F5F Stay Refreshed Hardware Desktop Extreme variation in core temperature by 7900 times

Extreme variation in core temperature by 7900 times

Extreme variation in core temperature by 7900 times

I
iRaine
Posting Freak
800
08-25-2022, 10:19 PM
#1
I've been using a 7900X for a couple of years. Recently, it's been direct die cooled (using der8auer tools). Before this change, I experienced higher CPU temperatures—around 50°C on the core 2 and 100°C on the core 8, with thermal throttling on the core 8. It's been an issue that hasn't fully come to my attention, but it's been bothering me lately. UNTIL NOW! (maybe) I think the problem might be related to the die being more convex than standard dies. I've tried various mounting pressures and torque settings to rule out a simple mounting issue, but I'm convinced it's about die height consistency. Since these chips have been in use for some time, do you know anyone who has run a Skylake X chip with a lapped die using liquid metal? Is there any evidence that it eventually leaks into the chip and causes damage? I recall this was a concern back in 2018, but I haven't seen many long-term users with setups like mine. If fixing it is your best bet, do you have any advice on affordable calipers and very fine sandpaper? I already checked Harbor Freight, but they don’t carry what I need. Also, could you suggest a way to make the cooler’s cold plate more concave? That might reduce the risk to my CPU?
I
iRaine
08-25-2022, 10:19 PM #1

I've been using a 7900X for a couple of years. Recently, it's been direct die cooled (using der8auer tools). Before this change, I experienced higher CPU temperatures—around 50°C on the core 2 and 100°C on the core 8, with thermal throttling on the core 8. It's been an issue that hasn't fully come to my attention, but it's been bothering me lately. UNTIL NOW! (maybe) I think the problem might be related to the die being more convex than standard dies. I've tried various mounting pressures and torque settings to rule out a simple mounting issue, but I'm convinced it's about die height consistency. Since these chips have been in use for some time, do you know anyone who has run a Skylake X chip with a lapped die using liquid metal? Is there any evidence that it eventually leaks into the chip and causes damage? I recall this was a concern back in 2018, but I haven't seen many long-term users with setups like mine. If fixing it is your best bet, do you have any advice on affordable calipers and very fine sandpaper? I already checked Harbor Freight, but they don’t carry what I need. Also, could you suggest a way to make the cooler’s cold plate more concave? That might reduce the risk to my CPU?

A
Aimone
Member
73
09-04-2022, 03:05 AM
#2
ShrimpBrime could be aware of how flat your cooler looks.
A
Aimone
09-04-2022, 03:05 AM #2

ShrimpBrime could be aware of how flat your cooler looks.

_
_Annabelle_
Member
72
09-04-2022, 10:44 AM
#3
I have some experience with these topics. The 50c between cores isn't too much, usually I see differences around 20c. A direct die should stay under a water block. It tends to need a bit more clamping force than an air-cooled one, though it might risk cracking the silicon. Personally, I wouldn’t stick with LM for long-term use. An MX-4 would work just as well. If you’re reusing the IHS plate, it’s best to lap the bottom side where the glue bonds to the silicon. Use wet sanding paper from auto stores. For the silicon die, go with 2000 grit. Dry the process and apply light pressure in figure 8 patterns. Cooler lapping is similar, but start at about 800 grit and increase gradually. The best results come on a thick mirror surface. Secure the paper in place so it stays steady.
_
_Annabelle_
09-04-2022, 10:44 AM #3

I have some experience with these topics. The 50c between cores isn't too much, usually I see differences around 20c. A direct die should stay under a water block. It tends to need a bit more clamping force than an air-cooled one, though it might risk cracking the silicon. Personally, I wouldn’t stick with LM for long-term use. An MX-4 would work just as well. If you’re reusing the IHS plate, it’s best to lap the bottom side where the glue bonds to the silicon. Use wet sanding paper from auto stores. For the silicon die, go with 2000 grit. Dry the process and apply light pressure in figure 8 patterns. Cooler lapping is similar, but start at about 800 grit and increase gradually. The best results come on a thick mirror surface. Secure the paper in place so it stays steady.

M
Maisiemoo12
Member
154
09-09-2022, 01:05 AM
#4
All good details. Thanks everyone. Right now I'm using the big Noctua air cooler. I had a liquid loop but it was made of aluminum, so I learned the hard way that aluminum and gallium don't mix. I'm checking the cooler's cold plate for flatness since I'm not sure what to measure without special equipment. While I'm in school, I'm also working part-time at an auto parts shop, hoping to find tools or sandpaper for engine building that could help. Noted on the LM and reusing the IHS. I'll try some MX-4 paste to see if it helps with temperature, just because it's thicker. If that doesn't work, I might use liquid metal and just push my CPU until it fails. I'd rather spend money on my car than my PC right now. Appreciate the help!
M
Maisiemoo12
09-09-2022, 01:05 AM #4

All good details. Thanks everyone. Right now I'm using the big Noctua air cooler. I had a liquid loop but it was made of aluminum, so I learned the hard way that aluminum and gallium don't mix. I'm checking the cooler's cold plate for flatness since I'm not sure what to measure without special equipment. While I'm in school, I'm also working part-time at an auto parts shop, hoping to find tools or sandpaper for engine building that could help. Noted on the LM and reusing the IHS. I'll try some MX-4 paste to see if it helps with temperature, just because it's thicker. If that doesn't work, I might use liquid metal and just push my CPU until it fails. I'd rather spend money on my car than my PC right now. Appreciate the help!