Hi there, with the release of the new Intel's 7th gen. KabyLake cpu, we heard a lot of talk about bad TIM under IHS or bad IHS quality or is better mount your aio directly on DIE etc. Here i want share some data collected on a retail i7 7700k.
First of all: is worth delid my cpu replacing stock TIM? The answer is yes, but please considere metallic grase mandatory. Looks like high quality silicon based TIM is not enough to worth the job of deliding your cpu. Here you can see an i7 7700k cpu running Cinebench R15 at 5000Mhz CPU / 5000Mhz CACHE and 16GB of DDR4 running 3600 with XMP profile, hitting 92°c on the hotter core. The cpu was cooled with a Raijintek Triton 360 AiO. Between the IHS and Aio was applied the Thermal Grizzly Kryonaut TIM.
After Delid the IHS and replace the stock Intel's TIM with Thermal Grizzly Conductonaut, metallic based TIM, the hotter core has reached 63°c. Still, like before delid, a Raijintek Triton 360 AiO was used, between the IHS and Aio was applied the Thermal Grizzly Kryonaut TIM. The decreasing of the temperature running the the same benchmark in the same environment is in the orther of 29°C lower on the hotter core.
The comparison will be between 3 kind of IHS: -Intel Stock IHS 3,2mm tall -CNC made 4mm tall -CNC made 10mm tall (aka big boy )
The HW involved: -CPU: Intel i7 7700k -Motherboard: AsRock z170m OC FORMULA Motherboard: Z170m OC FORMULA -Cooling: Raijinteck Triton 360 AiO -Memory: Apacer DDR4 (stock speed) -Videocard: Nvidia GT 710 -PSU: Seasonic 1200w Platinum -TIM: Thermal Grizzly Kryonut between IHS and AiO/ Thermal Grizzly Conductonaut between cpu's DIE and IHS
For test the temperature i've chosen prime95 8k/8k mode (careful will kick your temperature sky high):
Cpu will run at 5000Mhz CPU/ 5000MHZ CACHE @ 1.35V
For monitoring the temperature and collecting data, will be used a Quad Channel Type-k Thermocouple Data Logger OMEGA HH374. CH1=IHS temp. CH2=AMB temp. CH3=Water temp. of Triton AiO
Picture of messy setup
So here we go, starting with stock intel IHS: 4mm tall CNC made IHS: 10mm tall CNC made IHS:
As you can see from screenshot, there's not at all difference between different IHS, with a costant load.
A surprising data, coming out from 6000 count (2000 each test) is this showed below on the graphic:
Seems strange but the taller IHS have the worse temperature, during the recording data period, the 4mm tall have a slightly better temperature then Intel's stock IHS.
For a non expert eyes like mine, seems cpu (DIE) don't realy care about wich IHS you put over it. This make more sense when you think about IHS like something necessary to trasfer the heat to your active cooler, in this case the AiO.
Still there's a difference who is impossible to see with constant load. Quick benchmark like cinebench R15, decreed a winner. With the impressive frequncy of 5400Mhz CPU/5200Mhz CACHE @ 1.52v and 16GB of DDR4 memory running 4133Mhz XMP, the only IHS who allowed me to successful complete the benchmark is the taller 10mm aka Big Boy, probably due to a much slower response when load start kick.
Last but not least, is the possibility to mount without IHS at all, means AiO straight at direct contact with CPU DIE. Despite the tool you can see in the picture, i was unable to make cpu work. The contact between DIE and AiO was good but the pressure to site CPU in the socket wasn't enough, due to the corner of the socket.
I've some cpu to resell, left from a very tight binning done for extreme overclock purpose. The idea is reselling left cpu offering some services with them, since the cpu are basicaly new, i'll try to offer a good deal for a fear price.
For give you an idea of the difference you can see the screenshoot here: -BEFORE DELID
HARDWARE INVOLVED: -Motherboard: Z170m OC FORMULA -Cooling: Raijinteck Triton 3x120 AiO -Memory: G.Skill Ripjaws V 3600mhz xmp -Videocard: Nvidia GT 710 -TIM: Thermal Grizzly Konductonaut between IHS and AiO
After Delid process, the cpu will be tested under stress for 30 minutes with prime95 in 1344k mode, for give to each cpu the proper value (binning): [This test is done with Cache and Memory at stock speed] CPU Type D = 4900Mhz CPU/4200Mhz CACHE @ 1,350V or less CPU Type C = 5000Mhz CPU/4200Mhz CACHE @ 1,375V or less CPU Type B = 5100Mhz CPU/4200Mhz CACHE @ 1,400V or less CPU Type A = 5200Mhz CPU/4200Mhz CACHE @ 1,450V or less
For each cpu i'll test also the maximum frequency possible for CPU and CACHE with memory at 3600mhz xmp profile, on Cinebench R15.
The price cover the shipping cost for Italy, within EU will be 10€ more to add at cpu price.
CPU is BOX version, 3 years direct waranty from Intel. I'll don't take care of your RMA if necesary. As waranty of the performance i'll give you a screenshot taken during the test with cpu ID on it. Still i can't waranty you can run on your system with same voltage as on my system. Frequencies achieved can be ±100MHz.