explorer window ramped the fans.

It could have been thermal breakdown of the paste. I don’t know. The IHS just looked a lot cleaner than what I am used to. I even checked by pulling it to observe while it had good temps and there was a difference. A sporadic film with dry patches with most of the film concentrated at the top and bottom and sides with the die area mostly bare metal.
 
equinox654 said:
Yeah, but in this case… I had good scores and thermals and after a day I would have full speed fans and hard cpu throttling. Hell opening an explorer window ramped the fans.

It could have been thermal breakdown of the paste. I don’t know. The IHS just looked a lot cleaner than what I am used to. I even checked by pulling it to observe while it had good temps and there was a difference. A sporadic film with dry patches with most of the film concentrated at the top and bottom and sides with the die area mostly bare metal.
Click to expand...
It should look like in the pictures as TIM becomes an insulator if there is too much. If you look at your pictures you can see that most of the paste is on the coldplate. The custom loop water blocks from EK have very very high mounting pressure so it has been like that with all my mounts on my 5900x with most of the paste squeezed out, but temps were good. Did swap for techn as it was significantly better at lower flow rates and a bit better at max flow rates. Wasn't quite as happy with kryonaut as it lost a few degrees performance after a few months, but hydronaut was good for at least 8+ months with little or no degradation. The stuff that comes with the custom loop stuff from EK was probably hydronaut, at least it looked and behaved very similar to the hydronaut paste I bought.

It is important to lock fan speed and run to stable state when comparing mounting/TIM and ideally monitor the water temp. In a water cooled system the delta between water temp and CPU temp will generally be quite constant unless something is wrong. If you set fans to follow a slow curve then the water temp will slowly rise and offset results quite a bit between a cold booted system and one that has been running with small background tasks. It can be as much as 6-7 degrees difference. This can be offset by running to steady state, basically the point where there is no more change. On a 360 AIO it will usually take 5-10 minutes while a large custom loop can take 30 minutes to reach steady state. My understanding is that the 7950x systems are supposed to throttle, at least on an AIO, and a few degrees might be the difference between lower performance and average watts.

Is it not possible to run the fans off the water temp on EK AIOs? At least that is what works best for water cooled systems as the temp will slowly creep up on idle if the fans are coupled to CPU temp and run speeds in the sub 400 rpm range. Most fans don't do much on a radiator below 600-700rpm due to need for static pressure. EK Vardar don't move that much air through the radiator until you reach 800-900RPM in my experience, but difference between 300 and 600 is still quite large as they barely move any air through the radiator at 300rpm.
 
It was just plain not cooling. The turbo in cinebench went from 5.2ghz to something in the low 4ghz range. And a repaste would always bump it back up. This occurred overnight. Every night for a couple of weeks until I built my custom loop. I was repasting to be able to work lol.
 
You must log in or register to reply here.
Back
Top