Review Index:
Feedback

Delidding your Intel Haswell CPU

Manufacturer: PC Perspective

The Vise Method: Step by Step

Tools Required

To remove the heat spreader from the CPU PCB, you will need a few tools that are common to most home work benches or can be easily obtained from your local Lowe's or Home Depot store.

  • a rubber mallet
  • a wooden 2 x 4 cut to a 6 inch length
  • a bench vise
  • electrical tape and/or duct tape
  • paper or cotton towels

Process and Procedure

The vise method definitely sounds like the more error prone approach out of the two methods discussed, but is actually much easier and safer than the razor blade method because of the reasons discussed. The key is to set aside time to do it and take your time through the process. This also means check, double check, and triple check everything before plunging in because a mistake in the process could cost you significant money.

View Full Size

First step is prepping the vise. You want to put one or more layers of electrical tape or duct Tape on both arms of the vise to better grip the CPU heat spreader as well as to offer some protection for the CPU. Then you want to line your bench and all around the vise with several layers of towels just in case the the CPU PCB goes flying during the process. The CPU die wouldn't like contacting a hard surface, just trust me on this one.

View Full Size

View Full Size

Next, you want to very carefully place your CPU in the vise and tighten until it is secure. Not too much though because you want to hold fast the heat spreader only, not crush it. Basically, you put the CPU in heat spreader down and lock it in so that the vise arms grid the heat spreader by the top portion of the spreader, gripping the vertical portion above the step in the spreader to keep the vice surface from directly contacting the CPU PCB surface. Make sure to place the CPU in the vice so that the sides of the spreader with the hold down "arms" are not gripped. This orientation is essential so that you don't risk banging the heat spreader into the circuitry along the sides of the CPU die.

View Full Size

View Full Size

Once the CPU is secured in the vise, place the wood block on the edge of the CPU PCB as shown holding it in place with your secondary hand. Then, gently tap the block several times into the CPU PCB until you feel the PCB begin to slide. At that point, you have delidded you CPU. Remove the block and carefully remove the CPU PCB from the vise.

View Full Size

If everything went according to plan, you should end up with the CPU in two parts - the CPU PCB and the heat spreader as shown in the picture. You will immediately notice the sub-par TIM on your 4770K Haswell also. The TIM can be carefully removed from the CPU die and the heat spreader using some Q-Tips and high-grade alcohol. Then you can hit it with contact cleaner to remove the left over TIM residue. You don't have to remove the RTV residue from the CPU PCB surface (the black goo around the outer edge of the PCB), but it can be removed by carefully scraping it off with a razor blade, alcohol or contact cleaner, and some elbow grease.

View Full Size

Once all cleaned up, the underside of your heat spreader and the CPU PCB should look something like the one pictured above, minus the die cracks. Cracking the die is one potential hazard of delidding the CPU, but we will go into exactly how that happened shortly. The actual vise-method delidding process did not cause that though, rest assured.


August 25, 2014 | 10:54 AM - Posted by Anonymous (not verified)

if you do go direct to die, if you are running and ek block you could use their adapter kit so that the mount is correct.

August 25, 2014 | 03:24 PM - Posted by Morry Teitelman

The MSI Die Guard *should* give the same amount of protection as the EK mounting system, theorectically.  But after my *adventures*, I don't think I'll be returning to the land of "direct-die cooling" (and I don't think my wife will let me either)...

November 20, 2015 | 04:14 PM - Posted by Lorenzitto (not verified)

Why?
You bought CPU with your wife money?

August 25, 2014 | 11:21 AM - Posted by Anonymous (not verified)

Vise?
Is that a sort of pun on it being the wiser method or do you not know how to spell 'vice'?

August 25, 2014 | 11:31 AM - Posted by Ryan Shrout

Or maybe....

http://www.harborfreight.com/hand-tools/vises-anvils.html

August 25, 2014 | 12:18 PM - Posted by Daisyahoy

seems harbor freight doesn't know how to spell it either. Or maybe......

