How to FIX AW17R4/15R3/13R3 CPU...
Hi all, i'm starting a new discussion and as the title suggest, this is a place to discuss ways and methods used to fix CPU core differential problem on all new alienware models.
Background of the issue.
New system shipped from factory, even as recent as May 2017 are suffering from extreme core differential problem even at stock speed. True the problem is a little bit better now with the Kaby lake models, but it is still there and wide spread. I must say, over 70% of the systems being shipped came with this problem, if not all.
How to detect this issue.
Easy. Run a CPU stress program like OCCT or wPRIME and monitor your CPU temperature. If all 4 CPU cores are within 3-5C degrees (example: 75C, 76C, 74C, 77C), you are one of the very few lucky ones. If 2 of your CPU cores are 10C+ hotter than the other 2 cores (example: 80C, 90C, 82C, 92C), your system is suffering from this issue. note that the core differential problem always come in pairs. Core 0, 2, or Core 1, 3 depending on which area on the CPU die has less contact with the heatsink. The problem is also AMPLIFIED significantly for those who OC their 6820HK/7820HK CPU without undervolt.
Cause of the problem.
The heatsink of the CPU is not making proper contact with the CPU die surface. Especially the first half upper region of the CPU. This is due to the heatsink of the CPU resting at an angle (when viewed sideways) on the CPU die surface due to 2 main problems.
1. Poor CPU tripod tension arm design, weakest point at the single arm top part of the CPU. not providing enough pressure to allow the CPU heatsink to sit flush with the CPU die.
2. Thermal pads. use of thermal pads thicker than you should on the power delivery components on top of the CPU raised the heatsink up at the top side and the already weak cpu tension arm does not help at all. the initial skylake models were even worse using thick thermal pads on the larger silver chokes which were rectified later with the very thin white pads. This enhancement was done in part due to iunlock & Team LHz holding discussion with dell engineering team. the CPU core temperature differential reduced from over 20C to 10C.
How to fix the problem.
From what i see around the forum, most people who have successfully fixed this issue are using liquid metal (LM). LM has such high thermal conductivity, by just applying thicker layer on the top half of the cpu, it effectively plugged the gap between the heatsink and cpu die.
For those who use traditional paste like i do, Grizzly kryonaut, it is near impossible to fix this issue easily. even with thermal paste as good as kryonaut, the thermal conductivity is pale in comparison to LM. the only person whom i've seen successfully get even temperatures across all core using traditional paste is again, iunlock.
How he did it? i have no idea. iunlock if you are reading this, your input is greatly appreciated.
I have tried many non invasive methods, including
1. changing thermal paste, putting a thicker layer at the top part of the cpu
2. repad the VRM (smaller black chips on top of the larger square silver Chokes) with a thinner 1mm fujipoly pad and leaving the 0.1mm white thermal pads on the silver chokes from factory.
3. tried different screw combination (the heatsink screws are labeled 1,2,3 which indicates which screw you should tighten first).
4. pressing the heatsink against the CPU when fastening the scews
5. bending the tripod top tension arm as iunlock suggested
Basically, NON of the above method works. to date, i have possibly done repad and repaste more than 10 times combined with 6820HK and 7820HK AW17R4. to date i have not been able to reduce core differential to less than 10C.
Things yet to try
1. Mobius 1 suggested to replace the 0.1mm white thermal pad from factory with actual thermal paste for an even thinner contact to reduce the gap between the cpu and heatsink.
2. invasive methods like sand paper the heatsink, using mallet to knock the heatsink copper, etc.
Those who have this issue, or those who are able to fix this issue, please share and discuss over here. and note that NOT all heatsink are created equal, but it's a good point of reference.
Read responses in forum.printercomparison.com