HUGE core temp difference 7900x

[Schiggy]

I have been running a 7900x for a couple years now. For about the last year it has been direct die cooled (der8auer tools).

 

Since I first did direct die cooling, I have had worse CPU temps. I'm talking like 50c on core 2 and 100 and thermal throttling on core 8. It has been one of those problems that is just at the back of my mind but I haven't gotten around to fixing (busy year)... UNTIL NOW! (maybe)

 

I believe the problem is that my specific die is more convex that usual dies. I tried all sorts of different mounting pressures and torque rotations to make sure it was not just a simple mounting issue, but I am convinced it is a problem with die height consistency.

 

Since these chips have been out a while now, is there anyone who has ran a skylake x chip with a lapped die using liquid metal? Is there any data showing that it eventually seeps through into the chip and breaks it? I remember it being a concern in 2018, but I have not seen any examples of people that ran a setup like mine long term.

 

If this is my best option to fix it, is there any suggestions for decently priced calipers and super fine sand paper? I checked harbor freight already and they don't have what I need lol. Also, is there potentially a way to make my coolers cold plate more concave? that might solve the issue with less risk to my cpu.

[EmoChipmonk]

6 minutes ago, Schiggy said:

I have been running a 7900x for a couple years now. For about the last year it has been direct die cooled (der8auer tools).

 

Since I first did direct die cooling, I have had worse CPU temps. I'm talking like 50c on core 2 and 100 and thermal throttling on core 8. It has been one of those problems that is just at the back of my mind but I haven't gotten around to fixing (busy year)... UNTIL NOW! (maybe)

 

I believe the problem is that my specific die is more convex that usual dies. I tried all sorts of different mounting pressures and torque rotations to make sure it was not just a simple mounting issue, but I am convinced it is a problem with die height consistency.

 

Since these chips have been out a while now, is there anyone who has ran a skylake x chip with a lapped die using liquid metal? Is there any data showing that it eventually seeps through into the chip and breaks it? I remember it being a concern in 2018, but I have not seen any examples of people that ran a setup like mine long term.

 

If this is my best option to fix it, is there any suggestions for decently priced calipers and super fine sand paper? I checked harbor freight already and they don't have what I need lol. Also, is there potentially a way to make my coolers cold plate more concave? that might solve the issue with less risk to my cpu.

i think LTT had a chip that had liquid metal eventually spill on to the substrate.

[Moonzy]

@ShrimpBrime might know

 

how flat is your cooler?

[Guest]

2 hours ago, Moonzy said:

@ShrimpBrime might know

 

how flat is your cooler?

I kinda know some things. ...

 

50c between cores is quite a bit, naturally I've seen them as far as 20c difference.

 

Direct die should be under a water block.

Generally needs more clamping force than clipped on air cooler by a bit.

But with risk to cracking the silicon.

 

I wouldn't do LM for long term personally. MX-4 would do just fine. 

 

When and if reusing the IHS plate, its recommended to lap the bottom side where the glue goes on the IHS plate so it mates well with the silicon.

 

You want wet sanding paper found at auto stores. 2000 grit for the silicon die. Do it dry and very light pressure in figure 8 motions.

 

Cooler lapping is the same, but start at maybe 800 grit and work your way up.

 

Best lapping takes place on a thick mirror. Tape the paper down so it doesnt slide around.

 

 

[Schiggy]

All good info. Thanks guys. I am running the big boi Noctua air cooler at the moment. I had a liquid loop but it was aluminum and I figured out the hard way that aluminum and gallium don't mix. I could check the cold plate on the noctua cooler for flatness cause I really don't know what it is, but I have no idea how I would measure it without really odd tools. Also right now while I'm in school I am working part time at an autoparts store so ill check and see if I can get my hands on any tools/ sandpaper for engine building that might double for this purpose.

 

Noted on the LM and re-using IHS. I will try some MX-4 and see if it helps temps just because its a thicker paste. If that doesn't do it ill probably lap the die, use liquid metal, and just go wild with my cpu till it breaks. I wouldn't mind getting a second gen x299 chip anyway. I would just rather put the money into my car instead of my pc right now.

 

thanks for the help guys! much appreciated!