August 25, 2014 | 02:31 PM - Posted by Anonymous (not verified)

Thats how you spell It for crying out loud!

August 25, 2014 | 02:47 PM - Posted by sixstringrick

The British spell it "vice" in America we spell it "vise".

GOOGLE HAS SPOKEN!!!!

August 25, 2014 | 09:09 PM - Posted by Morry Teitelman

Thanks for checking.  I did double check that before posting, but you never know...

August 25, 2014 | 02:37 PM - Posted by Anonymous (not verified)

http://www.merriam-webster.com/dictionary/vise
http://www.merriam-webster.com/dictionary/vice

Though, in the case of the Haswell CPU TIM, both may be acceptable depending upon interpretation. :P

August 25, 2014 | 11:28 AM - Posted by razor512

With the older AMD CPUs, the dies were much smaller, which allowed for easier and safer heat sink use without a heat spreader in place.

The larger dies, also cause issues for some users who only use a tiny drop of thermal compound which fails to property cover the die area.

August 26, 2014 | 12:02 PM - Posted by Dusty

That is not true, over time die size has remained similar, but transistor count has increased significantly. This means, that even though TDP has gone down, there is more heat being produced in a smaller area. Smaller transistors are also less resilient to heat. This is the need for a the IHS. Secondly, if contact between the CPU and the heatsink is good, the less TIM is used, and the better the transfer of heat will be. This is because heat transfer is a function of thickness of the material.

August 25, 2014 | 11:37 AM - Posted by Anonymous (not verified)

dont use the MSI guard, just exchange the TIM and put the old heatspreader back on! my 4770K runs just fine like this.

August 25, 2014 | 11:42 AM - Posted by Anonymous (not verified)

Sorry that happened to you, Morry. That is one strong reason why I don't bother with extreme overclocking, and thus never need to resort to such "in-depth" methods to boost overclocking potential. The benefits never seemed to outweigh the potential loss of a chip.

August 25, 2014 | 12:45 PM - Posted by Lord_Exodia (not verified)

What type of temps are people getting with this method using air or standard enclosed kits found in most stores? Also what type of overclocks are people getting with these impoved temps? None of this was addressed so I'm not even sure if it's worth the trouble.

August 25, 2014 | 02:35 PM - Posted by Anonymous (not verified)

Yeah, I was expecting some sort of comparison between temps/stability or something.

August 25, 2014 | 03:21 PM - Posted by Morry Teitelman

Unfortunately, my CPU die cracked before i could really get any in-depth analysis done on temps and overclocking.  However, I did see a 20C drop in temps initially (until that of which will not be spoken occurred :))

At the first incident, I was not really willing to possibly sacrifice another CPU to the CPU-gods...

August 26, 2014 | 09:38 AM - Posted by Lord_Exodia (not verified)

Understood Morry, 20c is nothing to scoff at. That will at the very least extend the life and stability of your CPU since intel's boost will likely react better to temps in the much healthier threshold. I'm curious if this will net someone the elusive 5Ghz + overclocks that were initially rumored but then denounced once the chips released and were reviewed.

Anyone have a good experience with this? What were your OC results? We can't expect Morry to possibly sacrifice another flagship chip :)

August 26, 2014 | 04:00 AM - Posted by Anonymous (not verified)

yes mine droped arround 20°C under water.

August 25, 2014 | 02:50 PM - Posted by sixstringrick

Sigh...if only Morry had the used MSI Die Guard

August 25, 2014 | 08:10 PM - Posted by willmore (not verified)

Do you ever get the impression that people don't even read the articles? Heck, at least look at the pretty pictures!!!

August 26, 2014 | 01:14 AM - Posted by razor512

It is absolutely amazing how well it worked! :)

If only all things worked that well; I could try to turn on my TV and have the stove leak gas instead.

August 25, 2014 | 04:51 PM - Posted by Anonymous (not verified)

If only that big x86 CPU monopoly had used better thermal interface material to begin with, no one would have to flip their lids, or crack their dies.

