- Joined
- Aug 16, 2005
- Messages
- 27,937 (3.84/day)
- Location
- Alabama
| System Name | RogueOne |
|---|---|
| Processor | Xeon W9-3495x |
| Motherboard | ASUS w790E Sage SE |
| Cooling | SilverStone XE360-4677 |
| Memory | 128gb Gskill Zeta R5 DDR5 RDIMMs |
| Video Card(s) | MSI SUPRIM Liquid 5090 |
| Storage | 1x 2TB WD SN850X | 2x 8TB GAMMIX S70 |
| Display(s) | 49" Philips Evnia OLED (49M2C8900) |
| Case | Thermaltake Core P3 Pro Snow |
| Audio Device(s) | Moondrop S8's on Schitt Gunnr |
| Power Supply | Seasonic Prime TX-1600 |
| Mouse | Razer Viper mini signature edition (mercury white) |
| Keyboard | Wooting 80 HE White, Gateron Jades |
| VR HMD | Quest 3 |
| Software | Windows 11 Pro Workstation |
| Benchmark Scores | I dont have time for that. |
Starting out I picked up a NUC 12 Extreme. This variant with the 12900 (non-k. More on this later). This small project (GET IT??) is the exploration into modification and making it run the best I can given its limited upgrade options.
My specs are in my profile.
To start cooling. The reason I even own this is the replacement of my 7980XE and my midtower/H2O machine. It is amazing the amount they manage to pack into this. The compute unit itself is small and about the size of a midrange GPU but holds the RAM/CPU mobo and all I/O in one. Overall its impressive as far as engineering is concerned. Like the past few gens of nuc extreme specifically this CPU is socketed and otherwise a normal desktop processor.
The first upgrade on the list was a 12900k. You can see this was done previously by der8auer to save you time it didn't work. Recent BIOS updates lockout the BIOS override which locks the CPU at 400mhz. Further new versions of Intel XTU do not expose the power overrides and now show the system as unsupported. The 12900k was sent back. That is fine though. I only wanted it for the slightly higher boost clock. No overclocking was going to be done on this machine. With my vanilla 12900 in hand we continue.
With this formfactor and CPU generation however come obvious concerns.
Gen 12 mobo flex
Heat
Thats about it. So lets take a look. First we need to verify there is a problem. Temps will be posted later but out of the box it was high. I tested most of this using the same in game settings in FF14. With a replicatable area. Is it as specific and scientific as I have been? No, but I watch my temps so for what it may be worth to you salt or not temps were stable regarding the figures that come later.
First was to replace the TIM to drop temps. This worked and the factory goo while applied fantastically just wasnt as good as the thermal grizzly kryonaught I applied. This was a good time to test my warping. Yeah even these suffer from it.
So lets continue.
A quick look showed my PCH temps surpassing 100C under load. Thats absolutely wild to me. So I took a look. The backplate for the NUC had a thermal pad that appeared to be too thin and didnt even appear to touch the PCH. Lets get the boys together.
We need to fix the warping:
We need to better cool the PCH with a better pad and get a proper size on it:
Now we need to help the NUC the cooler is tiny:
Perfect lets get to work. SEVERAL DAYS LATER
Now all the stuff is in. Lets get it worked on.
First we need to work on the PCH, this was simple and I forgot to take pics. I replaced the pad and it could finally touch.
The next was ripping apart the cooler so we could replace the CPU bracket.
Get ye eld paste off. Just look at the dead spot.
Get our new bracket back on using the OEM screws and retention plate.
Now we throw LM on it and re-assemble. Lets check the retention plate again.
Oh yeah much better.
Finally lets take a look at temps.
PCH Range (Before): 97-112ºC
PCH Range (After): 79-88ºC
CPU Range (Before OEM): 74-89ºC
CPU Range (Kryonaught): 72-81º
CPU Range (Conductonaut): 67-74ºC
This is impressive to me. Even giving the variable CPU load (not 100%) the movement of the temp brackets is impressive, doubly so given that it wasnt 100% CPU load. For them to move that much paints the picture of the importance of cooling in SFF builds.
