- Joined
- Jul 21, 2018
- Messages
- 773 (0.31/day)
- Location
- Germany
| System Name | FATTYDOVE-R-SPEC |
|---|---|
| Processor | Intel i9 10980XE |
| Motherboard | EVGA X299 Dark |
| Cooling | Water (1x 240mm, 1x 280mm, 1x 420mm + 2x Mo-Ra 360 external radiator) |
| Memory | 64GB DDR4 |
| Video Card(s) | RTX 2080 Super / RTX 3090 |
| Storage | Crucial MX500 |
| Display(s) | 24", 1440p, freesync, 144hz |
| Case | Open Benchtable (OBT) |
| Audio Device(s) | beyerdynamic MMX 300 |
| Power Supply | EVGA Supernova T2 1600W |
| Mouse | OG steelseries Sensei |
| Keyboard | steelseries 6Gv2 |
| Software | Windows 10 |
Hello lads and lasses,
Today I tested liquid metal (Thermal Grizzly Conductonaut) vs. thermal paste (Thermal Grizzly Kryonaut) on top of my stock 10980XE IHS.
I got inspired by @buildzoid :
He noticed for his Ryzen 3950X that there is a significant improvement in load-temp if you use liquid metal between cooler and IHS instead of paste.
This got me interested in trying it for myself on my 10980XE, as that thing can use every little headroom it can get.
I use watercooling with a Mo-Ra 360 radiator in a loop just for the CPU with the Aquacomputer 'cuplex kryos NEXT' block. The block is full metal with nickel plating.
The water-temperature was stable at 23°C during both tests.
Heads-up I did not test very well, as I had a higher voltage on my liquid metal run. This was due to the system being unstable with prime testing a day earlier and I failed to re-run cb20 on paste and I don't want to re-apply everything just to fix that. It will still be a good demonstration.
10980XE @4.7GHz / 1.15V
Thermal paste:
This resulted in 77°C on the hottest core, average core temp was 70.67°C.
Next up is my result from today with liquid metal on the stock IHS. No lapping done!
This time again 4.7GHz (with BCLK, see score for proof) but at 1.17V.
Now the hottest core is at 70°C. The average is 64.61°C.
So there is a whole 7°C drop on my hottest core and a 6°C drop on my average temp across all cores while at the same time I ran a slightly higher voltage.
As kinda expected, the more heat you have to move, the more liquid metal can display its higher transfer capability. I'm really happy with the result.
Still you have to keep in mind that you need to handle liquid metal with caution. Only use it when you feel comfortable doing so and know what you are doing.
It is highly conductive, it will probably remove the etching on your chip even if you don't lap it and void any warranty!
I took some pre-caution steps by masking the CPU around the upper edges of the substrate with Kapton-Tape and extending that roughly 1cm around the CPU in the socket so that even IF there are any spills they are caught by the tape. Luckily I seem to have gotten the right amount on the first try, I could not see any spills from the side of the socket and temps indicate a good contact.
Probably not necessary but I also plumbed the hole on the IHS with some paste and put a thin layer of liquid rubber on top of that hole to prevent any liquid metal from ever entering the CPU. I preferred the liquid rubber approach over glue as the rubber can still give to pressure and be a little flexible just in case that pressure under the IHS from heat could be a problem (I seriously doubt that but I feel better this way).
I'm a lunatic and chase optimal performance without going sub-ambient. Thank you buildzoid for bringing this up in one of your videos and bringing my attention to it.
Today I tested liquid metal (Thermal Grizzly Conductonaut) vs. thermal paste (Thermal Grizzly Kryonaut) on top of my stock 10980XE IHS.
I got inspired by @buildzoid :
He noticed for his Ryzen 3950X that there is a significant improvement in load-temp if you use liquid metal between cooler and IHS instead of paste.
This got me interested in trying it for myself on my 10980XE, as that thing can use every little headroom it can get.
I use watercooling with a Mo-Ra 360 radiator in a loop just for the CPU with the Aquacomputer 'cuplex kryos NEXT' block. The block is full metal with nickel plating.
The water-temperature was stable at 23°C during both tests.
Heads-up I did not test very well, as I had a higher voltage on my liquid metal run. This was due to the system being unstable with prime testing a day earlier and I failed to re-run cb20 on paste and I don't want to re-apply everything just to fix that. It will still be a good demonstration.
10980XE @4.7GHz / 1.15V
Thermal paste:
This resulted in 77°C on the hottest core, average core temp was 70.67°C.
Next up is my result from today with liquid metal on the stock IHS. No lapping done!
This time again 4.7GHz (with BCLK, see score for proof) but at 1.17V.
Now the hottest core is at 70°C. The average is 64.61°C.
So there is a whole 7°C drop on my hottest core and a 6°C drop on my average temp across all cores while at the same time I ran a slightly higher voltage.
As kinda expected, the more heat you have to move, the more liquid metal can display its higher transfer capability. I'm really happy with the result.
Still you have to keep in mind that you need to handle liquid metal with caution. Only use it when you feel comfortable doing so and know what you are doing.
It is highly conductive, it will probably remove the etching on your chip even if you don't lap it and void any warranty!
I took some pre-caution steps by masking the CPU around the upper edges of the substrate with Kapton-Tape and extending that roughly 1cm around the CPU in the socket so that even IF there are any spills they are caught by the tape. Luckily I seem to have gotten the right amount on the first try, I could not see any spills from the side of the socket and temps indicate a good contact.
Probably not necessary but I also plumbed the hole on the IHS with some paste and put a thin layer of liquid rubber on top of that hole to prevent any liquid metal from ever entering the CPU. I preferred the liquid rubber approach over glue as the rubber can still give to pressure and be a little flexible just in case that pressure under the IHS from heat could be a problem (I seriously doubt that but I feel better this way).
I'm a lunatic and chase optimal performance without going sub-ambient. Thank you buildzoid for bringing this up in one of your videos and bringing my attention to it.