August 26, 2014 | 12:40 AM - Posted by Anonymous (not verified)

And the world would be a more boring place.

August 26, 2014 | 12:49 AM - Posted by Laststop (not verified)

Thank the gods haswell-e will be going back to soldered tim like my i7-980x. Finally upgrading my x58 system and I thought I was going to have to delid to be able to properly overclock the i7-5960x 8 core. Since 8 core haswell-e will only be 3.0ghz at stock getting the best OC potential out of this chip is very important as single thread performance will suffer without a healthy overclock.

Now that haswell-e is back to quality soldering as the tim delidding is no longer necessary. I know 8 core haswell-e is not going to hit 4.6-4.7ghz like the quad cores but I am hoping I can get it up to 4.2ghz at least like my i7-980x 6 core is clocked.

I'll be water cooling it. EKWB supremacy full nickel waterblock, Alphacool Nexxos 60mm thick Full Copper Triple 140mm Radiator (420mm total) with 6x 140mm noctua a15 fans in push pull (really need push/pull with 60mm thick rads), XSPC dual 5.25" bay reservoir with dual mcp655 pumps in serial with 1/2" ID tubing at pump speed 3(i use dual pumps in serial for my water cooling because I need the extra power running through my loop as it cools not only the cpu but full coverage of the motherboard as well as my videocard and using 2 pumps allows me to lower the pump speed on both to reduce noise. Thats also why i have a huge 420mm 140mm x 3 and triple thick 60mm radiator. Just 1 massive loop that cools everything. I'm not totally crazy with that huge a setup just to cool the cpu.

Basically if I can't get the 8 core haswell to 4.2ghz nobody is.

August 26, 2014 | 07:38 AM - Posted by Anonymous (not verified)

"Basically if I can't get the 8 core haswell to 4.2ghz nobody is."

Laststop is my heeeeeero. Just be sure to keep those radiator fans running quiet. You don't want the noise to leave the basement and anger your mom and dad.

August 26, 2014 | 09:40 AM - Posted by Lord_Exodia (not verified)

They'd better use the Tim fit only for the Gods on that $1000+ CPU!

August 26, 2014 | 12:27 PM - Posted by BBMan (not verified)

Ahhh- to be able to afford to trashing, I mean delidding, your CPU ....

August 28, 2014 | 09:45 AM - Posted by Dr_b_ (not verified)

Does intel bin the Haswell-E chips before packaging them?

Since its just a Xeon with bad cores that failed QA and were shutoff/cutoff from the rest of the die, they are using the same solder pack process as they would for dice that had all 12 functioning cores that would be sold as a Xeon, and that is likely happening regardless of the binning.

August 26, 2014 | 02:00 PM - Posted by PhoneyVirus

I was considering this approach when I got no more then 200MHz overclocked with the system locking up from BSOD. Changed my mind and said it's not even worth the headache.

PhoneyVirus
https://twitter.com/PhoneyVirus
https://phoneyvirus.wordpress.com/

August 27, 2014 | 08:27 AM - Posted by Berserkism

Sick to death of hearing the "sub par TIM" argument.

In fact, it really fucking annoys me.

The reduction in distance between the die and heatspreader, due to the removal of the glue, is the reason you get lower temps.
It has next to nothing to do with the quality of the TIM.

God damn it!

August 28, 2014 | 09:57 AM - Posted by Dr_b_

Since Haswell-E is just a Xeon with bad cores that failed QA and were shutoff/cutoff from the rest of the die and down marketed to consumers, they are using the same solder pack process as they would for dice that had all 12 functioning cores that would be sold as a Xeon, and that is likely happening regardless of the binning.

Even without OC, regular Haswell reaches very high temps under certain work loads, 70C+, although it still works.

September 7, 2014 | 08:59 PM - Posted by Ben (not verified)

Why doesn't intel just solder their dies to the spreader? There wouldn't even be a need to delid.

October 26, 2015 | 12:25 AM - Posted by Anonymous (not verified)

"should reassemble" > "should resemble"

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.