Overclock.net › Forums › Intel › Intel CPUs › Deliding a 4770K (Haswell). Improving temperatures and maximizing overclockablity.
New Posts  All Forums:Forum Nav:

Deliding a 4770K (Haswell). Improving temperatures and maximizing overclockablity. - Page 13

post #121 of 1134
As soon as I can find some sort of stability on OC attempts, I'm going to delid my 4770K.

If I'm understanding this process correctly and from what I've gathered.

1. Delid using either a thin razor or hammer and vice method.

2. Carefully clean off access black glue with finger nails (?) until you can't anymore.

3. Cover up the iVRMs with something (Nail Polish, Hot glue, Electrical tape, etc), please advise on the best method here.

4. Apply a good amount of CL Pro in between Core and CPU HS.

5. Carefully mount cpu back onto motherboard, place HS on top and apply the CPU clamp.

6. Apply a good amount of CL Ultra? on top of CPU HS and mount watercooling block back onto CPU HS.

7. Walla! -20c temps?


Any advice is helpful guys as this is going to be my first attempt at delidding.
post #122 of 1134
Thread Starter 
Quote:
Originally Posted by Naviblue View Post

As soon as I can find some sort of stability on OC attempts, I'm going to delid my 4770K.

If I'm understanding this process correctly and from what I've gathered.

1. Delid using either a thin razor or hammer and vice method.

2. Carefully clean off access black glue with finger nails (?) until you can't anymore.

3. Cover up the iVRMs with something (Nail Polish, Hot glue, Electrical tape, etc), please advise on the best method here.

4. Apply a good amount of CL Pro in between Core and CPU HS.

5. Carefully mount cpu back onto motherboard, place HS on top and apply the CPU clamp.

6. Apply a good amount of CL Ultra? on top of CPU HS and mount watercooling block back onto CPU HS.

7. Walla! -20c temps?


Any advice is helpful guys as this is going to be my first attempt at delidding.

Cleaning off the excess glue is not really needed. I didn't cover the VRMs before putting the IHS back on but using some electrical tape on them might be safer. You only need a tiny amount of Liquid Pro/Ultra to cover the whole thing. Make sure you spread it evenly.
No Bottlenecks
(21 items)
 
 
File Server
(10 items)
 
CPUMotherboardGraphicsRAM
6700K @ 4.8 GHz 1.35V Gigabyte Z170X Gaming 7 GTX 1080 Reference @ 2.05 / 11 GHz 32GB G.SKILL Ripjaws 3350 MHz 15-14-14-34-2T 1.35V 
Hard DriveHard DriveHard DriveOptical Drive
Intel 750 400GB PCIe WD Red 8TB NAS 10 * 6TB HGST Desktar NAS (RAID Z2) Go away 
CoolingCoolingCoolingCooling
EK-FB GA Z170X Monoblock EK-FC1080 GTX Waterblock/Backplate 2 * Dazmode Darkside LP360 MCP355 DDC @ 7V 
CoolingOSMonitorKeyboard
9 * Corsair SP120 Quiet Edition @ 7V 8.1 Pro 12-bit Asus VG27HE 27" 1080p @ 145Hz Corsair Vengeance K60 
PowerCaseMouseMouse Pad
SeaSonic SS-520FL2 - Fanless Fractal Design Define S Logitech G502 Proteus Core XTRAC PADS Ripper XXL 
Audio
Asus Xonar DSX 
CPUMotherboardGraphicsRAM
4770K @ 4.7 GHz 1.39V Maximus VI Gene MSI GTX 1070 Seahawk EK X @ 2.1/9.0 Ghz 16GB G.Skill Sniper 2600 MHz 11-14-14-14-2T 1.65V 
Hard DriveHard DriveOptical DriveCooling
Samsung 840 Pro 256GB WD Green 6TB I will end you Dazmode Darkside Maple Leaf Special Edition 
CoolingCoolingCoolingCooling
EK GTX 1070 Seahawk Feser TFC Admiral 240 Black Ice GTX Xtreme M92 6 * Silverstone SST-FN121-P 
CoolingOSMonitorMonitor
2 * Noctua NF-B9 Redux 7 Ultimate 64-bit Samsung U28D590D 28" 4K Overlord Tempest X270OC 27" 1440P 
KeyboardPowerCaseMouse
Corsair Vengeance K60 EVGA SuperNova 750 G2 NZXT Vulcan (With Attached Handle) Logitech G500 
Mouse Pad
XTRAC PADS Ripper XXL 
CPUMotherboardRAMHard Drive
E5-2670 Supermicro X9SRA 64GB Samsung ECC 1333 MHz Mushkin Ventura Pro 32GB (Boot) 
Hard DriveOptical DriveCoolingOS
10 * 6TB HGST Deskstar NAS (RAID Z2) I will harm you Corsair H60 FreeNAS 9.10 
PowerCase
Corsair RM650 (Passive) Fractal Design Define R5 
  hide details  
Reply
No Bottlenecks
(21 items)
 
 
File Server
(10 items)
 
CPUMotherboardGraphicsRAM
6700K @ 4.8 GHz 1.35V Gigabyte Z170X Gaming 7 GTX 1080 Reference @ 2.05 / 11 GHz 32GB G.SKILL Ripjaws 3350 MHz 15-14-14-34-2T 1.35V 
Hard DriveHard DriveHard DriveOptical Drive
Intel 750 400GB PCIe WD Red 8TB NAS 10 * 6TB HGST Desktar NAS (RAID Z2) Go away 
CoolingCoolingCoolingCooling
EK-FB GA Z170X Monoblock EK-FC1080 GTX Waterblock/Backplate 2 * Dazmode Darkside LP360 MCP355 DDC @ 7V 
CoolingOSMonitorKeyboard
9 * Corsair SP120 Quiet Edition @ 7V 8.1 Pro 12-bit Asus VG27HE 27" 1080p @ 145Hz Corsair Vengeance K60 
PowerCaseMouseMouse Pad
SeaSonic SS-520FL2 - Fanless Fractal Design Define S Logitech G502 Proteus Core XTRAC PADS Ripper XXL 
Audio
Asus Xonar DSX 
CPUMotherboardGraphicsRAM
4770K @ 4.7 GHz 1.39V Maximus VI Gene MSI GTX 1070 Seahawk EK X @ 2.1/9.0 Ghz 16GB G.Skill Sniper 2600 MHz 11-14-14-14-2T 1.65V 
Hard DriveHard DriveOptical DriveCooling
Samsung 840 Pro 256GB WD Green 6TB I will end you Dazmode Darkside Maple Leaf Special Edition 
CoolingCoolingCoolingCooling
EK GTX 1070 Seahawk Feser TFC Admiral 240 Black Ice GTX Xtreme M92 6 * Silverstone SST-FN121-P 
CoolingOSMonitorMonitor
2 * Noctua NF-B9 Redux 7 Ultimate 64-bit Samsung U28D590D 28" 4K Overlord Tempest X270OC 27" 1440P 
KeyboardPowerCaseMouse
Corsair Vengeance K60 EVGA SuperNova 750 G2 NZXT Vulcan (With Attached Handle) Logitech G500 
Mouse Pad
XTRAC PADS Ripper XXL 
CPUMotherboardRAMHard Drive
E5-2670 Supermicro X9SRA 64GB Samsung ECC 1333 MHz Mushkin Ventura Pro 32GB (Boot) 
Hard DriveOptical DriveCoolingOS
10 * 6TB HGST Deskstar NAS (RAID Z2) I will harm you Corsair H60 FreeNAS 9.10 
PowerCase
Corsair RM650 (Passive) Fractal Design Define R5 
  hide details  
Reply
post #123 of 1134
Quote:
Originally Posted by Naviblue View Post

As soon as I can find some sort of stability on OC attempts, I'm going to delid my 4770K.

If I'm understanding this process correctly and from what I've gathered.

1. Delid using either a thin razor or hammer and vice method.

2. Carefully clean off access black glue with finger nails (?) until you can't anymore.

3. Cover up the iVRMs with something (Nail Polish, Hot glue, Electrical tape, etc), please advise on the best method here.

4. Apply a good amount of CL Pro in between Core and CPU HS.

5. Carefully mount cpu back onto motherboard, place HS on top and apply the CPU clamp.

6. Apply a good amount of CL Ultra? on top of CPU HS and mount watercooling block back onto CPU HS.

7. Walla! -20c temps?


Any advice is helpful guys as this is going to be my first attempt at delidding.

Covering up the iVRMs has been suggested as a safety precaution but of the several deliddings I've seen, nobody has bothered to do it. You just have to make sure you apply the CLU properly since it's conductive and make sure it doesn't touch the iVRMS. Also, Using CLU between the heat spreader and heatsink isn't going to do anything for you other than make it a complete ***** to pull the heatsink back off. Stick with AS5 or some other high quality TIM for the heatsink.
post #124 of 1134
I've read that CL Ultra comes off rather easily no?

And isn't the black glue the reason we have to delid anyways? So getting all that off is a good idea no?
Edited by Naviblue - 6/6/13 at 10:49am
post #125 of 1134
Quote:
Originally Posted by Exolaris View Post

Stick with AS5 or some other high quality TIM for the heatsink.
AS5 is still considered high quality? I thought it was outperformed by many more modern TIMs like MX-2 and IC diamond.
post #126 of 1134
Quote:
Originally Posted by opt33 View Post

Quote:
Originally Posted by dr/owned View Post

There's no point in delidding Ivy/Haswell to re-solder the IHS back on. Direct die cooling will yield results superior to this, without risking a literal meltdown.

Not unless either soldering your waterblock to your die, or using liquid metal with thermal conductance high enough and bondline thin enough to counteract the loss in surface area (that is gained at 87w/mk through intels solder), prior to hitting thermal roadblock at critical low surface area stage of tim1..especially if using crappy end user tim (5-10w/mk). It is theoretically incorrect to even assume direct die is better cooling than a soldered IHS on a high power density cpu, and has been proven wrong experimentally on Xtreme if using paste on die, and more than once, there is more than one thread on subject. I told the guy doing it, he would likely get worse temps, he did it anyway, and got much worse temps.

With liquid metal it will be a close crapshoot. If you solder your waterblock to your die, then yes you will gain a few C.

The success of direct die depends on power density, surface area and dissipated power. GPUs have low power density, so direct die works well. CPUs have high power density, hence heat must be spread to much larger area AT A HIGH HEAT CONDUCTANCE, prior to running into any serious heat roadblocks like relative crappy end user tim like PK1. Intels solder is 87 Wmk, PK1 is 10x lower. Liquid metal is only 4x lower. PK1 will yield worse results, liquid metal may be close.

Myth 1). removing a layer like IHS is always beneficial. (if this were true, then also beneficial to remove your heatsink fins from air cooler, as all that surface area is just in your way...much better to just blow air on IHS. This does not work, because neither air nor water (and water is eventually cooled via air, ie rad) can cool without massive surface area, because thermal conductance of water is paltry 0.6 w/mk, and air is much worse than water. And heat conductance in expanding that surface area is king.

Truth, 1) removing a layer is beneficial if you replace the TIM with exact same thermal conductance (87 w/mk for intels solder). For example, removing a soldered IHS and soldering waterblock to die, yes then removing a soldered IHS will net a few C.

Solder: intels die is copper laden silicon with heat conductance of 180 w/Mk. Then hits solder (~150um thick) at 87 W/mk. Then hits IHS at 400 w/mk. But surface is MANY times higher from die to IHS AT A CRITICALLY HIGH SPEED OF HEAT CONDUCTANCE, before suffering slow speed of piss poor tim of paste 5-10 w/mk, liquid metal little better at 20 w/mk. Which is why direct die cooling using tim paste like mx2 or pk1, doesnt work as well as solder, but liquid metal even though 4x lower heat conductance, has a THINNER bondline, if you remove all the adhesive, and may get close result depending on how thin/lack of voids the liquid metal is applied..

This same principle is what cause the relatively poor results of the direct to IHS water cooling...removing bottom of water block did remove a layer of copper, but in so doing, removed critical surface area of pins at critical stage, and ended up with worse results.

Removing solder IHS and direct die cooling with paste tim = failure
Removing solder IHS and direct die cooling with liquid metal = you might see similar temps, but solder (per intel white paper) will have lower spread between temps...unless you manage to avoid any voids...in end that would be a crapshoot, and just an exercise in futility.

Removing paste (5 w/mk) IHS that has a ~150 bondline thickness like intel ivy/haswell cpu, and replacing with direct die using paste with thinner bondline will yield better temps, direct die or replacing IHS, because replacing critical TIM 1 interface at lower surface area with same thermal conductance or better AND decreasing bondline, ie win/win. Using liquid metal, now your increasing thermal conductance 4x, decreasing interface resistance (metal/metal lower than tim/metal), and decreasing bondline thickness in addition to removing a layer, ie 4 improvements 0 negatives.

It however does not follow that removing a solder IHS is beneficial, totally different animal.

I had a thought experiment on this last night, and agree with your conclusion. Solder + IHS will probably be about the same as CLU + direct die. Solder + IHS has the advantage of being applied thinner and more consistently with better thermal conductivity (~80 W/m K). CLU + direct die has the advantage of being closer to the die and not restricted by a copper IHS layer, but CLU will be thicker than solder and have 1/3 the thermal conductivity (~ 30 W/m K).

I'm going to look into the cost of having my waterblock base professionally lapped to be flat and mirror finish. Since we already know the silicon is flat, a matched waterblock would negate the need for any TIM at all.
post #127 of 1134
in one of the videos, he seemed to indicate that the surface mount components on the pcb were the VRM's. if this is the case, I wonder how much heat they generate and would it be a possibility to use a thermal pad to help transfer some of that heat to the IHS?
HAL
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel I7 7700k, 5.2GHz @ 1.44v Asus Maximus IX Apex 2x EVGA gtx 970 SC acx 2.0  G.SKILL TridentZ RGB Series 16GB (2 x 8GB) 288... 
Hard DriveCoolingCoolingOS
2 X 480GB Samsung 840 EVO (raid 0) CPU: MCP655, XSPC 360 Rad, MIPS Iceforce HF GPUs: MCP655, external 360 p/p, EK 970 acx full... Windows 10 
MonitorPowerCase
asus 24" 144Hz Corsair AX 1200i Corsair 900D 
  hide details  
Reply
HAL
(15 items)
 
  
CPUMotherboardGraphicsRAM
Intel I7 7700k, 5.2GHz @ 1.44v Asus Maximus IX Apex 2x EVGA gtx 970 SC acx 2.0  G.SKILL TridentZ RGB Series 16GB (2 x 8GB) 288... 
Hard DriveCoolingCoolingOS
2 X 480GB Samsung 840 EVO (raid 0) CPU: MCP655, XSPC 360 Rad, MIPS Iceforce HF GPUs: MCP655, external 360 p/p, EK 970 acx full... Windows 10 
MonitorPowerCase
asus 24" 144Hz Corsair AX 1200i Corsair 900D 
  hide details  
Reply
post #128 of 1134
Quote:
Originally Posted by mystiksinner View Post

in one of the videos, he seemed to indicate that the surface mount components on the pcb were the VRM's. if this is the case, I wonder how much heat they generate and would it be a possibility to use a thermal pad to help transfer some of that heat to the IHS?


I don't think they are.  They look like very small surface mount capacitors to me.

post #129 of 1134
They're probably the capacitors needed for the integrated voltage regulator. The inductors are in the substrate, but the caps have to be separate.
post #130 of 1134
Great post!

I'm very curious to see what kind of numbers you get after doing this... my haswell chip arrives tomorrow, so I may delid the thing right off the bat.
Nerdgasm
(19 items)
 
  
CPUMotherboardGraphicsGraphics
Intel Processor Z87X-OC Force-CF EVGA NVIDIA GeForce GTX 780 Classified Hydro Co... EVGA NVIDIA GeForce GTX 780 Classified Hydro Co... 
GraphicsRAMHard DriveHard Drive
EVGA NVIDIA GeForce GTX 780 Classified Hydro Co... Corsair Vengence Mushkin Enhanced MKNSSDCR240GB-DX7 2.5" 240GB S... Seagate STBD4000400 
CoolingCoolingCoolingCooling
XSPC AX360 Radiator XSPC Twin D5 Dual 5.25" Bay Clear Reservoir XSPC Raystorm EK GIGABYTE GA-Z87X-OC Force Full Board Cooling... 
CoolingOSMonitorMonitor
Black Ice GT Xtreme 140 Radiator Windows 7 64 Overlord Computers Tempest X270OC @120hz Overlord Computers Tempest X270OC @120hz 
PowerCaseMouse
Corsair AX1200 NZXT Phantom 820 Razer Mamba 
  hide details  
Reply
Nerdgasm
(19 items)
 
  
CPUMotherboardGraphicsGraphics
Intel Processor Z87X-OC Force-CF EVGA NVIDIA GeForce GTX 780 Classified Hydro Co... EVGA NVIDIA GeForce GTX 780 Classified Hydro Co... 
GraphicsRAMHard DriveHard Drive
EVGA NVIDIA GeForce GTX 780 Classified Hydro Co... Corsair Vengence Mushkin Enhanced MKNSSDCR240GB-DX7 2.5" 240GB S... Seagate STBD4000400 
CoolingCoolingCoolingCooling
XSPC AX360 Radiator XSPC Twin D5 Dual 5.25" Bay Clear Reservoir XSPC Raystorm EK GIGABYTE GA-Z87X-OC Force Full Board Cooling... 
CoolingOSMonitorMonitor
Black Ice GT Xtreme 140 Radiator Windows 7 64 Overlord Computers Tempest X270OC @120hz Overlord Computers Tempest X270OC @120hz 
PowerCaseMouse
Corsair AX1200 NZXT Phantom 820 Razer Mamba 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Deliding a 4770K (Haswell). Improving temperatures and maximizing overclockablity.