Its important to note here in the case of a 65w product like the 12900 the most beneficial result of the lowered temps isnt the temps themselves exactly, as even at its worst it was within operating parameters, but instead the duration and frequency of the boost clocks. Which are regulated by both power draw and heat. Given the limited nature of SFF and low watt products in general, the more we do and the more we maintain boost clocking the better the overall experience.
Q/A:
Are you scared of using LM on small expensive things and why LM because I'm mad and %reasons%?
No: https://www.techpowerup.com/forums/threads/7980xe-delid.251203/
Why not measure max load?
Good question! it would be useless. For these kind of system you want to measure the temp bracket you operate in. The cooler is too small to maintain 100% load without heat soaking. You could probably bring the delta down, you know time till saturation. But you wouldnt be able to cure it.
My specs are in my profile.
To start cooling. The reason I even own this is the replacement of my 7980XE and my midtower/H2O machine. It is amazing the amount they manage to pack into this. The compute unit itself is small and about the size of a midrange GPU but holds the RAM/CPU mobo and all I/O in one. Overall its impressive as far as engineering is concerned. Like the past few gens of nuc extreme specifically this CPU is socketed and otherwise a normal desktop processor.
The first upgrade on the list was a 12900k. You can see this was done previously by der8auer to save you time it didn't work. Recent BIOS updates lockout the BIOS override which locks the CPU at 400mhz. Further new versions of Intel XTU do not expose the power overrides and now show the system as unsupported. The 12900k was sent back. That is fine though. I only wanted it for the slightly higher boost clock. No overclocking was going to be done on this machine. With my vanilla 12900 in hand we continue.
With this formfactor and CPU generation however come obvious concerns.
Gen 12 mobo flex
Heat
Thats about it. So lets take a look. First we need to verify there is a problem. Temps will be posted later but out of the box it was high. I tested most of this using the same in game settings in FF14. With a replicatable area. Is it as specific and scientific as I have been? No, but I watch my temps so for what it may be worth to you salt or not temps were stable regarding the figures that come later.
First was to replace the TIM to drop temps. This worked and the factory goo while applied fantastically just wasnt as good as the thermal grizzly kryonaught I applied. This was a good time to test my warping. Yeah even these suffer from it.
So lets continue.
A quick look showed my PCH temps surpassing 100C under load. Thats absolutely wild to me. So I took a look. The backplate for the NUC had a thermal pad that appeared to be too thin and didnt even appear to touch the PCH. Lets get the boys together.
We need to fix the warping:
We need to better cool the PCH with a better pad and get a proper size on it:
Now we need to help the NUC the cooler is tiny:
Perfect lets get to work. SEVERAL DAYS LATER
Now all the stuff is in. Lets get it worked on.
First we need to work on the PCH, this was simple and I forgot to take pics. I replaced the pad and it could finally touch.
The next was ripping apart the cooler so we could replace the CPU bracket.
Get ye eld paste off. Just look at the dead spot.
Get our new bracket back on using the OEM screws and retention plate.
Now we throw LM on it and re-assemble. Lets check the retention plate again.
Oh yeah much better.
Finally lets take a look at temps.
PCH Range (Before): 97-112ºC
PCH Range (After): 79-88ºC
CPU Range (Before OEM): 74-89ºC
CPU Range (Kryonaught): 72-81º
CPU Range (Conductonaut): 67-74ºC
This is impressive to me. Even giving the variable CPU load (not 100%) the movement of the temp brackets is impressive, doubly so given that it wasnt 100% CPU load. For them to move that much paints the picture of the importance of cooling in SFF builds.
Its important to note here in the case of a 65w product like the 12900 the most beneficial result of the lowered temps isnt the temps themselves exactly, as even at its worst it was within operating parameters, but instead the duration and frequency of the boost clocks. Which are regulated by both power draw and heat. Given the limited nature of SFF and low watt products in general, the more we do and the more we maintain boost clocking the better the overall experience.
Q/A:
Are you scared of using LM on small expensive things and why LM because I'm mad and %reasons%?
No: https://www.techpowerup.com/forums/threads/7980xe-delid.251203/
Why not measure max load?
Good question! it would be useless. For these kind of system you want to measure the temp bracket you operate in. The cooler is too small to maintain 100% load without heat soaking. You could probably bring the delta down, you know time till saturation. But you wouldnt be able to cure it.
Last edited:
